Oracle Databases on ONTAP - tele.ga: export, import and stat your

Technical Report
Oracle Databases on ONTAP
Jeffrey Steiner, NetApp
July 2017 | TR-3633
Important
Consult the Interoperability Matrix Tool (IMT) to determine whether the environment,
configurations, and versions specified in this report support your environment.
TABLE OF CONTENTS
1
Introduction ........................................................................................................................................... 6
2
ONTAP Platforms.................................................................................................................................. 6
3
4
5
6
7
2
2.1
ONTAP with AFF and FAS Controllers ...........................................................................................................6
2.2
NPS for Cloud .................................................................................................................................................7
2.3
ONTAP Select.................................................................................................................................................7
2.4
ONTAP Cloud .................................................................................................................................................8
ONTAP Configuration........................................................................................................................... 8
3.1
RAID Levels ....................................................................................................................................................8
3.2
Capacity Limits................................................................................................................................................9
3.3
Snapshot-Based Backups ...............................................................................................................................9
3.4
Snapshot-Based Recovery ...........................................................................................................................10
3.5
Snapshot Reserve ........................................................................................................................................11
3.6
read_realloc ..................................................................................................................................................11
3.7
ONTAP and Third-Party Snapshots ..............................................................................................................12
3.8
Cluster Operations—Takeover and Switchover ............................................................................................12
Storage Virtual Machines and Logical Interfaces ........................................................................... 14
4.1
SVMs ............................................................................................................................................................14
4.2
LIF Types ......................................................................................................................................................15
4.3
SAN LIF Design ............................................................................................................................................15
4.4
NFS LIF Design ............................................................................................................................................16
Quality of Service ............................................................................................................................... 18
5.1
IOPS QoS .....................................................................................................................................................19
5.2
Bandwidth QoS .............................................................................................................................................19
5.3
Guaranteed QoS ...........................................................................................................................................19
Compression, Compaction, and Deduplication .............................................................................. 19
6.1
Compression .................................................................................................................................................19
6.2
Inline Data Compaction .................................................................................................................................22
6.3
Deduplication ................................................................................................................................................23
Thin Provisioning ............................................................................................................................... 23
7.1
Space Management ......................................................................................................................................23
7.2
LUN Thin Provisioning ..................................................................................................................................24
7.3
Fractional Reservations ................................................................................................................................24
7.4
Compression and Deduplication ...................................................................................................................24
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
7.5
8
9
Compression and Fractional Reservations ...................................................................................................25
Performance Optimization and Benchmarking ............................................................................... 25
8.1
Oracle Automatic Workload Repository and Benchmarking .........................................................................26
8.2
Oracle AWR and Troubleshooting ................................................................................................................26
8.3
calibrate_io....................................................................................................................................................27
8.4
SLOB2 ..........................................................................................................................................................27
8.5
Swingbench ..................................................................................................................................................27
8.6
HammerDB ...................................................................................................................................................27
8.7
Orion .............................................................................................................................................................27
General Oracle Configuration ........................................................................................................... 28
9.1
filesystemio_options ......................................................................................................................................28
9.2
db_file_multiblock_read_count......................................................................................................................29
9.3
Redo Block Size............................................................................................................................................29
9.4
Checksums and Data Integrity ......................................................................................................................29
10 Flash .................................................................................................................................................... 30
10.1 Flash Cache ..................................................................................................................................................30
10.2 SSD Aggregates ...........................................................................................................................................31
10.3 Flash Pool .....................................................................................................................................................32
10.4 AFF Platforms ...............................................................................................................................................33
11 Ethernet Configuration....................................................................................................................... 33
11.1 Ethernet Flow Control ...................................................................................................................................33
11.2 Jumbo Frames ..............................................................................................................................................34
11.3 TCP Parameters ...........................................................................................................................................34
12 General NFS Configuration ............................................................................................................... 35
12.1 NFS Versions ................................................................................................................................................35
12.2 TCP Slot Tables ............................................................................................................................................35
12.3 Installation and Patching ...............................................................................................................................35
12.4 ONTAP and NFS Flow Control .....................................................................................................................35
12.5 Direct NFS ....................................................................................................................................................36
12.6 Direct NFS and Host File System Access .....................................................................................................36
12.7 ADR and NFS ...............................................................................................................................................37
13 General SAN Configuration ............................................................................................................... 37
13.1 Zoning ...........................................................................................................................................................37
13.2 LUN Alignment ..............................................................................................................................................37
3
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
13.3 LUN Misalignment Warnings.........................................................................................................................38
13.4 LUN Sizing ....................................................................................................................................................38
13.5 LUN Resizing and LVM-Based Resizing .......................................................................................................38
13.6 LUN Count ....................................................................................................................................................39
13.7 Datafile Block Size ........................................................................................................................................40
13.8 Redo Block Size............................................................................................................................................40
14 Virtualization ....................................................................................................................................... 40
14.1 Overview .......................................................................................................................................................40
14.2 Storage Presentation ....................................................................................................................................40
14.3 Paravirtualized Drivers ..................................................................................................................................41
14.4 Overcommitting RAM ....................................................................................................................................42
15 Clustering ............................................................................................................................................ 42
15.1 Oracle Real Application Clusters ..................................................................................................................42
15.2 Solaris Clusters .............................................................................................................................................43
15.3 Veritas Cluster Server ...................................................................................................................................43
16 IBM AIX ................................................................................................................................................ 44
16.1 Concurrent I/O ..............................................................................................................................................44
16.2 AIX NFSv3 Mount Options ............................................................................................................................45
16.3 AIX jfs/jfs2 Mount Options .............................................................................................................................46
17 HP-UX ................................................................................................................................................... 46
17.1 HP-UX NFSv3 Mount Options.......................................................................................................................46
17.2 HP-UX VxFS Mount Options .........................................................................................................................47
18 Linux .................................................................................................................................................... 47
18.1 Linux NFS .....................................................................................................................................................47
18.2 Linux NFSv3 Mount Options .........................................................................................................................48
18.3 General Linux SAN Configuration .................................................................................................................50
18.4 ASM Mirroring ...............................................................................................................................................51
18.5 ASMlib Block Sizes .......................................................................................................................................51
18.6 Linux ext3 and ext4 Mount Options ...............................................................................................................52
19 Microsoft Windows ............................................................................................................................. 52
19.1 NFS ...............................................................................................................................................................52
19.2 SAN ..............................................................................................................................................................52
20 Solaris .................................................................................................................................................. 52
20.1 Solaris NFSv3 Mount Options .......................................................................................................................52
4
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
20.2 Solaris UFS Mount Options...........................................................................................................................54
20.3 Solaris ZFS ...................................................................................................................................................54
21 Conclusion .......................................................................................................................................... 57
Appendix A: Stale NFS Locks ................................................................................................................. 57
Appendix B: WAFL Alignment Verification ............................................................................................ 58
Aligned ................................................................................................................................................................. 58
Misaligned ............................................................................................................................................................ 60
Redo Logging ....................................................................................................................................................... 60
LIST OF TABLES
Table 1) AIX NFSv3 mount options—single instance. ..................................................................................................45
Table 2) AIX NFSv3 mount options—RAC. ..................................................................................................................45
Table 3) AIX jfs/jfs2 mount options—single instance. ...................................................................................................46
Table 4) HP-UX NFSv3 mount options—single instance. ............................................................................................46
Table 5) HP-UX NFSv3 mount options—RAC..............................................................................................................46
Table 6) Linux NFSv3 mount options—single instance. ...............................................................................................48
Table 7) Linux NFSv3 mount options—RAC. ...............................................................................................................48
Table 8) Solaris NFSv3 mount options—single instance. .............................................................................................53
Table 9) Solaris NFSv3 mount options—RAC. .............................................................................................................53
5
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
1 Introduction
NetApp® ONTAP® is a powerful data-management platform with native capabilities that include inline
compression, nondisruptive hardware upgrades, and the ability to import a LUN from a foreign storage
array. Up to 24 nodes can be clustered together, simultaneously serving data through Network File
System (NFS), Common Internet File System (CIFS), iSCSI, Fibre Channel (FC), and Fibre Channel over
Ethernet (FCoE) protocols. In addition, NetApp Snapshot® technology is the basis for creating tens of
thousands of online backups and fully operational database clones.
In addition to the rich feature set of ONTAP, there is a wide variety of user requirements, including
database size, performance requirements, and data protection needs. Known deployments of NetApp
storage include everything from a virtualized environment of approximately 6,000 databases running
under VMware ESX to a single-instance data warehouse currently sized at 996TB and growing. As a
result, there are few clear best practices for configuring an Oracle database on NetApp storage.
This document addresses the requirements for operating an Oracle database on NetApp storage in two
ways. First, when a clear best practice exists, it is called out specifically. Second, this document reviews
the many design considerations that must be addressed by architects of Oracle storage solutions based
on their specific business requirements.
This document first discusses general considerations for all environments followed by specific
recommendations based on the choice of virtualization and OS. Special topics such as the choice of file
system layout and NFS lock breaking are included in the appendixes.
For more details, see the following additional resources:
•
TR-4591: Database Data Protection
•
TR-4592: Oracle on MetroCluster
•
TR-4534: Migration of Oracle Databases to NetApp Storage Systems
2 ONTAP Platforms
ONTAP software is the foundation for advanced data protection and management. However, ONTAP only
refers to software. There are several ONTAP hardware platforms to choose from:
•
ONTAP on All Flash FAS (AFF) and FAS
•
NetApp Private Storage (NPS) for Cloud
•
ONTAP Select
•
ONTAP Cloud
The key concept is that ONTAP is ONTAP. Some hardware options offer better performance, others offer
lower costs, and some run within hyperscaler clouds. The core functions of ONTAP are unchanged, with
multiple replication options available to bind different ONTAP platforms into a single solution. As a result,
data protection and disaster recovery strategies can be built on real-world needs, such as performance
requirements, capex/opex considerations, and overall cloud strategy. The underlying storage technology
runs anywhere in any environment.
2.1
ONTAP with AFF and FAS Controllers
For maximum performance and control of data, ONTAP on a physical AFF or FAS controller remains the
leading solution. This is the standard option that thousands of customers have relied upon for more than
20 years. ONTAP delivers solutions for any environment, ranging from three mission-critical databases to
60,000-database service provider deployments, instant restores of petabyte-scale databases, and DBaaS
involving hundreds of clones of a single database.
6
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
2.2
NPS for Cloud
NetApp introduced the NPS option to address the needs of data-intensive workloads in the public cloud.
Although many public cloud storage options exist, most of them have limitations in terms of performance,
control, or scale. With respect to database workloads, the primary limitations are as follows:
•
Many public cloud storage options do not scale to the IOPS levels required by modern database
workloads in terms of cost, efficiency, or manageability.
•
Even when the raw IOPS capabilities of a public cloud provider meet requirements, the I/O latencies
are frequently unacceptable for database workloads. This has become even more true as databases
have migrated to all-flash storage arrays, and businesses have begun to measure latency in terms of
microseconds, not milliseconds.
•
Although public cloud storage availability is good overall, it does not yet meet the demands of most
mission-critical environments.
•
Backup and recovery capabilities exist within public cloud storage services, but they generally cannot
meet the zero RPO and near-zero RTO requirements of most databases. Data protection requires
true instant snapshot-based backup and recovery, not streaming backup and recovery to and from
elsewhere in a cloud.
•
Hybrid cloud environments must move data between on-premises and cloud storage systems,
mandating a common foundation for storage management.
•
Many governments have strict data sovereignty laws that prohibit relocating data outside national
borders.
NPS systems deliver maximum storage performance, control, and flexibility to public cloud providers,
including Amazon AWS, Microsoft Azure, and IBM SoftLayer. This capability is delivered by AFF and FAS
systems, including MetroCluster options, in data centers connected directly to public clouds. The full
power of the hyperscaler compute layer can be used without the limitations of hyperscaler storage.
Furthermore, NPS enables cloud-independent and multicloud architectures because the data, such as
application binaries, databases, database backups, and archives, all remain wholly within the NPS
system. There is no need to expend time, bandwidth, or money moving data between cloud providers.
Notably, some NetApp customers have used the NPS model on their own initiative. In many locations,
high-speed access to one of the hyperscaler providers is readily available to customer data center
facilities. In other cases, customers use a colocation facility that is already capable of providing highspeed access to hyperscaler cloud providers. This had led to the use of Amazon AWS, Azure, and
SoftLayer as essentially on-demand, consumption-based sources of virtualized servers. In some cases,
nothing has changed about the customers' day-to-day operations. They simply use the hyperscaler
services as a more powerful, flexible, and cost-efficient replacement for their traditional virtualization
infrastructure.
Options are also available for NPS as a service (NPSaaS). In many cases, the demands of database
environments are substantial enough to warrant purchasing an NPS system at a colocation facility.
However, in some cases, customers prefer to utilize both cloud servers and cloud storage as an
operational expense rather than a capital expense. In these cases, they want to use storage resources
purely as an as-needed, on-demand service. Several providers now offer NPS as a service for such
customers.
2.3
ONTAP Select
ONTAP Select runs on a customer’s own virtualization infrastructure and delivers ONTAP intelligence and
data fabric connectivity to the drives inside of white box hardware. ONTAP Select allows ONTAP and
guest operating systems to share the same physical hardware as a highly-converged infrastructure. The
best practices for running Oracle on ONTAP are not affected. The primary consideration is performance,
but ONTAP Select should not be underestimated.
7
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
An ONTAP Select environment does not match the peak performance of a high-end AFF system, but
most databases do not require 300K IOPS. Typical databases only require around 5K to 10K IOPS, a
target that can be met by ONTAP Select. Furthermore, most databases are limited more by storage
latency than storage IOPS, a problem that can be addressed by deploying ONTAP Select on SSD drives.
2.4
ONTAP Cloud
ONTAP Cloud is similar to ONTAP Select, except that it runs in a hyperscaler Cloud environment,
bringing intelligence and data fabric connectivity to hyperscaler storage volumes. The best practices for
running Oracle on ONTAP are not affected. The primary considerations are performance and to a lesser
extent cost.
ONTAP Cloud is partially limited by the performance of the underlying volumes managed by the cloud
provider. The result is more manageable storage, and, in some cases, the caching capability of ONTAP
Cloud offers a performance improvement. However, there are always some limitations in terms of IOPS
and latency due to the reliance on the public cloud provider. This does not mean that database
performance is unacceptable. It simply means that the performance ceiling is lower than options such as
an actual physical AFF system. Furthermore, the performance of storage volumes offered by the various
cloud providers that are utilized by ONTAP Cloud are continuously improving.
The prime use case for ONTAP Cloud is currently for development and testing work, but some customers
have used ONTAP Cloud for production activity as well. One particularly notable report was the use of
Oracle’s In-Memory feature to mitigate storage performance limitations. This allows more data to be
stored in RAM on the virtual machine hosting the database server, thus reducing performance demands
on storage.
3 ONTAP Configuration
A complete description of the configuration of the ONTAP OS is beyond the scope of this document. A
best practice for an environment with 2,000 virtualized databases might be inappropriate for a
configuration of three very large enterprise resource planning databases. Even small changes in data
protection requirements can significantly affect storage design. Some basic details are reviewed in this
section. A more complete explanation can be found in TR-4591. For comprehensive assistance with
design, contact NetApp or a NetApp partner.
3.1
RAID Levels
Questions occasionally arise concerning RAID levels in the configuration of NetApp storage. Many older
Oracle documents and books on Oracle configuration contain warnings about using RAID mirroring
and/or avoiding certain types of RAID. Although they raise valid points, these materials do not apply to
RAID 4 and the NetApp RAID DP® and RAID-TEC™ technologies used in ONTAP.
RAID 4, RAID 5, RAID 6, RAID DP, and RAID-TEC all leverage parity so that data is not lost because of a
drive failure. These RAID options offer much better storage efficiency in comparison to mirroring, but most
RAID implementations have a drawback that affects write operations. Completion of a write operation on
other RAID implementations requires multiple disk reads to regenerate the parity data, a process
commonly called the RAID penalty.
Leveraging ONTAP, however, does not incur a RAID penalty. This is because of the integration of
NetApp WAFL® (Write Anywhere File Layout) with the RAID layer. Write operations are coalesced in RAM
and prepared as a complete RAID stripe, including parity generation. There is no need to perform a read
in order to complete a write, which means that ONTAP and WAFL avoid the RAID penalty. Performance
for latency-critical operations, such as redo logging, is unimpeded, and random data-file writes do not
incur any RAID penalty resulting from a need to regenerate parity.
8
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
With respect to statistical reliability, even RAID DP offers better protection than RAID mirroring. The
primary problem is the demand made on disks during a RAID rebuild. With a mirrored RAID set, the risk
of data loss from a disk failing while rebuilding to its partner in the RAID set is much greater than the risk
of a triple-disk failure in a RAID DP set.
3.2
Capacity Limits
In order to provide high and predictable performance on a storage array, some free space is required for
metadata and data organizational tasks. Free space is defined as any space that is not used for actual
data, and includes unallocated space on the aggregate itself and unused space within the constituent
volumes. Thin provisioning must also be considered. For example, a volume might contain a 1TB LUN of
which only 50% is utilized by actual data. In a thin provisioned environment, this would correctly appear to
be consuming 500GB of space. However, in a fully provisioned environment, the full capacity of 1TB will
appear to be in use. The 500GB of unallocated space will be hidden. This space is unused by actual data
and should therefore be included in the calculation of total free space.
NetApp recommendations for storage systems used for databases are described in the sections that
follow.
SSD Aggregates, Including AFF Systems
NetApp recommends a minimum of 10% free space. This includes all unused space, including free space
within the aggregate or a volume and any free space that is allocated due to the use of full provisioning
but is not used by actual data.
The recommendation of 10% free space is conservative. SSD aggregates can support database
workloads at even higher levels of utilization without any effect on performance. However, as the
utilization of the aggregate increases the risk of running out of space also increases if utilization is not
monitored carefully.
HDD Aggregates, Including Flash Pool Aggregates
NetApp recommends a minimum of 15% free space. This includes all unused space, including free space
within the aggregate or a volume and any free space that is allocated due to the use of full provisioning
but is not used by actual data.
There should be no measurable performance effect when utilization is less than 85%. As utilization
approaches 90%, some reduction in performance might become noticeable for certain workloads. As
utilization reaches 95%, most database workloads experience a degradation in performance.
3.3
Snapshot-Based Backups
The most important consideration for a file system layout is the plan for leveraging NetApp Snapshot
technology. There are two primary approaches:
•
Crash-consistent backups
•
Snapshot-protected hot backups
A crash-consistent backup of a database requires the capture of the entire database structure, including
datafiles, redo logs, and control files, at a single point in time. If the database is stored in a single NetApp
FlexVol® flexible volume, then the process is simple; a Snapshot can be created at any time. If a
database spans volumes, a consistency group (CG) Snapshot copy must be created. Several options
exist for creating CG Snapshot copies, including NetApp SnapCenter® software, the NetApp Snap
Creator® framework, NetApp SnapManager® for Oracle (SMO), NetApp SnapDrive® for UNIX, and usermaintained scripts.
9
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
Crash-consistent Snapshot backups are primarily used when point-of-the-backup recovery is sufficient.
Archive logs can be applied under some circumstances, but when more granular point-in-time recovery is
required, a hot backup is preferable.
The basic procedure for a snapshot-based hot backup is as follows:
1. Place the database in backup mode.
2. Create a Snapshot copy of all volumes hosting datafiles.
3. Exit backup mode.
4. Run the command alter system archive log current to force log archiving.
5. Create Snapshot copies of all volumes hosting the archive logs.
This procedure yields a set of Snapshot copies containing datafiles in backup mode and the critical
archive logs generated while in backup mode. These are the two requirements for recovering a database.
Files such as control files should also be protected for convenience, but the only absolute requirement is
protection for datafiles and archive logs.
Although different customers might have very different strategies, almost all of these strategies are
ultimately based on the principles outlined in this section.
3.4
Snapshot-Based Recovery
When designing volume layouts for Oracle databases, the first decision is whether to use volume-based
NetApp SnapRestore® (VBSR) technology.
Volume-based SnapRestore allows a volume to be almost instantly reverted to an earlier point in time.
Because all of the data on the volume is reverted, VBSR might not be appropriate for all use cases. For
example, if an entire database, including datafiles, redo logs, and archive logs, is stored on a single
volume and this volume is restored with VBSR, then data is lost because the newer archive log and redo
data are discarded.
VBSR is not required for restore. Many databases can be restored by using file-based single-file
SnapRestore (SFSR) or by simply copying files from the Snapshot copy back into the active file system.
VBSR is preferred when a database is very large or when it must be recovered as quickly as possible,
and the use of VBSR requires isolation of the datafiles. In an NFS environment, the datafiles of a given
database must be stored in dedicated volumes that are uncontaminated by any other type of file. In a
SAN environment, datafiles must be stored in dedicated LUNs on dedicated FlexVol volumes. If a volume
manager is used (including Oracle Automatic Storage Management [ASM]), the disk group must also be
dedicated to datafiles.
Isolating datafiles in this manner allows them to be reverted to an earlier state without damaging other file
systems.
Enhancements in ONTAP 8.2
A significant enhancement to restore capabilities was added in ONTAP 8.2. Previous versions of ONTAP
could only create file-level clones from the active file system. With 8.2, it is now possible to create a filelevel clone from a Snapshot copy. As a result, it is now much easier to use a file system layout that
includes multiple database file types in a single volume or even multiple databases in a single volume.
Prior to version 8.2, a 10TB database would have likely required the isolation of its datafiles in a
dedicated volume to deliver an acceptably fast restore procedure. Foreign files in the volume would have
precluded the use of VBSR because those foreign files would be destroyed by the restoration process.
Recovery would have been performed by copying data instead. This process can still be very fast when
managed by a product such as SnapManager for Oracle (SMO), which can invoke an internal copy
operation within the array. However, it is still not as fast as VBSR.
10
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
With ONTAP 8.2, files and LUNs can be cloned directly from a Snapshot copy. This is a nearly
instantaneous process that preserves space efficiency. Therefore, VBSR is no longer required for rapid
recovery of large databases, and multiple databases can share a common volume.
In LUN-based environments, store datafiles in dedicated disk groups and LUNs so that they can be
restored as a unit. If other files are stored within a datafile disk group, cloning from a Snapshot copy
cannot be used for rapid recovery because doing so destroys these foreign files.
Note:
3.5
When a file is restored by cloning from a Snapshot copy, a background process updates any
metadata. There should be no effect on performance, but this process blocks the creation of
Snapshot copies until it is complete. The processing rate is approximately 5GBps (18TB/hour)
based on the total size of the files restored.
Snapshot Reserve
For each volume with Oracle data in a SAN environment, the percent-snapshot-space should be set
to zero because reserving space for a Snapshot copy in a LUN environment is not useful. If the fractional
reserve is set to 100, a Snapshot copy of a volume with LUNs requires enough free space in the volume,
excluding the snapshot reserve, to absorb 100% turnover of all of the data. If the fractional reserve is set
to a lower value, then a correspondingly smaller amount of free space is required, but it always excludes
the Snapshot copy reserve. This means that the Snapshot copy reserve space in a LUN environment is
wasted.
In an NFS environment, there are two options:
•
Set the percent-snapshot-space based on expected Snapshot copy space consumption.
•
Set the percent-snapshot-space to zero and manage active and Snapshot copy space
consumption collectively.
With the first option, percent-snapshot-space is set to a nonzero value, typically around 20%. This
space is then hidden from the user. This value does not, however, create a limit on utilization. If a
database with a 20% reservation experiences 30% turnover, the Snapshot copy space can grow beyond
the bounds of the 20% reserve and occupy unreserved space.
The main benefit of setting a reserve to a value such as 20% is to verify that some space is always
available for Snapshot copies. For example, a 1TB volume with a 20% reserve would only permit a
database administrator (DBA) to store 800GB of data. This configuration guarantees at least 200GB of
space for Snapshot copy consumption.
When percent-snapshot-space is set to zero, all space in the volume is available to the end user,
which delivers better visibility. A DBA must understand that, if he or she sees a 1TB volume that
leverages Snapshot copies, this 1TB of space is shared between active data and Snapshot turnover.
There is no clear preference between option one and option two among end users.
3.6
read_realloc
Most write activity to an Oracle datafile consists of random overwrites. As these overwrites occur, the
changed data is placed on a new physical location within the storage system. This action has no effect on
random I/O, which is typically the most performance-critical I/O type. However, it can affect sequential I/O
throughput because the storage system is forced to perform more physical disk I/Os to assemble the
response to a multiblock read request and to perform readahead.
On an AFF system, the additional I/Os are not significant, but on an array with spinning media, including
Flash Pool aggregates, the additional drive head movement results in increased latency and in turn
lowers throughput.
Enabling read_realloc on a volume results in real-time optimization of the file system layout. When the
data on a WAFL volume is poorly allocated, the bulk of the work required to address the problem is read
11
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
activity. After the block reads are complete, writing the data back to disk in a single contiguous RAID
stripe is a low-cost operation. The read_realloc option enables this process in a way that does not
affect overall performance.
For example, if a full table scan is being performed, meaning that a datafile is being read sequentially,
and read_realloc detects blocks that were suboptimally organized on the disk, then 90% of the work
to address the problem is already complete. The blocks are already in RAM on the storage system.
Therefore, after servicing the read request from the database server, read_realloc performs the next
step and writes them back to disk in an optimized format. The next time a full table scan is performed, the
data is optimized. In the long term, the use of read_realloc creates a constant data cleanup process
that optimizes the layout of the datafiles on a disk.
There are two read_realloc methods: the general on and space_optimized. The general setting
optimizes block layout for both the live file system and the blocks contained in a Snapshot copy. This can
result in increased space consumption when Snapshot copies are present, but with the benefit of
improved performance during sequential reads on the live file system, Snapshot copies, and clones. If
space_optimized is used, the blocks contained within Snapshot copies are not reorganized.
These parameters can be changed at any time, but do not enable read_realloc across the entire
environment at once because the additional work required could affect performance. Enabling it on one or
two datafile-containing volumes per day should be a safe approach.
NetApp recommends the following:
•
Set read_realloc on volumes containing the datafiles and then monitor the space consumption.
Enabling this option is unnecessary on volumes containing an archive log, control file, or other Oracle
file data, but doing so should not cause problems.
•
If the Snapshot copies appear to cause excess space consumption, change the setting to
space_optimized.
•
As stated previously, read_realloc is not applicable to AFF systems.
3.7
ONTAP and Third-Party Snapshots
Oracle Doc ID 604683.1 explains the requirements for third-party snapshot support and the multiple
options available for backup and restore operations.
The third-party vendor must guarantee that the company’s snapshots conform to the following
requirements:
•
Snapshots must integrate with Oracle's recommended restore and recovery operations.
•
Snapshots must be database crash consistent at the point of the snapshot.
•
Write ordering is preserved for each file within a snapshot.
ONTAP and NetApp Oracle management products comply with these requirements.
3.8
Cluster Operations—Takeover and Switchover
An understanding storage takeover and switchover function is required to ensure database operations are
not disrupted by these operations.
•
Under normal conditions, incoming writes to a given controller are synchronously mirrored to its
partner. In a NetApp MetroCluster™ environment, writes are also mirrored to a remote controller. Until
a write is stored in nonvolatile media in all locations, it is not acknowledged to the host application.
•
The media storing the write data is called nonvolatile memory or NVMEM. It is also sometimes
referred to nonvolatile random access memory (NVRAM), and it can be thought of as a write cache
although it functions as a journal. In a normal operation, the data from NVMEM is not read; it is only
12
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
used to protect data in the event of a software or hardware failure. When data is written to disk, the
data is transferred from the RAM in the system, not from NVMEM.
•
During a takeover operation, one node in a high availability (HA) pair takes over the operations from
its partner. A switchover is essentially the same, but it applies to MetroCluster configurations in which
a remote node takes over the functions of a local node.
During routine maintenance operations, a storage takeover or switchover operation should be
transparent, other than for a potential brief pause in database operations as the network paths change.
Networking can be complicated and it is easy to make errors, so NetApp strongly recommends testing
takeover and switchover operations thoroughly with a database before putting a storage system into
production. Doing so is the only way to be sure that all network paths are configured correctly. In a SAN
environment, carefully check the output of the command sanlun lun show -p to make sure that all
expected primary and secondary paths are available.
Care must be taken when issuing a forced takeover or switchover. Forcing a change to storage
configuration with these options means that the state of the controller that owns the disks is disregarded
and the alternative node forcibly takes control of the disks. Incorrect forcing of a takeover can result in
data loss or corruption. This is because a forced takeover or switchover can discard the contents of
NVMEM. After the takeover or switchover is complete, the loss of that data means that the data stored on
disk might revert to a slightly older state from the point of view of the database.
A forced takeover with a normal HA pair should rarely be required. In almost all failure scenarios, a node
shut downs and informs the partner so that an automatic failover takes place. There are some edge
cases, such as a rolling failure in which the interconnect between nodes is lost and then one controller is
lost, in which a forced takeover is required. In such a situation, the mirroring between nodes is lost before
the controller failure, which means that the surviving controller would have no longer has a copy of the
writes in progress. The takeover then needs to be forced, which means that data potentially is lost.
The same logic applies to a MetroCluster switchover. In normal conditions, a switchover is nearly
transparent. However, a disaster can result in a loss of connectivity between the surviving site and the
disaster site. From the point of view of the surviving site, the problem could be nothing more than an
interruption in connectivity between sites, and the original site might still be processing data. If a node
cannot verify the state of the primary controller, only a forced switchover is possible.
NetApp recommends taking the following precautions:
•
Be very careful to not accidentally force a takeover or a switchover. Normally, forcing should not be
required, and forcing the change can cause data loss.
•
If a forced takeover or switchover is required, make sure that the database is shut down, dismount all
file systems, shut down any ASM instances, and varyoff any logical volume manager (LVM) volume
groups.
•
In the event of a forced MetroCluster switchover, fence off the failed node from all surviving storage
resources. For more information, see the MetroCluster Management and Disaster Recovery Guide for
the relevant version of ONTAP.
MetroCluster and Multiple Aggregates
MetroCluster is a synchronous replication technology that switches to asynchronous mode if connectivity
is interrupted. This is the most common request from customers, because guaranteed synchronous
replication means that interruption in site connectivity leads to a complete stall of database I/O, taking the
database out of service.
With MetroCluster, aggregates rapidly resynchronize after connectivity is restored. Unlike other storage
technologies, MetroCluster should never require a complete remirroring after site failure. Only delta
changes must be shipped.
13
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
In databases that span aggregates, there is a small risk that additional data recovery steps would be
required in a rolling disaster scenario. Specifically, if (a) connectivity between sites is interrupted, (b)
connectivity is restored, (c) the aggregates reach a state in which some are synchronized and some are
not, and then (d) the primary site is lost, the result is a surviving site in which the aggregates are not
synchronized with one another. If this happens, parts of the database are synchronized with one another
and it is not possible to bring up the database without recovery. If a database spans aggregates, NetApp
strongly recommends leveraging Snapshot-based backups with one of the many available tools to verify
rapid recoverability in this unusual scenario.
NVFAIL
Databases are vulnerable to corruption if a failover or switchover is forced because databases maintain
large internal caches. If a forced failover occurs, previously acknowledged changes are effectively
discarded. The contents of the storage array effectively jump backward in time, and the state of the
database cache no longer reflects the state of the data on disk. The result is data corruption.
Caching can occur at the application or server layer. For example, an Oracle database server caches
data within the Oracle system global area (SGA). An operation that resulted in lost data would put the
database at risk of corruption because the blocks stored in the SGA might not match the blocks on the
array. A less obvious use of caching is at the OS file system layer. Blocks from a mounted NFS file
system might be cached in the OS, or a file system based on LUNs can cache data in the OS buffer
cache. A failure of NVRAM or a forced takeover in these situations could result in file system corruption.
ONTAP systems protect databases and operating systems from this scenario with NVFAIL and its
associated parameters.
4 Storage Virtual Machines and Logical Interfaces
This section provides an overview of key management principles. For more comprehensive
documentation, see the ONTAP Network Management Guide for the version of ONTAP in use. As with
other aspects of database architecture, the best options for storage virtual machine (SVM; formally known
as Vserver) and logical interface (LIF) design depend heavily on scaling requirements and business
needs.
Consider the following primary topics when building a LIF strategy:
•
Performance. Is the network bandwidth sufficient?
•
Resiliency. Are there any single points of failure in the design?
•
Manageability. Can the network be scaled nondisruptively?
These topics apply to the end-to-end solution, from the host through the switches to the storage system.
4.1
SVMs
SVMs are the basic functional unit of storage, so it is useful to compare an SVM to a guest on a VMware
ESX server. When first installed, ESX has no preconfigured capabilities, such as hosting a guest OS or
supporting an end-user application. It is an empty container until a virtual machine (VM) is defined.
ONTAP is similar. When first installed, this OS has no data-serving capabilities, and an SVM must be
defined. It is the SVM personality that defines the data services.
Some customers operate one primary SVM for most of their day-to-day requirements but then create a
small number of SVMs for special needs, including the following situations:
•
An SVM for a critical business database managed by a specialist team
•
An SVM for a development group to whom complete administrative control has been given so that
they can manage their own storage independently
14
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
•
An SVM for sensitive business data, such as human resources or financial reporting data, for which
the administrative team must be limited
In a multi-tenant environment, each tenant's data can be given a dedicated SVM. The recommended limit
for SVMs is approximately 125 per cluster node, but in general the LIF maximums are reached before the
SVM limit is reached. There is a point in which a multi-tenant environment is best separated based on
network segments rather than isolating them into dedicated SVMs.
4.2
LIF Types
There are multiple LIF types. Official ONTAP documentation provides more complete information on this
topic, but from a functional perspective LIFs can be divided into the following groups:
•
Cluster and node management LIFs. LIFs used to manage the storage cluster.
•
SVM management LIFs. Interfaces that permit access to an SVM through the ONTAP API (known
as ZAPI) for functions such as Snapshot copy creation or volume resizing. Products such as SMO
must have access to an SVM management LIF.
•
Data LIFs. Interfaces that carry FC, iSCSI, NFS, or CIFS data.
Note:
4.3
A data LIF used for NFS traffic can also be used for management by changing the firewall
policy from data to mgmt or another policy that allows HTTP, HTTPS, or SSH. This change
can simplify network configuration by avoiding the configuration of each host for access to
both the NFS data LIF and a separate management LIF. It is not possible to configure an
interface for both iSCSI and management traffic, despite the fact that both use an IP protocol.
A separate management LIF is required in iSCSI environments.
SAN LIF Design
LIF design in a SAN environment is relatively simple for one reason: multipathing. All modern SAN
implementations allow a client to access data over multiple network paths and select the best path or
paths for access. As a result, performance with respect to LIF design is simpler to address because SAN
clients automatically load-balance I/O across the best available paths.
If a path becomes unavailable, the client automatically selects a different path. The resulting simplicity of
design makes SAN LIFs generally more manageable. This does not mean that a SAN environment is
always more easily managed, because there are many other aspects of SAN storage that are much more
complicated than NFS. It simply means that SAN LIF design is easier.
Performance
The most important consideration with LIF performance in a SAN environment is bandwidth. For example,
a four-node ONTAP cluster with two 16Gb FC ports per node allows up to 32Gb of bandwidth from each
node. I/O is automatically balanced between ports, and all I/O is directed down the most optimal path.
Resiliency
SAN LIFs do not fail over. If a SAN LIF fails, then the client's multipathing ability detects the loss of a path
and redirects I/O to a different LIF.
Manageability
LIF migration is a much more common task in an NFS environment because LIF migration is often
associated with relocating volumes around the cluster. There is no need to migrate a LIF in a SAN
environment when volumes are relocated. That is because, after the volume move has completed,
ONTAP sends a notification to the SAN about a change in paths, and the SAN clients automatically
reoptimize. LIF migration with SAN is primarily associated with major physical hardware changes. For
15
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
example, if a nondisruptive upgrade of the controllers is required, a SAN LIF is migrated to the new
hardware. If an FC port is found to be faulty, a LIF can be migrated to an unused port.
Design Recommendations
NetApp makes the following primary recommendations:
•
Do not create more paths than are required. Excessive numbers of paths make overall management
more complicated and can cause problems with path failover on some hosts. Furthermore, some
hosts have unexpected path limitations for configurations such as SAN booting.
•
Very few LUNs should require more than four paths to storage. The value of having more than two
nodes advertising paths to LUNs is limited because the aggregate hosting a LUN is inaccessible if the
node that owns the LUN and its HA partner fail. Creating paths on nodes other than the primary HA
pair is not helpful in such a situation.
•
Although the number of visible LUN paths can be managed by selecting which ports are included in
FC zones, it is generally easier to include all potential target points in the FC zone and control LUN
visibility at the ONTAP level.
•
In ONTAP 8.3 and later, the selective LUN mapping (SLM) feature is the default. With SLM, any new
LUN is automatically advertised from the node that owns the underlying aggregate and the node's HA
partner. This arrangement avoids the need to create port sets or configure zoning to limit port
accessibility. Each LUN is available on the minimum number of nodes required for both optimal
performance and resiliency.
In the event a LUN must be migrated outside of the two controllers, the additional nodes can be
added with the lun mapping add-reporting-nodes command so that the LUNs are advertised
on the new nodes. Doing so creates additional SAN paths to the LUNs for LUN migration. However,
the host must perform a discovery operation to use the new paths.
•
Do not be overly concerned about indirect traffic. It is best to avoid indirect traffic in a very I/Ointensive environment for which every microsecond of latency is critical, but the visible performance
effect is negligible for typical workloads.
•
Follow the zoning rules described in section 13.1.
4.4
NFS LIF Design
In contrast to SAN protocols, NFS has a limited ability to define multiple paths to data. The parallel NFS
(pNFS) extensions to NFSv4.1 address this limitation, but pNFS is not yet supported for Oracle databases
and is not covered in this document.
Performance and Resiliency
Although measuring SAN LIF performance is primarily a matter of calculating the total bandwidth from all
primary paths, determining NFS LIF performance requires taking a closer look at the exact network
configuration. For example, two 10Gb ports can be configured as raw physical ports, or they can be
configured as a Link Aggregation Control Protocol (LACP) interface group. If they are configured as an
interface group, multiple load balancing policies are available that work differently depending on whether
traffic is switched or routed. Finally, Direct NFS (DNFS) offers load-balancing configurations that do not
exist in any OS NFS clients at this time.
Unlike SAN protocols, NFS file systems require resiliency at the protocol layer. For example, a LUN is
always configured with multipathing enabled, meaning that multiple redundant channels are available to
the storage system, each of which uses the FC protocol. An NFS file system, on the other hand, depends
on the availability of a single TCP/IP channel that can only be protected at the physical layer. This
arrangement is why options such as port failover and LACP port aggregation exist.
In an NFS environment, both performance and resiliency are provided at the network protocol layer. As a
result, both topics are intertwined and must be discussed together.
16
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
Bind LIFs to Port Groups
To bind a LIF to a port group, associate the LIF IP address with a group of physical ports. The primary
method for aggregating physical ports together is LACP. The fault-tolerance capability of LACP is fairly
simple; each port in an LACP group is monitored and is removed from the port group in the event of a
malfunction. There are, however, many misconceptions about how LACP works with respect to
performance:
•
LACP does not require the configuration on the switch to match the endpoint. For example, ONTAP
can be configured with IP-based load balancing, while a switch can use MAC-based load balancing.
•
Each endpoint using an LACP connection can independently choose the packet transmission port,
but it cannot choose the port used for receipt. This means that traffic from ONTAP to a particular
destination is tied to a particular port, and the return traffic could arrive on a different interface. This
does not cause problems, however.
•
LACP does not evenly distribute traffic all the time. In a large environment with many NFS clients, the
result is typically even use of all ports in an LACP aggregation. However, any one NFS file system in
the environment is limited to the bandwidth of only one port, not the entire aggregation.
•
Although robin-robin LACP policies are available on ONTAP, these policies do not address the
connection from a switch to a host. For example, a configuration with a four-port LACP trunk on a
host and a four-port LACP trunk on ONTAP is still only able to read a file system using a single port.
Although ONTAP can transmit data through all four ports, no switch technologies are currently
available that send from the switch to the host through all four ports. Only one is used.
The most common approach in larger environments consisting of many database hosts is to build an
LACP aggregate of an appropriate number of 10Gb interfaces by using IP load balancing. This approach
enables ONTAP to deliver even use of all ports, as long as enough clients exist. Load balancing breaks
down when there are fewer clients in the configuration because LACP trunking does not dynamically
redistribute load.
When a connection is established, traffic in a particular direction is placed on only one port. For example,
a database performing a full table scan against an NFS file system connected through a four-port LACP
trunk reads data though only one network interface card (NIC). If only three database servers are in such
an environment, it is possible that all three are reading from the same port, while the other three ports are
idle.
Bind LIFs to Physical Ports
Binding a LIF to a physical port results in more granular control over network configuration because a
given IP address on a ONTAP system is associated with only one network port at a time. Resiliency is
then accomplished through the configuration of failover groups and failover policies.
Failover Policies and Failover Groups
The behavior of LIFs during network disruption is controlled by failover policies and failover groups.
Configuration options have changed with the different versions of ONTAP. Consult the ONTAP Network
Management Guide for specific details for the version of ONTAP being deployed.
Follow these general practices for ONTAP 8.2 and earlier:
1. Configure a failover group to be user defined.
2. Populate the failover group with ports on the storage failover (SFO) partner controller so that the LIFs
follow the aggregates during a storage failover. This configuration avoids the creation of indirect
traffic.
3. Use failover ports with performance characteristics that match the original LIF. For example, a LIF on
a single physical 10Gb port should include a failover group with a single 10Gb port. A four-port LACP
LIF should fail over to another four-port LACP LIF.
17
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
4. Set the failover policy to priority.
ONTAP 8.3 allows management of LIF failover based on broadcast domains. Therefore, an administrator
can define all of the ports that have access to a given subnet and allow ONTAP to select an appropriate
failover LIF. This approach can be used by some customers, but it has limitations in a high-speed
database storage network environment because of the lack of predictability. For example, an environment
can include both 1Gb ports for routine file system access and 10Gb ports for datafile I/O. If both types of
ports exist in the same broadcast domain, LIF failover can result in moving datafile I/O from a 10Gb port
to a 1Gb port.
NetApp recommends using the ONTAP 8.2 approach that defines which ports can be used for LIF
failover. In summary, consider the following practices:
1. Configure a failover group as user-defined.
2. Populate the failover group with ports on the SFO partner controller so that the LIFs follow the
aggregates during a storage failover. This avoids creating indirect traffic.
3. User failover ports with matching performance characteristics to the original LIF. For example, a LIF
on a single physical 10Gb port should include a failover group with a single 10Gb port. A four-port
LACP LIF should fail over to another four-port LACP LIF. These ports would be a subset of the ports
defined in the broadcast domain.
4. Set the failover policy to SFO-partner only. Doing so makes sure that the LIF follows the aggregate
during failover.
Auto-revert
Set the auto-revert parameter as desired. Most customers prefer to set this parameter to true to
have the LIF revert to its home port. However, in some cases, customers have set this to false so that
an unexpected failover can be investigated before returning a LIF to its home port.
LIF-to-Volume Ratio
A common misconception is that there must be a 1:1 relationship between volumes and NFS LIFs.
Although this configuration is required for moving a volume anywhere in a cluster while never creating
additional interconnect traffic, it is categorically not a requirement. Intercluster traffic must be considered,
but the mere presence of intercluster traffic does not create problems. Many of the published benchmarks
created for ONTAP include predominantly indirect I/O.
For example, a database project containing a relatively small number of performance-critical databases
that only required a total of 40 volumes might warrant a 1:1 volume to LIF strategy, an arrangement that
would require 40 IP addresses. Any volume could then be moved anywhere in the cluster along with the
associated LIF, and traffic would always be direct, minimizing every source of latency even at
microsecond levels.
As a counter example, a large hosted environment might be more easily managed with a 1:1 relationship
between customers and LIFs. Over time, a volume might need to be migrated to a different node, which
would cause some indirect traffic. However, the performance effect should be undetectable unless the
network ports on the interconnect switch are saturating. If there is concern, a new LIF can be established
on additional nodes and the host can be updated at the next maintenance window to remove indirect
traffic from the configuration.
5 Quality of Service
The increased adoption of all-flash storage has also resulted in consolidation of database workloads.
Storage arrays relying on spinning media tended to support only a limited number of databases because
of the limited IOPS capabilities of older drive technology. One or two highly active databases would
saturate the underlying disks long before the storage controllers reached their limits. This has changed. A
18
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
performance capability of a relatively small number of SSD drives can saturate even the most powerful
storage controllers. This means the full capabilities of the controllers can be leveraged without the fear of
sudden collapse of performance as spinning media latency spiked.
As a reference example, a simple two-node HA AFF8080 system is capable of servicing around 400K
random IOPS before latency climbs above one millisecond. Less than one percent of databases would be
expected to reach such levels, and allowing an AFF8080 to manage just 10K IOPS would be wasteful.
There are two types of quality of service (QoS) in ONTAP: IOPS and bandwidth. QoS controls can be
applied to SVMs, volumes, LUNs, and files.
5.1
IOPS QoS
An IOPS QoS control is obviously based on the total IOPS of a given resource, but there are a number of
aspects of IOPS QoS that might not be intuitive. A few customers have been initially puzzled by the
apparent increase in latency when an IOPS threshold is reached. This is the only viable method to limit
IOPS. Logically, it functions similar to a token system. For example, if a given volume containing datafiles
has a 10K IOPS limit, each IO that arrives must first receive a token to continue processing. So long as
no more than 10K tokens have been consumed in a given second, no delays are present.
5.2
Bandwidth QoS
First, not all I/O sizes are the same. Databases might be performing a large number of fully random block
reads which would result in the IOPS threshold being reached, but databases might also be performing a
full table scan operation which would consist of a very small number of large block reads, consuming a
very large amount of bandwidth but relatively few IOPS.
5.3
Guaranteed QoS
Many customers seek a solution that includes guaranteed QoS, which is more difficult than it seems and
potentially wasteful. For example, placing 10 databases with a 10K IOPS guarantee would require a
system to be sized for a scenario where all 10 databases are simultaneously running at 10K IOPS, for a
total of 100K. This is a highly unlikely scenario, but if it is required, the best option is to guarantee
performance through the sizing effort
For example, each AFF8080 HA pair could constitute a 400K IOPS building block. The total guaranteed
IOPS of all hosted databases should not add up to more than 400K IOPS. Furthermore, to prevent a
given database from consuming more than its allotted IOPS capability it should have the standard QoS
applied as a limit.
6 Compression, Compaction, and Deduplication
Compression and deduplication are two storage efficiency options that increase the amount of logical
data that fits on a given amount of physical storage. At a high level, compression is a mathematical
process whereby patterns in data are detected and encoded in a way that reduces space requirements.
In contrast, deduplication detects actual repeated blocks of data and removes the extraneous copies.
Although they deliver similar results, they work in significantly different ways and therefore must be
managed differently.
6.1
Compression
There are multiple ways to compress a database. Until recently, compression was of limited value
because most databases required a very large number of spindles to provide sufficient performance. One
side effect of building a storage array with acceptable performance was that the array generally offered
more capacity than required. The situation has changed with the rise of solid-state storage. There is no
longer a need to vastly overprovision the drive count to obtain good performance.
19
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
Even without compression, migrating a database to a partially or fully solid-state storage platform can
yield significant cost savings because doing so avoids the need to purchase drives only needed to
support I/O. For example, NetApp has examined some storage configurations from recent large database
projects and compared the costs with and without the use of NetApp Flash Cache ™ or Flash Pool™
intelligent data caching. These flash technologies decreased costs by approximately 50% because IOPSdense flash media permits a significant reduction in the number of spinning disks and shelves than would
otherwise be required.
As stated above, the increased IOPS capability of solid-state drives (SSDs) almost always yields cost
savings, but compression can achieve further savings by increasing the effective capacity of solid-state
media. Although compression can be performed by the database itself, this is rarely observed in an
Oracle environment. The built-in compression option is not suitable for rapidly changing data, and the
advanced compression option has a high licensing cost. In addition, the cost of the Oracle database itself
is high. It makes little sense to pay a high per-CPU license cost for a CPU that performs data
compression and decompression rather than real database work. A better option is to offload the
compression work on to the storage system.
The value of compression is not limited to a pure SSD environment. Hybrid options such as Flash Pool
aggregates also benefit from compression. When data is compressed in the SSD layer, the effective
result is an increase in SSD storage capacity.
ONTAP 8.3.1
ONTAP 8.3.1 introduces adaptive compression, an inline compression method that works with blocks of
varying sizes. The performance effect is minimal, and enabling compression can improve overall
performance in some cases. In addition, the compression method available in ONTAP 8.3 and earlier is
still available and is now renamed secondary compression.
There is no single best practice for the use of compression; the best option depends on business
practices. Most databases can be placed on volumes with adaptive compression enabled with no
requirement for separation of data or special treatment of files. As the name implies, adaptive
compression adapts to multiple data types and I/O patterns.
Adaptive Compression
Adaptive compression has been thoroughly tested with Oracle workloads, and the performance effect has
been found to be negligible, even in an all-flash environment in which latency is measured in
microseconds. In initial testing, some customers have reported a performance increase with the use of
compression. This increase is the result of compression effectively increasing the amount of Flash Pool
SSD available to the database.
ONTAP manages physical blocks in 4KB units. Therefore, the maximum possible compression ratio is 2:1
with a typical Oracle database using an 8KB block. Early testing with real customer data has shown
compression ratios approaching this level, but results vary based on the type of data stored.
Secondary Compression
Secondary compression uses a larger block size that is fixed at 32KB. This feature enables ONTAP to
compress data with increased efficiency, but secondary compression is primarily designed for data at rest
or data that is written sequentially and requires maximum compression.
NetApp recommends secondary compression for data such as archive logs or Recovery Manager
(RMAN) backups. These types of files are written sequentially and not updated. This point does not mean
that adaptive compression is discouraged. However, if the volume of data being stored is large, then
secondary compression delivers better savings when compared to adaptive compression.
Consider secondary compression of datafiles when the amount of data is very large and the datafiles
themselves are either read-only or rarely updated. Datafiles using a 32KB block size should see more
20
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
compression under secondary compression that has a matching 32KB block size. However, care must be
taken to verify that data using block sizes other than 32KB are not placed on these volumes. Only use
this method in cases in which the data is not frequently updated.
Caution
Secondary compression and deduplication should not be used together with RMAN backups. The
reason is small changes to the backed-up data will affect the 32KB compression window. If the window
shifts, the resulting compressed data will differ across the entire file. Deduplication occurs after
compression, which means the deduplication engine will see each compressed backup differently. If
deduplication of RMAN backups is required, secondary compression should not be used. Adaptive
compression is preferable, because it works at a smaller block size and will not disrupt deduplication
efficiency. For similar reasons host-side compression will also interfere with deduplication efficiency.
Compression and Thin Provisioning
Compression is a form of thin provisioning. For example, a 100GB LUN occupying a 100GB volume might
compress down to 50GB. There are no actual savings realized yet because the volume is still 100GB.
The volume must first be reduced in size so the space saved can be used elsewhere on the system. If
later changes to the 100GB LUN result in the data becoming less compressible, then the LUN grows in
size and the volume could fill up. Thin provisioning can yield a substantial improvement in usable capacity
with associated cost savings, but space utilization must be monitored to make sure that capacity is not
unexpectedly exhausted.
Alignment
Adaptive compression in a database environment requires some consideration of compression block
alignment. Doing so is only a concern for data that is subject to random overwrites of very specific blocks.
This approach is similar in concept to overall file system alignment, as is described in the section
“Zoning.”
For example, an Oracle 8KB write to a datafile is compressed only if it aligns with an 8KB boundary within
the file system itself. This point means that it must fall within the first 8KB of the file, the second 8KB of
the file, and so forth. Data such as RMAN backups or archive logs are sequentially written operations that
span multiple blocks, all of which are compressed. There is no need to consider alignment. The only I/O
pattern of concern is the random overwrites of datafiles.
NFS
With the use of NFS, datafiles are aligned. Each block of the datafile is aligned with respect to the start of
the file.
SAN
SAN environments require data to be aligned to an 8KB boundary for optimum compression. There are
two aspects of alignment for SAN: the LUN and the file system. The LUN must be configured as either a
whole-disk device (no partition) or with a partition that aligns to an 8K boundary. See the OS-specific
sections that follow for details on compression and alignment on a configuration-by-configuration basis.
Recommendations for ONTAP 8.3.1 and Later
NetApp provides the following recommendations for ONTAP 8.3.1 and later:
•
The simplest approach for leveraging compression is enabling adaptive compression for all database
volumes. As stated previously, adaptive compression is suitable for all I/O patterns. Note the following
exceptions:
21
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
•

If a volume is not thin provisioned, do not enable compression because doing so provides no
benefit.

If a very large number of archive logs are retained, moving the archive logs to a volume using
secondary compression improves storage efficiency.

Some databases have very high redo logging rates. Redo logs are comparatively small and are
constantly overwritten, so any space savings from compression is negligible. This data should be
moved to a volume without compression.

If datafiles contain a significant amount of uncompressible data, for example when compression
is already enabled or encryption is used, place this data on volumes without compression.
The exceptions above should not be overemphasized. ONTAP offers the flexibility to choose when
compression is enabled, and these exceptions are listed in the interest of offering the broadest
choices for customers. Although in most tests the effect of compression is not detectable, it is not
nonzero. Some customers seek the highest possible platform performance and minimize every
microsecond of latency. In such cases, compression might not be the best option.
Recommendations for ONTAP 8.3 and Earlier
Use ONTAP compression with care in ONTAP 8.3 and earlier because the compression block size is 32K
and a typical database uses an 8KB block. As a result, updating a single 8KB block requires ONTAP to
read four Oracle blocks, update a single 8KB unit, and write it back to disk.
Some customers have successfully used ONTAP compression in 8.3 and earlier with data that is either
written sequentially, such as archived logs, or is not frequently updated, such as archival data in a
datafile.
•
Do not enable compression on any volumes containing Oracle data unless (a) the change rate is
extremely low or (b) the data is written sequentially and not subject to updates, such as for RMAN
backup files or archive logs.
ONTAP Version-Specific Notes
The following notes are specific to the ONTAP version in use:
•
In ONTAP 8.2, the use of compression on a volume prevents data from being cached by Flash Pool.
•
In ONTAP 8.3, compressed blocks are eligible for Flash Pool read caching but not write caching.
•
In ONTAP 8.3.1, compressed blocks are eligible for both Flash Pool read and write caching.
•
Flash Cache can be used with compressed volumes, but the data is stored in the flash layer in an
uncompressed format.
Note:
6.2
See the section “Fractional Reservations” for an explanation of the interaction between
compression and fractional reservation.
Inline Data Compaction
Inline data compaction is a technology introduced in ONTAP 9, which improves compression efficiency.
As stated previously, adaptive compression alone can provide at best 2:1 savings because it is limited to
storing an 8K IO in a 4K WAFL block. Compression methods such as Secondary Compression use a
larger block size and deliver better efficiency but are not suitable for data that is subject to small block
overwrites. Decompressing 32KB units of data, updating an 8K portion, recompressing, and writing back
to disk creates overhead.
Inline data compaction works by allowing logical WAFL blocks to be stored within physical WAFL blocks.
For example, a database with highly compressible data such as text or partially full blocks might
compress from 8KB to 1KB. Without compaction, that 1KB of data would still occupy an entire 4KB block.
Inline data compaction allows that 1KB of compressed data to be stored in just 1KB of physical space
22
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
alongside other compressed data. It is not a compression technology, it is simply a more efficient way of
allocating space on disk and therefore should not create any detectable performance impact.
The degree of savings obtained will vary. Data that is already compressed or encrypted cannot generally
be further compressed and therefore such datasets will not benefit from compaction. Newly initialized
Oracle datafiles that contain little more than block metadata and zeros compress up to 80:1. This creates
an extremely wide range of possibilities. The best way to evaluate potential savings is using the NetApp
Space Savings Estimation Tool (SSET) available on NetApp Field Portal or through your NetApp
representative.
6.3
Deduplication
Do not use deduplication with Oracle database files primarily because this process is almost entirely
ineffective. An Oracle block contains a header that is globally unique to the database and a trailer that is
nearly unique. One percent space savings are possible, but at the expense of significant overhead
caused by data deduplication.
Space savings of up to 15% in databases with 16k and large block sizes have been observed in a few
cases. The initial 4KB of each block contains the globally unique header, and the final 4KB block contains
the nearly unique trailer. The intervening blocks are candidates for deduplication, although in practice this
is almost entirely attributed to the deduplication of zeroed data.
Many competing arrays claim the ability to deduplicate Oracle databases based on the presumption that a
database is copied multiple times. In this respect, NetApp deduplication could also be used, but ONTAP
offers a better option: NetApp FlexClone® technology. The end result is the same; multiple copies of an
Oracle database that share most of the underlying physical blocks are created. Using FlexClone is much
more efficient than taking the time to copy datafiles and then deduplicate them. It is, in effect,
nonduplication rather than deduplication, because a duplicate is never created in the first place.
In the unusual case in which multiple copies of the same datafiles exist, deduplication can be used.
NetApp recommends disabling deduplication on any volume containing Oracle datafiles unless the
volume is known to contain multiple copies of the same data, such as repeated RMAN backups to a
single volume.
7 Thin Provisioning
Thin provisioning refers to configuring more space on a storage system than is technically available. Such
configuring comes in many forms and is integral to the many features that ONTAP offers to an Oracle
database environment.
Almost any use of Snapshot involves thin provisioning. For example, a typical 10TB database on NetApp
storage contains 30 days of Snapshot copies. This arrangement results in approximately 10TB of data
visible in the active file system and 300TB dedicated to Snapshot copies. The total 310TB of storage
usually resides on approximately 12TB to 15TB of space. The active database consumes 10TB, and the
remaining 300TB of data only requires 2TB to 5TB of space because only the changes to the original data
are stored.
Cloning is also an example of thin provisioning. One of NetApp’s major customers has created 40 clones
of an 80TB database for use by development. If all 40 developers overwrote every block in every datafile,
over 3.2PB of storage would be required. In practice, turnover is low and the collective space requirement
is closer to 40TB, because only the changes are stored on disk.
7.1
Space Management
Some care must be taken with thin provisioning an Oracle environment because data change rates can
increase unexpectedly. For example, space consumption due to Snapshot copies can grow rapidly if
23
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
tables are reindexed, or a misplaced RMAN backup can write a large amount of data in a very short time.
Finally, it can be difficult to recover an Oracle database if a file system runs out of free space during
datafile extension.
Fortunately, these risks can be addressed with careful configuration of volume-autogrow and
snapshot-autodelete policies. As their names imply, these options enable a user to create policies
that automatically clear space consumed by Snapshot copies or grow a volume to accommodate
additional data. Many options are available, and needs vary by customer.
See the “ONTAP Logical Storage Management Guide” for a complete discussion of these features.
7.2
LUN Thin Provisioning
Thin provisioning of active LUNs is of limited use in an Oracle environment because Oracle initializes
datafiles to their full size at the time of creation. The efficiency of thin provisioning of active LUNs in a file
system environment can be lost over time as deleted and erased data occupy more and more unallocated
whitespace in the file system.
There is one exception when logical volume managers (LVM) are used. When an LVM such as Veritas
VxVM or Oracle ASM is used, the underlying LUNs are divided into extents that are only used when
needed. For example, if a database begins at 2TB in size but could grow to 10TB over time, this database
can be placed on 10TB of thin-provisioned LUNs organized in an LVM disk group. It would occupy only
2TB of disk space at the time of creation and would only claim additional space as extents are allocated
to accommodate database growth. This process is safe as long as space is monitored.
7.3
Fractional Reservations
Fractional reserve refers to the behavior of a LUN in a volume with respect to space efficiency. When the
option fractional-reserve is set to 100%, all data in the volume can experience 100% turnover with
any data pattern without exhausting space on the volume.
For a Snapshot example, consider a database on a single 250GB LUN in a 1TB volume. Creating a
Snapshot copy would immediately result in the reservation of an additional 250GB of space in the volume
to guarantee that the volume does not run out of space for any reason. Using fractional reserves is
generally wasteful because it is extremely unlikely that every byte in the database volume would need to
be overwritten. There is no reason to reserve space for an event that never happens. Still, if a customer
cannot monitor space consumption in a storage system and must be certain that space never runs out,
100% fractional reservations would be a requirement to use Snapshot copies.
7.4
Compression and Deduplication
Compression and deduplication are both forms of thin provisioning. For example, a 50TB database
footprint might compress to 30TB, resulting in a savings of 20TB. For compression to yield any benefits,
some of that 20TB must be used for other data or the storage system must be purchased with less than
50TB. The result is storing more data than is technically available on the storage system. From the
database point of view, there is 50TB of data, even though it occupies only 30TB on disk.
There is always a possibility that the compressibility of a database changes, which would result in
increased consumption of real space. This increase in consumption means that compression must be
managed as with other forms of thin provisioning in terms of monitoring and using volume-autogrow
and snapshot-autodelete.
Compression and deduplication are discussed in further detail in the sections “Compression” and
“Deduplication.”
24
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
7.5
Compression and Fractional Reservations
Compression is a form of thin provisioning. Fractional reservations affect the use of compression, with
one important note; space is reserved in advance of the Snapshot copy creation. Normally fractional
reserve is only important if a Snapshot copy exists. If there is no Snapshot copy, fractional reserve is not
important. This is not the case with compression. If a LUN is created on a volume with compression,
ONTAP preserves space to accommodate a Snapshot copy. This behavior can be confusing during
configuration, but it is expected.
As an example, consider a 10GB volume with a 5GB LUN that has been compressed down to 2.5GB with
no Snapshot copies. Consider these two scenarios:
•
Fractional reserve = 100 results in 7.5GB utilization
•
Fractional reserve = 0 results in 2.5GB utilization
The first scenario includes 2.5GB of space consumption for current data and 5GB of space to account for
100% turnover of the source in anticipation of Snapshot copy use. The second scenario reserves no extra
space.
Although this situation might seem confusing, it is unlikely to be encountered in practice. Compression
implies thin provisioning, and thin provisioning in a LUN environment requires fractional reservations. It is
always possible for compressed data to be overwritten by something uncompressible, which means a
volume must be thin provisioned for compression to result in any savings.
NetApp recommends the following reserve configurations:
•
Set fractional-reserve to 0 when basic capacity monitoring is in place along with volumeautogrow and snapshot-autodelete.
•
Set fractional-reserve to 100 if there is no monitoring ability or if it is impossible to exhaust
space under any circumstance.
8 Performance Optimization and Benchmarking
Accurate testing of database storage performance is an extremely complicated subject. It requires not just
an understanding of IOPS and throughput, but also the understanding of the difference between
foreground and background I/O operations, the impact of latency upon the database, and numerous OS
and network settings that also affect storage performance. In addition, there are nonstorage databases
tasks to consider. There is a point where optimizing storage performance yields no useful benefits
because storage performance is no longer a limiting factor for performance.
A majority of database customers now select an all-flash array, which creates some additional
considerations. As an example, consider performance testing on a two-node AFF8080 system:
•
With a 75/25 read/write ratio, two AFF8080 nodes can deliver over 300K random database IOPS
before latency even crosses the 1ms mark. This is so far beyond the current performance demands
of most databases that it is difficult to predict the expected improvement. Storage would be largely
erased as a bottleneck.
•
Network bandwidth is an increasingly common source of performance limitations. For example,
spinning disk solutions are often bottlenecks for database performance because the I/O latency is
very high. When latency limitations are removed by an all-flash array, the barrier frequently shifts to
the network. This is especially notable with virtualized environments and blade systems where the
true network connectivity is difficult to visualize. This can complicate performance testing if the
storage system itself cannot be fully utilized due to bandwidth limitations.
•
Comparing the performance of an all-flash array with an array containing spinning disks is generally
not possible because of the dramatically improved latency of all-flash arrays. Test results are typically
not meaningful.
25
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
•
Comparing peak IOPS performance with an all-flash array is frequently not a useful test because
databases are not limited by storage I/O. For example, assume one array can sustain 500K random
IOPS, whereas another can sustain 300K. The difference is irrelevant in the real world if a database
is spending 99% of its time on CPU processing. The workloads never utilize the full capabilities of the
storage array. In contrast, peak IOPS capabilities might be critical in a consolidation platform in which
the storage array is expected to be loaded to its peak capabilities.
•
Always consider latency as well as IOPS in any storage test. Many storage arrays in the market make
claims of extreme levels of IOPS, but the latency renders those IOPS useless at such levels. The
typical target with all-flash arrays is the 1ms mark. A better approach to testing is not to measure the
maximum possible IOPS, but to determine how many IOPS a storage array can sustain before
average latency is greater than 1ms.
8.1
Oracle Automatic Workload Repository and Benchmarking
The gold standard for Oracle performance comparison is an Oracle Automatic Workload Repository
(AWR) report.
There are multiple types of AWR reports. From a storage point of view, a report generated by running the
awrrpt.sql command is the most comprehensive and valuable because it targets a specific database
instance and includes some detailed histograms that break down storage I/O events based on latency.
Comparing two performance arrays ideally involves running the same workload on each array and
producing an AWR report that precisely targets the workload. In the case of a very long-running workload,
a single AWR report with an elapsed time that encompasses the start and stop time can be used, but it is
preferable to break out the AWR data as multiple reports. For example, if a batch job ran from midnight to
6 a.m., create a series of one-hour AWR reports from midnight–1 a.m., 1 a.m.–2 a.m., and so on.
In other cases, a very short query should be optimized. The best option is an AWR report based on an
AWR snapshot created when the query begins and a second AWR snapshot created when the query
ends. The database server should be otherwise quiet to minimize the background activity that would
obscure the activity of the query under analysis.
Note:
8.2
Where AWR reports are not available, Oracle statspack reports are a good alternative. They
contain most of the same I/O statistics as an AWR report.
Oracle AWR and Troubleshooting
An AWR report is also the most important tool for analyzing a performance problem.
As with benchmarking, performance troubleshooting requires that you precisely measure a particular
workload. When possible, provide AWR data when reporting a performance problem to the NetApp
support center or when working with a NetApp or partner account team about a new solution.
When providing AWR data, consider the following requirements:
•
Run the awrrpt.sql command to generate the report. The output can be either text or HTML.
•
If Oracle RAC is used, generate AWR reports for each instance in the cluster.
•
Target the specific time the problem existed. The maximum acceptable elapsed time of an AWR
report is generally one hour. If a problem persists for multiple hours or involves a multihour operation
such as a batch job, provide multiple one-hour AWR reports that cover the entire period to be
analyzed.
•
If possible, adjust the AWR snapshot interval to 15 minutes. This setting allows a more detailed
analysis to be performed. This also requires additional executions of awrrpt.sql to provide a report
for each 15-minute interval.
•
If the problem is a very short running query, provide an AWR report based on an AWR snapshot
created when the operation begins and a second AWR snapshot created when the operation ends.
26
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
The database server should be otherwise quiet to minimize the background activity that would
obscure the activity of the operation under analysis.
•
If a performance problem is reported at certain times but not others, provide additional AWR data that
demonstrates good performance for comparison.
8.3
calibrate_io
The calibrate_io command should never be used to test, compare, or benchmark storage systems.
As stated in the Oracle documentation, this procedure calibrates the I/O capabilities of storage.
Calibration is not the same as benchmarking. The purpose of this command is to issue I/O to help
calibrate database operations and improve their efficiency by optimizing the level of I/O issued to the
host. Because the type of I/O performed by the calibrate_io operation does not represent actual
database user I/O, the results are not predictable and are frequently not even reproducible.
8.4
SLOB2
SLOB2, the Silly Little Oracle Benchmark, has become the preferred tool for evaluating database
performance. It was developed by Kevin Closson and is available here. It takes minutes to install and
configure, and it uses an actual Oracle database to generate I/O patterns on a user-definable tablespace.
It is one of the few testing options available that can saturate an all-flash array with I/O. It is also useful for
generating much lower levels of I/O to simulate storage workloads that are low IOPS but latency
sensitive.
8.5
Swingbench
Swingbench can be useful for testing database performance, but it is extremely difficult to use
Swingbench in a way that stresses storage. NetApp has not seen any tests from Swingbench that yielded
enough I/O to be a significant load on any AFF array. In limited cases, the Order Entry Test (OET) can be
used to evaluate storage from a latency point of view. This could be useful in situations where a database
has a known latency dependency for particular queries. Care must be taken to make sure that the host
and network are properly configured to realize the latency potentials of an all-flash array.
8.6
HammerDB
HammerDB is a database testing tool that simulates TPC-C and TPC-H benchmarks, among others. It
can take a lot of time to construct a sufficiently large data set to properly execute a test, but it can be an
effective tool for evaluating performance for OLTP and data warehouse applications.
8.7
Orion
The Oracle Orion tool was commonly used with Oracle 9, but it has not been maintained to ensure
compatibility with changes in various host operation systems. It is rarely used with Oracle 10 or Oracle 11
due to incompatibilities with OS and storage configuration.
Oracle rewrote the tool, and it is installed by default with Oracle 12c. Although this product has been
improved and uses many of the same calls that a real Oracle database uses, it does not use precisely the
same code path or I/O behavior used by Oracle. For example, most Oracle I/Os are performed
synchronously, meaning the database halts until the I/O is complete as the I/O operation completes in the
foreground. Simply flooding a storage system with random I/Os is not a reproduction of real Oracle I/O
and does not offer a direct method of comparing storage arrays or measuring the effect of configuration
changes.
That said, there are some use cases for Orion, such as general measurement of the maximum possible
performance of a particular host-network-storage configuration, or to gauge the health of a storage
system. With careful testing, usable Orion tests could be devised to compare storage arrays or evaluate
27
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
the effect of a configuration change so long as the parameters include consideration of IOPS, throughput,
and latency and attempt to faithfully replicate a realistic workload.
9 General Oracle Configuration
The following parameters are generally applicable to all configurations.
9.1
filesystemio_options
The Oracle initialization parameter filesystemio_options controls the use of asynchronous and
direct I/O. Contrary to common belief, asynchronous and direct I/O are not mutually exclusive. NetApp
has observed that this parameter is frequently misconfigured in customer environments, and this
misconfiguration is directly responsible for many performance problems.
Asynchronous I/O means that Oracle I/O operations can be parallelized. Before the availability of
asynchronous I/O on various OSs, users configured numerous dbwriter processes and changed the
server process configuration. With asynchronous I/O, the OS itself performs I/O on behalf of the database
software in a highly efficient and parallel manner. This process does not place data at risk, and critical
operations, such as Oracle redo logging, are still performed synchronously.
Direct I/O bypasses the OS buffer cache. I/O on a UNIX system ordinarily flows through the OS buffer
cache. This is useful for applications that do not maintain an internal cache, but Oracle has its own buffer
cache within the SGA. In almost all cases, it is better to enable direct I/O and allocate server RAM to the
SGA rather than to rely on the OS buffer cache. The Oracle SGA uses the memory more efficiently. In
addition, when I/O flows through the OS buffer, it is subject to additional processing, which increases
latencies. The increased latencies are especially noticeable with heavy write I/O when low latency is a
critical requirement.
The options for filesystemio_options are:
•
async. Oracle submits I/O requests to the OS for processing. This process allows Oracle to perform
other work rather than waiting for I/O completion and thus increases I/O parallelization.
•
directio. Oracle performs I/O directly against physical files rather than routing I/O through the host
OS cache.
•
none. Oracle uses synchronous and buffered I/O. In this configuration, the choice between shared
and dedicated server processes and the number of dbwriters are more important.
•
setall. Oracle uses both asynchronous and direct I/O.
In almost all cases, the use of setall is optimal, but consider the following issues:
•
Some customers have encountered asynchronous I/O problems in the past, especially with previous
Red Hat Enterprise Linux 4 (RHEL4) releases. These problems are no longer reported, however, and
asynchronous I/O is stable on all current OSs.
•
If a database has been using buffered I/O, a switch to direct I/O might also warrant a change in the
SGA size. Disabling buffered I/O eliminates the performance benefit that the host OS cache provides
for the database. Adding RAM back to the SGA repairs this problem. The net result should be an
improvement in I/O performance.
•
Although it is almost always better to use RAM for the Oracle SGA than for OS buffer caching, it
might be impossible to determine the best value. For example, it might be preferable to use buffered
I/O with very small SGA sizes on a database server with many intermittently active Oracle instances.
This arrangement allows the flexible use of the remaining free RAM on the OS by all running
database instances. This is a highly unusual situation, but it has been observed at some customer
sites.
28
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
Note:
The filesystemio_options parameter has no effect in DNFS and ASM environments.
The use of DNFS or ASM automatically results in the use of both asynchronous and direct
I/O.
NetApp recommends the following:
•
Set filesystemio_options to setall, but be aware that under some circumstances the loss of
the host buffer cache might require an increase in the Oracle SGA.
9.2
db_file_multiblock_read_count
The db_file_multiblock_read_count parameter controls the maximum number of Oracle
database blocks that Oracle reads as a single operation during sequential I/O. This parameter does not,
however, affect the number of blocks that Oracle reads during any and all read operations, nor does it
affect random I/O. Only sequential I/O is affected.
Oracle recommends that the user leave this parameter unset. Doing so allows the database software to
automatically set the optimum value. This generally means that this parameter is set to a value that yields
an I/O size of 1MB. For example, a 1MB read of 8KB blocks would require 128 blocks to be read, and the
default value for this parameter would therefore be 128.
Most database performance problems observed by NetApp at customer sites involve an incorrect setting
for this parameter. There were valid reasons to change this value with Oracle versions 8 and 9. As a
result, the parameter might be unknowingly present in init.ora files because the database was
upgraded in place to Oracle 10 and later. A legacy setting of 8 or 16, compared to a default value of 128,
significantly damages sequential I/O performance.
NetApp recommends the following:
•
The db_file_multiblock_read_count parameter should not be present in the init.ora file.
NetApp has never encountered a situation in which changing this parameter improved performance,
but there are many cases in which it caused clear damage to sequential I/O throughput.
9.3
Redo Block Size
Oracle supports either a 512-byte or 4KB redo block size. The default is 512 bytes. The best option is
expected to be 512 bytes because this size minimizes the amount of data written during redo operations.
However, it is possible that the 4KB size could offer a performance benefit at very high logging rates. For
example, a single database with 50MBps of redo logging might be more efficient if the redo block size is
larger. A storage system supporting many databases with a large total amount of redo logging might
benefit from a 4KB redo block size. This is because this setting would eliminate inefficient partial I/O
processing when only a part of a 4KB block must be updated.
It is not correct that all I/O operations are performed in single units of the redo log block size. At very high
logging rates, the database generally performs very large I/O operations composed of multiple redo
blocks. The actual size of those redo blocks does not generally affect the efficiency of logging.
NetApp recommends the following:
•
Only change the default block size for cause, such as a documented requirement for a particular
application or because of a recommendation made by NetApp or Oracle customer support.
9.4
Checksums and Data Integrity
One question commonly directed to NetApp is how to secure the data integrity of a database. This
question is particularly common when a customer who is accustomed to using Oracle RMAN streaming
backups migrates to snapshot-based backups. One feature of RMAN is that it performs integrity checks
during backup operations. Although this feature has some value, its primary benefit is for a database that
is not used on a modern storage array. When physical disks are used for an Oracle database, it is nearly
29
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
certain that corruption eventually occurs as the disks age, a problem that is addressed by array-based
checksums in true storage arrays.
With a real storage array, data integrity is protected by using checksums at multiple levels. If data is
corrupted in an IP-based network, the Transmission Control Protocol (TCP) layer rejects the packet data
and requests retransmission. The FC protocol includes checksums, as does encapsulated SCSI data.
After it is on the array, ONTAP has RAID and checksum protection. Corruption can occur, but, as in most
enterprise arrays, it is detected and corrected. Typically, an entire drive fails, prompting a RAID rebuild,
and database integrity is unaffected. Less often, ONTAP detects a checksum error, meaning that data on
the disk is damaged. The disk is then failed out and a RAID rebuild begins. Once again, data integrity is
unaffected.
The Oracle datafile and redo log architecture is also designed to deliver the highest possible level of data
integrity, even under extreme circumstances. At the most basic level, Oracle blocks include checksum
and basic logical checks with almost every I/O. If Oracle has not crashed or taken a tablespace offline,
then the data is intact. The degree of data integrity checking is adjustable, and Oracle can also be
configured to confirm writes. As a result, almost all crash and failure scenarios can be recovered, and in
the extremely rare event of an unrecoverable situation, corruption is promptly detected.
Most NetApp customers using Oracle databases discontinue the use of RMAN and other backup
products after migrating to snapshot-based backups. There are still options in which RMAN can be used
to perform block-level recovery with SMO. However, on a day-to-day basis, RMAN, NetBackup, and other
products are only used occasionally to create monthly or quarterly archival copies.
Some customers choose to run dbv periodically to perform integrity checks on their existing databases.
NetApp discourages this practice because it creates unnecessary I/O load. As discussed above, if the
database was not previously experiencing problems, the chance of dbv detecting a problem is close to
zero, and this utility creates a very high sequential I/O load on the network and storage system. Unless
there is reason to believe corruption exists, such as exposure to a known Oracle bug, there is no reason
to run dbv.
10 Flash
A comprehensive explanation of the use of flash and SSD technologies with Oracle databases is beyond
the scope of this document, but some common questions and misconceptions must be addressed. All
principles explained in this section apply equally to all protocols and file systems, including Oracle ASM.
10.1 Flash Cache
NetApp Flash Cache intelligent data caching has been the leading flash-based technology used in Oracle
deployments for a simple reason: most databases are limited by random read latency. Flash Cache is a
simple method for accelerating random read performance.
However, Flash Cache does have the limitation of being tied to the particular node that hosts the PCIe
card containing the flash memory. As spindle sizes increase, customers deploy fewer spindles. This
increases the risk that controller failure results in a period of performance degradation while the Flash
Cache card on the takeover node warms up. Until warmup is complete, there can be significantly more
I/O on the spinning media. For this reason, NetApp Flash Pool intelligent data caching is the preferred
flash technology because the flash layer follows the spinning media during takeover. No warm-up time is
required, and the cache does not go cold. This does not mean, however, that configurations with Flash
Cache are flawed. Most systems contain enough spindles to cope with increased I/O during a controller
takeover.
In general, the default settings are optimal for database workloads. Make sure that Flash Cache is
enabled by setting flexscale. enable=on.
30
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
flexscale.lopri_blocks
The flexscale.lopri_blocks parameter applies to the use of Flash Cache intelligent caching. The
default for this option is off, which means that I/O from low-priority block operations such as random
overwrites and sequential I/O are not cached. The reason is simple; most databases are limited by
latency on random-read operations. When a random overwrite occurs, an Oracle database almost always
retains a copy of that block, and the block is highly unlikely to be reread soon. Caching such overwrites
wastes valuable space in Flash Cache. When Oracle performs sequential read I/O, it is a very large block
operation that is processed with inherent efficiency by the storage array, even if the underlying disk is
SATA. This type of I/O does not benefit from Flash Cache. Attempting to cache this I/O generally places
unnecessary load on the CPU and again wastes valuable space in Flash Cache that would be better used
for caching random I/O.
NetApp recommends the following:
•
Only change this parameter after careful consultation with NetApp customer support or professional
services or after thorough testing.
flexscale.read-ahead_blocks
The flexscale.read-head_blocks parameter was added in ONTAP 8.2, and is similar to
flexscale.lopri_blocks except that it only targets read data. Under normal operation, Flash Cache
only stores randomly accessed data. Enabling flexscale.read-ahead_blocks enables caching of
sequentially read data. As discussed above, sequential I/O is already very efficient and does not
generally benefit from being stored in flash memory.
NetApp recommends the following:
•
Only change this parameter after careful consultation with NetApp Customer Support or Professional
Services or after thorough testing.
flexscale.random_write_through_blocks
The flexscale.random_write_through_blocks parameter was added in ONTAP 8.3. Unlike the
prior two options, this parameter has the potential to help Oracle workloads. In most cases, randomly
written data does not need to be cached because the database retains a copy of the block. In situations in
which the Oracle cache is under pressure, randomly written blocks might be read back again quickly and
capturing that data in Flash Cache improves performance.
Refer to the following discussion of Flash Pool write caching for a more complete explanation of the value
of write caching.
NetApp recommends the following:
•
By default, set this parameter to off. However, NetApp recommends experimenting with alternative
settings. A database that is limited by db_file_sequential_read performance is a candidate for
enabling random_write_through_blocks. A decrease in latency is observed if the database
rapidly rereads recently written blocks.
10.2 SSD Aggregates
There is a lot of confusion about the use of SSD and Flash media for redo logs. Good redo logging
performance requires that the data be written to SSD. An SSD drive can be valuable for improving
logging performance when used with directly connected devices, but NetApp storage arrays already
contain nonvolatile, mirrored, NVRAM-based or NVMEM-based solid-state storage. When an Oracle
database performs a write operation, the write is acknowledged as soon as it is journaled into NVRAM or
NVMEM. Write performance is not directly affected by the type of drives that eventually receive the writes.
31
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
At best, use of an SSD aggregate or AFF platform for hosting sequential writes such as redo logging or
for temporary datafile I/O has no effect. There are circumstances where choosing AFF will improve write
performance indirectly, though. For example, a system with heavy random I/O that is overloading
spinning media might reach a point where the drives are no longer able to absorb the incoming writes
quick enough to keep NVMEM/NVRAM from filling up. In these cases, a change to an SSD aggregate or
AFF platform can improve redo performance, but it is an indirect benefit. The write performance problem
would be resolved because the system is better able to process random IO. The write behavior then
returns to normal, with all inbound writes committing to NVMEM/NVRAM without delay.
On occasion, customers have made planning errors which result in performance damage with an SSD
aggregate. Although SSD drives offer far higher performance than spinning media, SSD aggregates
sometimes have far fewer devices than do SAS or SATA aggregates on the system. For example,
NetApp has observed severe performance problems in customer environments caused by moving heavy
sequential-write workloads, including redo logs, from a large SAS aggregate that might contain 100 drives
to a small SSD aggregate with only 4 or 5 devices. SSD drives might be faster than SAS, but they are not
unlimited.
The primary application for an SSD aggregate is servicing random IO workloads. Indexes are particularly
good candidates for placement on SSD drives. Other types of IO should not suffer so long as there is not
an excessively small number of drives in the aggregate, but a performance improvement should not be
expected unless the prior system was badly overloaded.
10.3 Flash Pool
The same principles that underlie Flash Cache and SSD aggregates also apply to Flash Pool intelligent
caching. Flash Pool improves the latency of random reads, which is typically the primary performance
bottleneck in an Oracle database. Flash Pool is also a cost-saving technology. Many Oracle storage
systems have a significant number of spindles that service bursts of random-read activity at minimum
latency. A small Flash Pool allocation can replace a large number of spinning drives.
A further benefit of Flash Pool not realized with Flash Cache is write caching or, specifically, overwrite
caching. Using Flash Pool for write caching does not directly affect write performance because writes
commit to NVRAM or NVMEM first. From a latency perspective, I/O is complete when data is journaled
into NVRAM or NVMEM. The type of media on which an inbound write is subsequently stored does not
affect performance by itself. There can, however, be an indirect benefit to write performance if the use of
Flash Pool write caching decreases pressure on spinning media and thus leads to a general improvement
in I/O performance on the entire array.
Flash Pool write caching also improves read latency in which randomly overwritten blocks are rapidly read
again. This process is not applicable to all databases because the database typically retains a copy of a
written block. As the size of the Oracle buffer cache increases and more writes are cached, it becomes
less likely that the block must be reread from disk. In such cases, it might be preferable to disable write
caching and reserve valuable flash space for random read operations.
On the other hand, there is a benefit to capturing repeated overwrites of the same blocks on the SSD
layer to reduce pressure on spinning media. This issue can occur when the Oracle buffer cache is under
pressure and blocks are aging out of cache only to be read again quickly.
NetApp recommends the following:
•
Retain the default Flash Pool policy, which includes both random-read and random-write caching.
•
Although write caching might not be beneficial, the overall random-write levels observed in most
Oracle databases are not high enough to cause excessive use of SSD space. The defaults make
write caching available if needed.
The primary exception is a database workload with the following characteristics:
32
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
•
The workload dominates an aggregate and is therefore responsible for most of the Flash Pool
caching activity.
•
The workload is known to be limited by random-read latency.
•
Write activity is relatively low.
•
The Oracle buffer cache is relatively large.
In such cases, changing the Flash Pool write cache policy to none might be warranted. Maximum space
would then be available on the SSDs for read caching.
Flash Pool is often useful for Oracle standby databases, including use with Oracle DataGuard, because a
standby database usually lacks a true buffer cache. This situation results in a demanding I/O pattern in
which the same blocks are read, updated, and written repeatedly. Flash Pool captures this concentrated
overwrite activity in the SSD layer, which reduces pressure on spinning media. Prior to the availability of
technologies such as Flash Pool, it was not unusual for a standby database to require more spinning
disks than the primary database that was the source of replication.
10.4 AFF Platforms
NetApp AFF extends the value of SSD aggregates through increased performance and default behavior
tuned for an all-flash platform. Complete documentation is available on the NetApp Support site.
One particular consideration that should be made is that flash is not exclusively about IOPS. There are
other benefits such as consistency and predictability of performance, decreased power consumption,
decreased heat output, and general future-proofing of a solution.
In many cases, an all-flash platform can decrease costs as it avoids the need to deploy drive after drive of
spinning media purely to ensure good latency. Costs continue to decrease dramatically, which leads more
and more customers to select AFF as the default choice.
11 Ethernet Configuration
The TCP/IP settings required for Oracle database software installation are usually sufficient to provide
good performance for all NFS or iSCSI storage resources. In some cases, NetApp has seen performance
benefits in 10Gb environments after implementing specific recommendations from the network adapter
manufacturer.
11.1 Ethernet Flow Control
This technology allows a client to request that a sender temporarily stop data transmission. This is usually
done because the receiver is unable to process incoming data quickly enough. At one time, requesting
that a sender cease transmission was less disruptive than having a receiver discard packets because
buffers were full. This is no longer the case with the TCP stacks used in OSs today. In fact, flow control
causes more problems than it solves.
Performance problems caused by Ethernet flow control have been increasing in recent years. This is
because Ethernet flow control operates at the physical layer. If a network configuration permits any
database server to send an Ethernet flow control request to a storage system, the result is a pause in I/O
for all connected clients. Because an increasing number of clients are served by a single storage
controller, the likelihood of one or more of these clients sending flow control requests increases. The
problem has been seen frequently at customer sites with extensive OS virtualization.
A NIC on a NetApp system should not receive flow-control requests. The method used to achieve this
result varies based on the network switch manufacturer. In most cases, flow control on an Ethernet switch
can be set to receive desired or receive on, which means that a flow control request is not
forwarded to the storage controller. In other cases, the network connection on the storage controller might
not allow flow-control disabling. In these cases, the clients must be configured to never send flow control
33
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
requests, either by changing to the NIC configuration on the database server itself or the switch ports to
which the database server is connected.
NetApp recommends the following:
•
Make sure that NetApp storage controllers do not receive Ethernet flow-control packets. This can
generally be done by setting the switch ports to which the controller is attached, but some switch
hardware has limitations that might require client-side changes instead.
11.2 Jumbo Frames
The use of jumbo frames has been shown to offer some performance improvement in 1Gb networks by
reducing CPU and network overhead, but the benefit is not usually significant. Even so, NetApp
recommends implementing jumbo frames when possible, both to realize any potential performance
benefits and to future-proof the solution.
Using jumbo frames in a 10Gb network is almost mandatory. This is because most 10Gb implementations
reach a packets-per-second limit without jumbo frames before they reach the 10Gb mark. Using jumbo
frames improves efficiency in TCP/IP processing because it allows the database server, NICs, and the
storage system to process fewer larger packets. The performance improvement varies from NIC to NIC,
but it is significant.
In jumbo-frame implementations, there are common but incorrect beliefs that all connected devices must
support jumbo frames and that the MTU size must match end-to-end. Instead, the two network end points
negotiate the highest mutually acceptable frame size when establishing a connection. In a typical
environment, a network switch is set to an MTU size of 9216, the NetApp controller is set to 9000, and the
clients are set to a mix of 9000 and 1514. Clients that can support an MTU of 9000 can use jumbo
frames, and clients that can only support 1514 can negotiate a lower value.
Problems with this arrangement are rare in a completely switched environment. However, take care in a
routed environment that no intermediate router is forced to fragment jumbo frames.
NetApp recommends the following:
•
Jumbo frames are desirable but not required with 1Gb Ethernet (GbE)
•
Jumbo frames are required for maximum performance with 10GbE
11.3 TCP Parameters
Three settings are frequently misconfigured: TCP timestamps, selective acknowledgment (SACK), and
TCP window scaling. Many out-of-date documents on the Internet recommend disabling one or more of
these parameters to improve performance. There was some merit to this recommendation many years
ago when CPU capabilities were much lower and there was a benefit to reducing the overhead on TCP
processing whenever possible.
However, with modern OSs, disabling any of these TCP features usually results in no detectable benefit
or might result in damage performance. Performance damage is especially likely in virtualized networking
environments because these features are required for efficient handling of packet loss and changes in
network quality.
NetApp recommend the following:
•
Enable TCP timestamps, SACK, and TCP window scaling on the host
34
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
12 General NFS Configuration
12.1 NFS Versions
Oracle currently limits NFS support to NFS version 3. For this reason, NetApp does not support the use of
NFSv4, NFSv4.1, or pNFS with Oracle databases. This document will be updated if Oracle’s support
stance changes.
NetApp recommends the following:
•
NFSv3 is mandatory at this time
12.2 TCP Slot Tables
TCP slot tables are the NFS equivalent of host bus adapter (HBA) queue depth. These tables control the
number of NFS operations that can be outstanding at any one time. The default value is usually 16, which
is far too low for optimum performance. The opposite problem occurs on newer Linux kernels, which can
automatically increase the TCP slot table limit to a level that saturates the NFS server with requests.
For optimum performance and to prevent performance problems, adjust the kernel parameters that
control the TCP slot tables.
Run the sysctl -a | grep tcp.*.slot_table command, and observe the following parameters:
# sysctl -a | grep tcp.*.slot_table
sunrpc.tcp_max_slot_table_entries = 128
sunrpc.tcp_slot_table_entries = 128
All Linux systems should include sunrpc.tcp_slot_table_entries, but only some will include
sunrpc.tcp_max_slot_table_entries. They should both be set to 128.
12.3 Installation and Patching
The presence of the following mount options in ORACLE_HOME causes host caching to be disabled:
cio, actimeo=0, noac, forcedirectio.
This action can have a severe negative effect on the speed of Oracle software installation and patching.
Many customers temporarily remove these mount options during installation or patching of the Oracle
binaries. This removal can be performed safely if the user verifies that no other processes are actively
using the target ORACLE_HOME during the installation or patching process.
12.4 ONTAP and NFS Flow Control
Under some circumstances, the use of ONTAP requires changes in the Oracle or Linux kernel parameter.
The reason is related to NFS flow control, so do not confuse these changes with Ethernet flow control.
NFS flow control enables an NFS server such as ONTAP to limit network communication with an NFS
client that is not acknowledging receipt of data. This capability protects the NFS server in cases in which
a malfunctioning NFS client requests data at a rate beyond its ability to process the responses. Without
protection, the network buffers on the NFS server fill up with unacknowledged packets.
Under rare circumstances, I/O bursts from both Oracle DNFS clients and newer Linux NFS clients can
exceed the limits at which the ONTAP NFS server can protect itself. The NFS client lags in its processing
of inbound data while continuing to send requests for more data. This lag can lead to performance and
stability problems with NFS connectivity.
Although problems are rare, NetApp recommends the following protective measures as best practices.
These measures apply only to ONTAP, and the changes should not adversely affect performance.
NetApp recommends the following with ONTAP:
35
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
•
When Oracle DNFS is used, set the DNFS_BATCH_SIZE parameter to 128. This parameter is
available with Oracle 11.2.0.4 and later. If this cannot be done, do not use DNFS.
•
Make sure that both of the TCP slot tables parameters discussed previously are set to 128.
•
The Oracle calibrate_io command does not work if DNFS_BATCH_SIZE is set to anything other
than the default value. If I/O needs to be calibrated, temporarily remove the DNFS_BATCH_SIZE
parameter during calibration.
12.5 Direct NFS
Oracle’s DNFS client is designed to bypass the host NFS client and perform NFS file operations directly
on an NFS server. Enabling it only requires changing the Oracle Disk Manager library. Instructions for this
process are provided in the Oracle documentation.
Using DNFS results in a general improvement in I/O performance and decreases the load on the host and
the storage system because I/O is performed in the most efficient way possible. In addition, Oracle DNFS
provides multipathing and fault-tolerance. For example, two 10Gb interfaces can be bound together to
offer 20Gb of bandwidth. A failure of one interface results in I/Os being retried on the other interface. The
overall operation is very similar to FC multipathing.
When DNFS is used, it is critical that all patches described in Oracle Doc 1495104.1 are installed. If a
patch cannot be installed, the environment must be evaluated to make sure that the bugs described in
that document do not cause problems. In some cases, an inability to install the required patches prevents
the use of DNFS.
Caution
•
Before using DNFS, verify that the patches described in Oracle Doc 1495104.1 are installed.
•
Starting with Oracle 12c, DNFS includes support for NFSv3, NFSv4, and NFSv4.1. NetApp support
policies cover v3 and v4 for all clients, but at the time of writing NFSv4.1 is not supported for use
with Oracle DNFS.
•
Do not use DNFS with any type of round-robin name resolution, including DNS, DDNS, NIS or any
other method. This includes the DNS load balancing feature available in ONTAP. When an Oracle
database using DNFS resolves a host name to an IP address it must not change on subsequent
lookups. This can result in Oracle database crashes and possible data corruption.
12.6 Direct NFS and Host File System Access
Using DNFS can occasionally cause problems for applications or user activities that rely on the visible file
systems mounted on the host because the DNFS client accesses the file system out of band from the
host OS. The DNFS client can create, delete, and modify files without the knowledge of the OS.
When the mount options for single-instance databases are used, they enable caching of file and directory
attributes, which also means that the contents of a directory are cached. Therefore, DNFS can create a
file, and there is a short lag before the OS rereads the directory contents and the file becomes visible to
the user. This is not generally a problem, but, on rare occasions, utilities such as SAP BR*Tools might
have issues. If this happens, address the problem by changing the mount options to use the
recommendations for Oracle RAC. This change results in the disabling of all host caching.
Only change mount options when (a) DNFS is used and (b) a problem results from a lag in file visibility. If
DNFS is not in use, using Oracle Real Application Cluster (RAC) mount options on a single-instance
database results in degraded performance.
Note:
36
See the note about nosharecache in the section “Linux NFSv3 Mount Options” for a Linuxspecific DNFS issue that can produce unusual results.
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
12.7 ADR and NFS
Some customers have reported performance problems resulting from an excessive amount of I/O on data
in the ADR location. The problem does not generally occur until a lot of performance data has
accumulated. The reason for the excessive I/O is unknown, but this problem appears to be a result of
Oracle processes repeatedly scanning the target directory for changes.
Removal of the noac and/or actimeo=0 mount options allows host OS caching to occur and reduces
storage I/O levels.
NetApp recommends the following:
•
Do not place ADR data on a file system with noac or actimeo=0 because performance problems are
likely. Separate ADR data into a different mount point if necessary.
13 General SAN Configuration
13.1 Zoning
An FC zone should never contain more than one initiator. Such an arrangement might appear to work
initially, but crosstalk between initiators eventually interferes with performance and stability.
Multitarget zones are generally regarded as safe, although in rare circumstances the behavior of FC
target ports from different vendors has caused problems. For example, avoid including the target ports
from both a NetApp and an EMC storage array in the same zone. In addition, placing a NetApp storage
system and a tape device in the same zone is even more likely to cause problems.
13.2 LUN Alignment
LUN alignment refers to optimizing I/O with respect to the underlying file system layout. On a NetApp
system, storage is organized in 4KB units. Align an 8KB block on an Oracle datafile to exactly two 4KB
blocks. If an error in LUN configuration shifts the alignment by 1KB in either direction, each 8KB Oracle
block would exist on three different 4KB storage blocks rather than two. This arrangement would cause
increased latency and cause additional I/O to be performed within the storage system.
LUN alignment is generally only a concern when a logical volume manager is not used. As a practical
matter, this means that Linux and Solaris are of primary concern. If a physical volume within a logical
volume group is defined on the whole disk device (no partitions are created), the first 4KB block on the
LUN aligns with the first 4KB block on the storage system. This is a correct alignment. Problems arise
with partitions because they shift the starting location where the OS uses the LUN. As long as the offset is
shifted in whole units of 4KB, the LUN is aligned.
In Linux environments, build logical volume groups on the whole disk device. When a partition is required,
check alignment by running fdisk –u and verifying that the start of each partition is a multiple of eight.
This means that the partition starts at a multiple of eight 512-byte sectors, which is 4KB.
Also see the discussion about compression block alignment in the section “Compression.” Any layout that
is aligned with 8KB compression block boundaries is also aligned with 4KB boundaries.
Alignment in Solaris environments is more complicated. Refer to the appropriate Host Utilities
documentation for more information.
Caution
In Solaris x86 environments, take additional care about proper alignment because most configurations
have several layers of partitions. Solaris x86 partition slices usually exist on top of a standard master
boot record partition table.
37
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
13.3 LUN Misalignment Warnings
Oracle redo logging normally generates unaligned I/O that can cause misleading warnings about
misaligned LUNs on ONTAP. Oracle redo logging performs a sequential overwrite of the redo log file with
writes of varying size. A log write operation that does not align to 4KB boundaries does not ordinarily
cause performance problems because the next log write operation completes the block. The result is that
ONTAP is able to process almost all writes as complete 4KB blocks, even though the data in some 4KB
blocks was written in two separate operations.
Verify alignment by using by using utilities such as sio or dd that can generate I/O at a defined block
size. The I/O alignment statistics on the storage system can be viewed with the stats command. See
"Appendix B: WAFL Alignment Verification” for more information.
13.4 LUN Sizing
A LUN is a virtualized object on ONTAP that exists across all of the spindles in the hosting aggregate. As
a result, the performance of the LUN is unaffected by its size because the LUN draws on the full potential
of the aggregate no matter which size is chosen.
As a matter of convenience, customers might wish to use a LUN of a particular size. For example, if a
database is built on an ASM disk group composed of two LUNS of 1TB each, then that ASM disk group
must be grown in increments of 1TB. It might be preferable to build the ASM disk group from eight LUNs
of 500GB each so that the disk group can be increased in smaller increments.
The practice of establishing a universal standard LUN size is discouraged, because doing so can
complicate manageability. For example, a standard LUN size of 100GB might work well when a database
is in the range of 1TB to 2TB, but a database 20TB in size would require 200 LUNs. This means that
server reboot times are longer, there are more objects to manage in the various UIs, and products such
as SMO must perform discovery on many objects. Using fewer, larger LUNs avoids such problems.
Note:
•
The LUN count is more important than the LUN size.
•
LUN size is mostly controlled by LUN count requirements.
•
Avoid creating more LUNs than required.
13.5 LUN Resizing and LVM-Based Resizing
When a SAN-based file system has reached its capacity limit, there are two options for increasing the
space available:
•
Increase the size of the LUNs.
•
Add a LUN to an existing volume group and grow the contained logical volumes.
Both options are supportable, but increasing a LUN size is generally more difficult and can be risky. Some
of the considerations are as follows:
•
LUNs created by ONTAP can be increased to approximately 10X of their original size. The limitation
is based on the inherent structure of disk geometry. There are sometimes options to increase beyond
the 10X mark, but it can require changes to partition tables that require advanced understanding of
disk configuration at the host level.
•
It is recommended that a database be shut down before attempting a LUN resize, and to create
Snapshot copies as a fallback measure. Although this is not a requirement in all cases, there is some
risk of disruption and risk of user error when rediscovering the newly enlarged LUNs at the host OS
level.
•
One exception to LUN resizing complications is Microsoft Windows, which offers a safe and
nondisruptive method of increasing LUN sizes using NetApp SnapDrive for Windows.
38
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
Although LUN resizing is an option to increase capacity, it is generally better to use a logical volume
manager (LVM), including Oracle ASM. One of the principle reasons LVMs exist is to avoid the need for
a LUN resize. With an LVM, multiple LUNs are bonded together into a virtual pool of storage. The logical
volumes carved out of this disk pool are managed by the LVM and can be easily resized. An additional
benefit is the avoidance of hotspots on a particular disk by distributing a given logical volume across all
available LUNs. Transparent migration can usually be performed by using the volume manager to
relocate the underlying extents of a logical volume to new LUNs.
For the above reasons, a strategy involving resizing LUNs is discouraged in favor of an LVM approach.
13.6 LUN Count
Unlike the LUN size, the LUN count does affect performance. Oracle database performance is affected by
the capability to perform parallel I/O through the SCSI layer. As a result, two LUNs offer better
performance than a single LUN. Using a logical volume manager, such as Veritas VxVM, Linux LVM2, or
Oracle ASM is the simplest method to increase parallelism.
NetApp customers have generally experienced minimal benefit from increasing the number of LUNs
beyond eight, although the testing of 100%-SSD environments with very heavy random I/O has
demonstrated further improvement up to 64 LUNs. NetApp recommends building a volume group with an
extent size that enables the even distribution of I/O. For example, a 1TB volume group composed of 10
100GB LUNs and an extent size of 100MB would yield 10,000 extents in total (1,000 extents per LUN).
The resulting I/O on a database placed on this 1TB volume group should be evenly distributed across all
10 LUNs.
Distributing a logical volume across extents is not the same as striping, although the concept is similar.
Logical volume managers break up LUNs into relatively large extents in order to make data management
simpler. Larger extents are preferred because they deliver better efficiency of readahead operations. For
example, an Oracle ASM extent, also called an Allocation Unit, of 64MB allows storage array readahead
to assist transferring a full 64MB of data before ASM moves to the next extent. Smaller allocation units
mean readahead must reset more often as read operations move from extent to extent.
The more important random IO should still be evenly distributed across LUNs even with a large extent
size unless the database has extremely concentrated IO.
In contrast to distributed extents, true striping should be avoided. Striping was mostly targeting relatively
slow spinning disk technology. For example, if an application was known to read in 1MB chunks, a stripe
set could be created with eight LUNs with a stripe width of 128KB. The result is a 1MB operation could be
executed as eight simultaneous 128KB IOs on each LUN. This is almost never beneficial with a modern
database and storage system. Furthermore, incorrect tuning of a striped volume group will result in
damage to performance.
Most databases are limited by random I/O performance, not sequential performance. A datafile that exists
across a large number of extents enables a large amount of random I/O to be randomized across many
extents. This arrangement means that all LUNs in the volume group are used evenly and no individual
LUN limits performance.
NetApp recommends the following:
•
In general, four to eight LUNs are sufficient to support datafile I/O. Less than four LUNs might create
performance limitations because of limitations in host SCSI implementations.
•
Do not use fine-grained striping. Instead, enable an LVM policy that distributes data across large
extents on each LUN to ensure each datafile is spread across all available LUNs.
39
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
13.7 Datafile Block Size
Some OSs offer a choice of file system block sizes. For file systems supporting datafiles, the block size
should be 8KB when compression is used. When compression is not required, a block size of either 8KB
or 4KB can be used.
Some OSs offer a choice of file system block sizes. For file systems supporting datafiles, the block size
should be 4KB. If a datafile is placed on a file system with a 512-byte block, misaligned files are possible.
The LUN and the file system might be properly aligned based on NetApp recommendations, but the file
I/O would be misaligned. Such a misalignment would cause severe performance problems.
See additional information on the relationship between block sizes and compression in the section
“ONTAP 8.3.10.”
13.8 Redo Block Size
File systems supporting redo logs must use a block size that is a multiple of the redo block size. This
generally requires that both the redo log file system and the redo log itself use a block size of 512 bytes.
At very high redo rates, it is possible that 4KB block sizes perform better, because high redo rates allow
I/O to be performed in fewer and more efficient operations. If redo rates are greater than 50MBps,
consider testing a 4KB block size.
A few customer problems have been identified with databases using redo logs with a 512-byte block size
on a file system with a 4KB block size and many very small transactions. The overhead involved in
applying multiple 512-byte changes to a single 4KB file system block led to performance problems that
were resolved by changing the file system to use a block size of 512 bytes.
NetApp recommends the following:
•
Do not change the redo block size unless advised by a relevant customer support or professional
services organization or the change is based on official product documentation.
14 Virtualization
14.1 Overview
Virtualization of databases with VMware ESX, Oracle OVM, or KVM is an increasingly common choice for
NetApp customers who chose virtualization for even their most mission-critical databases.
Many misconceptions exist on the support policies for virtualization, particularly for VMware products.
Indeed, it is not uncommon to hear that Oracle does not support virtualization in any way. This notion is
incorrect and leads to missed opportunities for virtualization. Oracle Doc ID 249212.1 discusses known
issues in an Oracle environment and also specifies support for RAC.
A customer with a problem unknown to Oracle might be asked to reproduce the problem on physical
hardware. An Oracle customer running a bleeding-edge version of a product might not want to use
virtualization because of the potential for new bug discovery. However, this situation has not been a
problem in practice for virtualization customers using generally available product versions.
14.2 Storage Presentation
Customers considering virtualization of their databases should base their storage decisions on their
business needs. Although this is a generally true statement for all IT decisions, it is especially important
for virtualization, because the size and scope of projects vary considerably.
Regarding storage presentation, a storage resource should be managed directly by the VM guest.
Therefore, use one of the following storage configurations:
40
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
•
iSCSI LUNs managed by the iSCSI initiator on the VM, not the hypervisor
•
NFS file systems mounted by the VM, not a virtual machine disk (VMDK)
•
FC raw device mappings (RDMs) when the VM guest manages the file system
As a general rule, avoid using datastores for Oracle files. There are many reasons for this
recommendation:
•
Transparency. When a VM owns its file systems, it is easier for a database administrator or a system
administrator to identify the source of the file systems for their data.
•
Performance. Testing has shown that there is a performance effect from channeling all I/O through a
hypervisor datastore.
•
Manageability. When a VM owns its file systems, the use or nonuse of a hypervisor layer affects
manageability. The same procedures for provisioning, monitoring, data protection, and so on can be
used across the entire estate, including both virtualized and nonvirtualized environments.
•
Stability and troubleshooting. When a VM owns its file systems, delivering good, stable
performance and troubleshooting problems are much simpler because the entire storage stack is
present on the VM. The hypervisor's only role is to transport FC or IP frames. When a datastore is
included in a configuration, it complicates the configuration by introducing another set of timeouts,
parameters, log files, and potential bugs.
•
Portability. When a VM owns its file systems, the process of moving an Oracle environment
becomes much simpler. File systems can easily be moved between virtualized and nonvirtualized
guests.
•
Vendor lock-in. After data is placed in a datastore, leveraging a different hypervisor or take the data
out of the virtualized environment entirely becomes very difficult.
•
Snapshot enablement. In some cases, backups in a virtualized environment can become a problem
because of the relatively limited bandwidth. For example, a four-port 10GbE trunk might be sufficient
to support day-to-day performance needs of many virtualized databases. However, such a trunk
would be insufficient to perform backups using RMAN or other backup products that require
streaming a full-sized copy of the data.
Using VM-owned file systems makes it easier to leverage Snapshot-based backups and restores. A
VM-owned file system offloads the work of performing backups onto the storage system. There is no
need to overbuild the hypervisor configuration purely to support the bandwidth and CPU requirements
in the backup window.
NetApp recommends the following:
•
For optimum performance and manageability, avoid placing Oracle data on a datastore. Use guestowned file systems such as NFS or iSCSI file systems managed by the guest or with RDMs.
14.3 Paravirtualized Drivers
For optimum performance, the use of par virtualized network drivers is critical. When a datastore is used,
a paravirtualized SCSI driver is required. A paravirtualized device driver allows a guest to integrate more
deeply into the hypervisor, as opposed to an emulated driver in which the hypervisor spends more CPU
time mimicking the behavior of physical hardware.
The performance of most databases is limited by storage. Therefore, the extra latency introduced by a
network or SCSI driver is particularly noticeable. NetApp Customer Support has encountered many
performance complaints that were resolved by installing paravirtualized drivers. During one customer
proof of concept, databases showed better performance under ESX than with the same hardware running
as bare metal. The tests were very I/O intensive, and the performance difference was attributed to the
use of the ESX paravirtualized network drivers.
NetApp recommends the following:
•
Always use paravirtualized network drivers and SCSI drivers.
41
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
14.4 Overcommitting RAM
Overcommitting RAM means configuring more virtualized RAM on various hosts than exists on the
physical hardware. Doing so can cause unexpected performance problems. When virtualizing a
database, the underlying blocks of the Oracle SGA must not be swapped out to disk by the hypervisor.
Doing so causes highly unstable performance results.
NetApp recommends the following:
•
Do not configure a hypervisor in a way that allows Oracle SGA blocks to be swapped out.
15 Clustering
15.1 Oracle Real Application Clusters
This section applies to Oracle 10.2.0.2 and later. For earlier versions of Oracle, consult Oracle Doc ID
294430.1 in conjunction with this document to determine optimal settings.
disktimeout
The primary storage-related RAC parameter is disktimeout. This parameter controls the threshold
within which voting file I/O must complete. If the disktimeout parameter is exceeded, the RAC node is
evicted from the cluster. The default for this parameter is 200. This value should be sufficient for standard
storage takeover and giveback procedures.
NetApp strongly recommends testing RAC configurations thoroughly before placing them into production
because many factors affect a takeover or giveback. In addition to the time required for storage failover to
complete, additional time is also required for Link Aggregation Control Protocol (LACP) changes to
propagate, SAN multipathing software must detect an I/O timeout and retry on an alternate path, and, if a
database is extremely active, a large amount of I/O must be queued and retried before voting disk I/O is
processed.
If an actual storage takeover or giveback cannot be performed, the effect can be simulated with cable pull
tests on the database server.
NetApp recommends the following:
•
Leave the disktimeout parameter at the default value of 200.
•
Always test a RAC configuration thoroughly.
misscount
The misscount parameter normally affects only the network heartbeat between RAC nodes. The default
is 30 seconds. If the grid binaries are on a storage array or the OS boot disk is not local, this parameter
might become important. This includes hosts with boot disks located on an FC SAN, NFS-booted OSs,
and boot disks located on virtualization datastores such as a VMDK file.
If access to a boot disk is interrupted by a storage takeover or giveback, it is possible that the grid binary
location or the entire OS temporarily hangs. The time required for ONTAP to complete the storage
operation and for the OS to change paths and resume I/O might exceed the misscount threshold. As a
result, a node immediately evicts after connectivity to the boot LUN or grid binaries is restored. In most
cases, the eviction and subsequent reboot occur with no logging messages to indicate the reason for the
reboot. Not all configurations are affected, so test any SAN-booting, NFS-booting, or datastore-based
host in a RAC environment so that RAC remains stable if communication to the boot disk is interrupted.
42
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
In the case of nonlocal boot disks or the file system hosting grid binaries, misscount might need to be
changed to match disktimeout. If this parameter is changed, conduct further testing to also identify
any effects on RAC behavior, such as node failover time.
NetApp recommends the following:
•
Leave the misscount parameter at the default value of 30 unless one of the following conditions
applies:
•

grid binaries are located on a network-attached disk, including NFS, iSCSI, FC, and datastorebased disks

The OS is SAN booted
In such cases, evaluate the effect of network interruptions that affect access to OS or GRID_HOME file
systems. In some cases, such interruptions cause the Oracle RAC daemons to stall, which can lead
to a misscount-based timeout and eviction. The timeout defaults to 27 seconds, which is the value
of misscount minus reboottime. In such cases, increase misscount to 200 to match
disktimeout.
15.2 Solaris Clusters
Solaris clusters, an active-passive clustering technology, are much more highly integrated than other
clusterware options. This technology provides an almost plug-and-play capability to easily deployed
databases and applications as clustered resources and allows them to be easily moved around the
cluster (including associated IP addresses, configuration files, and storage resources). As a result of this
tight integration, Oracle has a rigid qualification procedure for Solaris clusters to make sure that all of the
components work properly together.
ONTAP provides broad support for Solaris clusters in a SAN environment. Consult the Interoperability
Matrix Tool (IMT) for further information.
In an NFS environment, support is limited. There is no supportability barrier with NFS in general, (for
example, the use of automounted NFS home directories), but databases cannot be placed under control
of Solaris clusters. Previously, an NFS agent was available, but support for this product ended in October
of 2012. Although it is possible to use the native ability of Solaris clusters to build a custom service that
can be clustered, this is probably not feasible for most deployments. The reason is the time and effort
required to write scripts that manage resources, including storage.
15.3 Veritas Cluster Server
Veritas Cluster Server (VCS) is similar to Solaris clusters in that it allows users to package a database or
application as a deployed service and deploy it on a cluster in an active-passive manner.
VCS and SAN
ONTAP provides broad support for VCS clustering in a SAN environment. Consult the Interoperability
Matrix Tool (IMT) for further information.
VCS and NFS
At one time, an NFS client was available from NetApp to provide quorum, monitoring, management,
fencing, and NFS lock-breaking capabilities. However, support was discontinued in October of 2012
primarily because these capabilities were no longer required. VCS can now natively manage and monitor
NFS file systems. Multiple options exist for quorum management in a NAS environment that do not
require an agent.
43
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
VCS and NFS Fencing
One consideration for any active-passive clustering is fencing, which means that storage resources are
available to only one node in the cluster. In a SAN environment, fencing usually means using SCSI
persistent reservations, which allows a node to claim exclusive control of a LUN. In an NFS context, it
means changing the export options for a file system to make it impossible to access a resource on more
than one node. The difference is that, in a SAN environment, fencing is performed by a node in the
cluster laying claim to a storage resource. In a NAS environment, the fencing must be performed on the
storage system.
Fencing with NFS is not strictly necessary. It is much more important to have fencing in a SAN
environment because the simple act of mounting a SAN file system on more than one server generally
corrupts data immediately. NFS is a clustered file system, which means that multiple servers can mount a
file system without problems.
Many customers use active-passive clustering with VCS and similar products such as HP ServiceGuard
and IBM PowerHA without any fencing in place. They trust the cluster software itself to make sure that a
resource runs on only one node. When fencing is desired, it can be deployed as part of the cluster
resource with a small scripting effort.
When a service starts up, it issues a command to the storage system to (a) cut off access for the target
file systems to all nodes and then (b) grant access to the one node on which the service starts. Therefore,
only one node is able to perform I/O on the target file systems. When a service shuts down, it issues a
command to the storage system to remove its access. Other variants exist, but this is the most
comprehensive approach.
Assistance with these systems is available from NetApp Professional Services. Contact your NetApp
representative for more information.
VCS and NFS Lock Breaking
NFS locks in an Oracle environment are a form of fencing. An Oracle database does not start if it finds an
NFS lock in place on the target files. In a VCS environment, NFS locking generally interferes with the
normal functioning of the VCS cluster. The only time locks must be broken is when one node takes over
the services of another node that has not shut down gracefully. During a clean shutdown of an Oracle
database, locks are removed. If the node crashes, locks are left in place and must be cleared before the
database can be restarted.
Most customers choose to disable NFS locking by including the appropriate NFS mount option that
prevents locks from being created in the first place. If this is not desirable, lock breaking can be scripted.
As with fencing, assistance for lock-break scripting is available from NetApp Professional Services, and,
in some cases, fully supported options might be available through Rapid Response Engineering. Contact
your NetApp representative for more information.
16 IBM AIX
This section addresses configuration topics specific to the IBM AIX operating system.
16.1 Concurrent I/O
Achieving optimum performance on IBM AIX requires the use of concurrent I/O. Without concurrent I/O,
performance limitations are likely because AIX performs serialized, atomic I/O, which incurs significant
overhead.
Originally, NetApp recommended using the cio mount option to force the use of concurrent I/O on the
file system, but this process had drawbacks and is no longer required. Since the introduction of AIX 5.2
44
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
and Oracle 10gR1, Oracle on AIX can open individual files for concurrent IO, as opposed to forcing
concurrent I/O on the entire file system.
The best method for enabling concurrent I/O is to set the init.ora parameter
filesystemio_options to setall. Doing so allows Oracle to open specific files for use with
concurrent I/O.
Using cio as a mount option forces the use of concurrent I/O, which can have negative consequences.
For example, forcing concurrent I/O disables readahead on file systems, which can damage performance
for I/O occurring outside the Oracle database software, such as copying files and performing tape
backups. Furthermore, products such as Oracle GoldenGate and SAP BR*Tools are not compatible with
using the cio mount option with certain versions of Oracle.
NetApp recommends the following:
•
Do not use the cio mount option at the file system level. Rather, enable concurrent I/O through the
use of filesystemio_options=setall.
•
Only use the cio mount option should if it is not possible to set filesystemio_options=setall.
16.2 AIX NFSv3 Mount Options
Table 1 and Table 2 list the AIX NFSv3 mount options.
Table 1) AIX NFSv3 mount options—single instance.
File Type
Mount Options
ADR_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536
Control files
Datafiles
Redo logs
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536
ORACLE_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,intr
Table 2) AIX NFSv3 mount options—RAC.
File Type
Mount Options
ADR_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536
Control files
Datafiles
Redo logs
rw,bg,hard,vers=3,proto=cp,timeo=600,rsize=65536,wsize=65536,nointr, noac
CRS/Voting
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,nointr,noac
Dedicated ORACLE_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536
Shared ORACLE_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,nointr
The primary difference between single-instance and RAC mount options is the addition of noac to the
mount options. This addition has the effect of disabling the host OS caching that enables all instances in
the RAC cluster to have a consistent view of the state of the data.
Although using the cio mount option and the init.ora parameter filesystemio_options=setall
has the same effect of disabling host caching, it is still necessary to use noac. Noac is required for
shared ORACLE_HOME deployments to facilitate the consistency of files such as Oracle password files and
45
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
spfile parameter files. If each instance in a RAC cluster has a dedicated ORACLE_HOME, then this
parameter is not required.
16.3 AIX jfs/jfs2 Mount Options
Table 3 lists the AIX jfs/jfs2 mount options.
Table 3) AIX jfs/jfs2 mount options—single instance.
File Type
Mount Options
ADR_HOME
Defaults
Control files
Data files
Redo logs
Defaults
ORACLE_HOME
Defaults
Before using AIX hdisk devices in any environment, including databases, check the parameter
queue_depth. This parameter is not the HBA queue depth; rather it relates to the SCSI queue depth of
the individual hdisk device. Depending on how the LUNs are configured, the value for queue_depth
might be too low for good performance. Testing has shown the optimum value to be 64.
17 HP-UX
This section addresses configuration topics specific to the HP-UX operating system.
17.1 HP-UX NFSv3 Mount Options
Table 4 lists the HP-UX NFSv3 mount options.
Table 4) HP-UX NFSv3 mount options—single instance.
File Type
Mount Options
ADR_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,suid
Control files
Datafiles
Redo logs
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,forcedirecti
o,nointr,suid
ORACLE_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,suid
Table 5) HP-UX NFSv3 mount options—RAC
File Type
Mount Options
ADR_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,noac,suid
Control files
Datafiles
Redo logs
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,nointr,noac
forcedirectio, suid
CRS/Voting
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,nointr,noac,
forcedirectio,suid
46
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
File Type
Mount Options
Dedicated ORACLE_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,suid
Shared ORACLE_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,nointr,noac,
suid
The primary difference between single-instance and RAC mount options is the addition of noac and
forcedirectio to the mount options. This addition has the effect of disabling host OS caching, which
enables all instances in the RAC cluster to have a consistent view of the state of the data. Although using
the init.ora parameter filesystemio_options=setall has the same effect of disabling host
caching, it is still necessary to use noac and forcedirectio.
The reason noac is required for shared ORACLE_HOME deployments is to facilitate consistency of files
such as Oracle password files and spfiles. If each instance in a RAC cluster has a dedicated
ORACLE_HOME, this parameter is not required.
17.2 HP-UX VxFS Mount Options
Use the following mount options for file systems hosting Oracle binaries:
delaylog,nodatainlog
Use the following mount options for file systems containing datafiles, redo logs, archive logs, and control
files in which the version of HP-UX does not support concurrent I/O:
nodatainlog,mincache=direct,convosync=direct
When concurrent I/O is supported (VxFS 5.0.1 and later, or with the ServiceGuard Storage Management
Suite), use these mount options for file systems containing datafiles, redo logs, archive logs, and control
files:
delaylog,cio
Note:
The parameter db_file_multiblock_read_count is especially critical in VxFS
environments. Oracle recommends that this parameter remain unset in Oracle 10g R1 and later
unless specifically directed otherwise. The default with an Oracle 8KB block size is 128. If the
value of this parameter is forced to 16 or less, remove the convosync=direct mount option
because it can damage sequential I/O performance. This step damages other aspects of
performance and should only be taken if the value of db_file_multiblock_read_count
must be changed from the default value.
18 Linux
This section addresses configuration topics specific to the Linux OS.
18.1 Linux NFS
Slot Tables
NFS performance on Linux depends on a parameter called tcp_slot_table_entries. This parameter
regulates the number of outstanding NFS operations that are permitted on a Linux OS.
The default in most 2.6-derived kernels, which includes RH5 and OL5, is 16, and this default frequently
causes performance problems. The opposite problem occurs on newer kernels in which the
tcp_slot_table_entries value is uncapped and can cause storage problems by flooding the system
with excessive requests.
47
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
The solution is to set this value statically. Use a value of 128 for any Linux OS using NetApp NFS storage
with an Oracle database.
To set this value in RHEL 6.2 and earlier, place the following entry in /etc/sysctl.conf:
sunrpc.tcp_slot_table_entries = 128
In addition, there is a bug in most Linux distributions using 2.6 kernels. The startup process reads the
contents of /etc/sysctl.conf before the NFS client is loaded. As a result, when the NFS client is
eventually loaded, it takes the default value of 16. To avoid this problem, edit /etc/init.d/netfs to
call /sbin/sysctl -p in the first line of the script so that tcp_slot_table_entries is set to 128
before NFS mounts any file systems.
To set this value in RHEL 6.3 and later, apply the following modification in the RPC configuration file:
echo "options sunrpc udp_slot_table_entries=64 tcp_slot_table_entries=128
tcp_max_slot_table_entries=128" >> /etc/modprobe.d/sunrpc.conf
18.2 Linux NFSv3 Mount Options
Table 6 and Table 7 list the Linux NFSv3 mount options.
Table 6) Linux NFSv3 mount options—single instance.
File Type
Mount Options
ADR_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536
Control files
Datafiles
Redo logs
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,nointr
ORACLE_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,nointr
Table 7) Linux NFSv3 mount options—RAC.
File Type
Mount Options
ADR_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,actimeo=0
Control files
Data files
Redo logs
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,nointr,actime
o=0
CRS/voting
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,nointr,noac,a
ctimeo=0
Dedicated ORACLE_HOME rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536
Shared ORACLE_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,nointr,actime
o=0
The primary difference between single-instance and RAC mount options is the addition of actimeo=0 to
the mount options. This addition has the effect of disabling the host OS caching, which enables all
instances in the RAC cluster to have a consistent view of the state of the data. Although using the
init.ora parameter filesystemio_options=setall has the same effect of disabling host
caching, it is still necessary to use actimeo=0.
48
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
The reason actimeo=0 is required for shared ORACLE_HOME deployments is to facilitate consistency of
files such as the Oracle password files and spfiles. If each instance in a RAC cluster has a dedicated
ORACLE_HOME, then this parameter is not required.
Generally, nondatabase files should be mounted with the same options used for single-instance datafiles,
although specific applications might have different requirements. Avoid the mount options noac and
actimeo=0 if possible because these options disable file system-level readahead and buffering. This can
cause severe performance problems for processes such as extraction, translation, and loading.
ACCESS and GETATTR
Some customers have noted that an extremely high level of other IOPS such as ACCESS and GETATTR
can dominate their workloads. In extreme cases, operations such as reads and writes can be as low as
10% of the total. This is normal behavior with any database that includes using actimeo=0 and/or noac
on Linux because these options cause the Linux OS to constantly reload file metadata from the storage
system. Operations such as ACCESS and GETATTR are low-impact operations that are serviced from
the ONTAP cache in a database environment. They should not be considered genuine IOPS, such as
reads and writes, that create true demand on storage systems. These other IOPS do create some load,
however, especially in RAC environments. To address this situation, enable DNFS, which bypasses the
OS buffer cache and avoids these unnecessary metadata operations.
Linux Direct NFS
One additional mount option, called nosharecache, is required when (a) DNFS is enabled and (b) a
source volume is mounted more than once on a single server (c) with a nested NFS mount. This
configuration is seen primarily in environments supporting SAP applications. For example, a single
volume on a NetApp system could have a directory located at /vol/oracle/base and a second at
/vol/oracle/home. If /vol/oracle/base is mounted at /oracle and /vol/oracle/home is
mounted at /oracle/home, the result is nested NFS mounts that originate on the same source.
The OS can detect the fact that /oracle and /oracle/home reside on the same volume, which is the
same source file system. The OS then uses the same device handle for accessing the data. Doing so
improves the use of OS caching and certain other operations, but it interferes with DNFS. If DNFS must
access a file, such as the spfile, on /oracle/home, it might erroneously attempt to use the wrong
path to the data. The result is a failed I/O operation. In these configurations, add the nosharecache
mount option to any NFS file system that shares a source FlexVol volume with another NFS file system
on that host. Doing so forces the Linux OS to allocate an independent device handle for that file system.
Linux Direct NFS and Oracle RAC
The use of DNFS has special performance benefits for Oracle RAC on the Linux OS because Linux does
not have a method to force direct I/O, which is required with RAC for coherency across the nodes. As a
workaround, Linux requires the use of the actimeo=0 mount option, which causes file data to expire
immediately from the OS cache. This option in turn forces the Linux NFS client to constantly reread
attribute data, which damages latency and increases load on the storage controller.
Enabling DNFS bypasses the host NFS client and avoids this damage. Multiple customers have reported
significant performance improvements on RAC clusters and significant decreases in ONTAP load
(especially with respect to other IOPS) when enabling DNFS.
Linux Direct NFS and oranfstab File
When using DNFS on Linux with the multipathing option, multiple subnets must be used. On other OSs,
multiple DNFS channels can be established by using the LOCAL and DONTROUTE options to configure
multiple DNFS channels on a single subnet. However, this does not work properly on Linux and
49
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
unexpected performance problems can result. With Linux, each NIC used for DNFS traffic must be on a
different subnet.
18.3 General Linux SAN Configuration
Compression Alignment—Partitions
Compression requires alignment to 8KB disk boundaries for optimum results. Check alignment by using
the fdisk utility with the -u option to view a disk based in sectors. See the following example:
[root@jfs0 etc]# fdisk -l -u /dev/sdb
Disk /dev/sdb: 10.7 GB, 10737418240 bytes
64 heads, 32 sectors/track, 10240 cylinders, total 20971520 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 4096 bytes
I/O size (minimum/optimal): 4096 bytes / 65536 bytes
Disk identifier: 0xb97f94c1
Device Boot
Start
End
Blocks
Id System
/dev/sdb1
36
20971519
10485742
83 Linux
Partition 1 does not start on physical sector boundary.
This partition is not 8KB aligned. Rather, the partition has an offset of 36 sectors. This offset aligns to a
4KB boundary, which is generally required for good performance but does not align to an 8KB boundary.
The start of a partition should be a multiple of 16 sectors (512 bytes * 16 = 8192) so that the partition is
aligned.
This example shows a correctly aligned partition:
[root@jfs0 etc]# fdisk -l -u /dev/sdb
Disk /dev/sdb: 10.7 GB, 10737418240 bytes
64 heads, 32 sectors/track, 10240 cylinders, total 20971520 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 4096 bytes
I/O size (minimum/optimal): 4096 bytes / 65536 bytes
Disk identifier: 0xb97f94c1
Device Boot
/dev/sdb1
Start
64
End
20971519
Blocks
10485728
Id
83
System
Linux
Compression Alignment—File Systems
In addition to the partition, the file system must also be aligned to 8KB boundaries. This means that the
block size of the file system must be 8KB. When using Oracle ASM, 8KB alignment is ensured because
of the way Oracle ASM performs extent allocation and striping.
When using other file systems, the block size must be specified at 8KB. Doing so might not be possible
with all file systems.
I/O Scheduler
The Linux kernel allows low-level control over the way that I/O to block devices is scheduled. The defaults
on various distribution of Linux vary considerably. Testing shows that Deadline usually offers the best
results, but on occasion NOOP has been slightly better. The difference in performance is minimal, but test
both options if it is necessary to extract the maximum possible performance from a database
configuration. CFQ is the default in many configurations, and it has demonstrated significant performance
problems with database workloads.
See the relevant Linux vendor documentation for instructions on configuring the I/O scheduler.
50
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
Multipathing
Some customers have encountered crashes during network disruption because the multipath daemon
was not running on their system. On recent versions of Linux, the installation process of the OS and the
multipathing daemon might leave these OSs vulnerable to this problem. The packages are installed
correctly, but they are not configured for automatic startup after a reboot.
For example, the default for the multipath daemon on RHEL5.5 might appear as follows:
[root@jfs0 iscsi]# chkconfig --list | grep multipath
multipathd
0:off
1:off
2:off
3:off
4:off
5:off
6:off
5:on
6:off
This can be corrected with the following commands:
[root@jfs0 iscsi]# chkconfig multipathd on
[root@jfs0 iscsi]# chkconfig --list | grep multipath
multipathd
0:off
1:off
2:on
3:on
4:on
18.4 ASM Mirroring
ASM mirroring might require changes to the Linux multipath settings to allow ASM to recognize a problem
and switch over to an alternate fail group. Most ASM configurations on ONTAP use external redundancy,
which means that data protection is provided by the external array and ASM does not mirror data. Some
sites use ASM with normal redundancy to provide two-way mirroring, normally across different sites.
The Linux settings shown in the NetApp Host Utilities documentation include multipath parameters that
result in indefinite queuing of I/O. This means an I/O on a LUN device with no active paths waits as long
as required for the I/O to complete. This is usually desirable because Linux hosts wait as long as needed
for SAN path changes to complete, for FC switches to reboot, or for a storage system to complete a
failover.
This unlimited queuing behavior causes a problem with ASM mirroring because ASM must receive an I/O
failure for it to retry I/O on an alternate LUN.
Set the following parameters in the Linux multipath.conf file for ASM LUNs used with ASM mirroring:
polling_interval 5
no_path_retry 24
These settings create a 120-second timeout for ASM devices. The timeout is calculated as the
polling_interval * no_path_retry as seconds. The exact value might need to be adjusted in
some circumstances, but a 120 second timeout should be sufficient for most uses. Specifically, 120
seconds should allow a controller takeover or giveback to occur without producing an I/O error that would
result in the fail group being taken offline.
A lower no_path_retry value can shorten the time required for ASM to switch to an alternate fail group,
but this also increases the risk of an unwanted failover during maintenance activities such as a controller
takeover. The risk can be mitigated by careful monitoring of the ASM mirroring state. If an unwanted
failover occurs, the mirrors can be rapidly resynced if the resync is performed relatively quickly. For
additional information, see the Oracle documentation on ASM Fast Mirror Resync for the version of
Oracle software in use.
18.5 ASMlib Block Sizes
ASMlib is an optional ASM management library and associated utilities. Its primary value is the capability
to stamp a LUN or an NFS-based file as an ASM resource with a human-readable label.
Recent versions of ASMlib detect a LUN parameter called Logical Blocks Per Physical Block Exponent
(LBPPBE). This value was not reported by the ONTAP SCSI target until recently. It now returns a value
that indicates that a 4KB block size is preferred. This is not a definition of block size, but it is a hint to any
51
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
application that uses LBPPBE that I/Os of a certain size might be handled more efficiently. ASMlib does,
however, interpret LBPPBE as a block size and persistently stamps the ASM header when the ASM
device is created.
This process can cause problems with upgrades and migrations in a number of ways, all based on the
inability to mix ASMlib devices with different block sizes in the same ASM disk group.
For example, older arrays generally reported an LBPPBE value of 0 or did not report this value at all.
ASMlib interprets this as a 512-byte block size. Newer arrays would be interpreted as having a 4KB block
size. It is not possible to mix both 512-byte and 4KB devices in the same ASM disk group. Doing so
would block a user from increasing the size of the ASM disk group using LUNs from two arrays or
leveraging ASM as a migration tool. In other cases, RMAN might not permit the copying of files between
an ASM disk group with a 512-byte block size and an ASM disk group with a 4KB block size.
The preferred solution is to patch ASMlib. The Oracle bug ID is 13999609, and the patch is present in
oracleasm-support-2.1.8-1 and higher. This patch allows a user to set the parameter
ORACLEASM_USE_LOGICAL_BLOCK_SIZE to FALSE in the /etc/sysconfig/oracleasm
configuration file. Doing so blocks ASMlib from using the LBPPBE parameter, which means that LUNs on
the new array are now recognized as 512-byte block devices.
Note:
The option does not change the block size on LUNs that were previously stamped by ASMlib. For
example, if an ASM disk group with 512-byte blocks must be migrated to a new storage system
that reports a 4KB block, the option ORACLEASM_USE_LOGICAL_BLOCK_SIZE must be set
before the new LUNs are stamped with ASMlib.
If ASMlib cannot be patched, ASMlib can be removed from the configuration. This change is disruptive
and requires the unstamping of ASM disks and making sure that the asm_diskstring parameter is set
correctly. This change does not, however, require the migration of data.
18.6 Linux ext3 and ext4 Mount Options
NetApp recommends using the default mount options.
19 Microsoft Windows
This section addresses configuration topics specific to the Microsoft Windows OS.
19.1 NFS
Oracle supports the use of Microsoft Windows with the direct NFS client. This capability offers a path to
the management benefits of NFS, including the ability to view files across environments, dynamically
resize volumes, and leverage a less expensive IP protocol. See the official Oracle documentation for
information on installing and configuring a database on Microsoft Windows using DNFS. No special best
practices exist.
19.2 SAN
For optimum compression efficiency, make sure that NTFS file systems use an 8192 byte or larger
allocation unit. Use of a 4096-byte allocation unit, which is generally the default, damages efficiency.
20 Solaris
This section addresses configuration topics specific to the Solaris OS.
20.1 Solaris NFSv3 Mount Options
Table 8 lists the Solaris NFSv3 mount options for a single instance.
52
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
Table 8) Solaris NFSv3 mount options—single instance.
File Type
Mount Options
ADR_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536
Control files
Datafiles
Redo logs
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,nointr,llock,suid
ORACLE_HOME
rw,bg,hard,vers=3,proto=tcp,timeo=600,rsize=65536,wsize=65536,suid
The use of llock has been proven to dramatically improve performance in customer environments by
removing the latency associated with acquiring and releasing locks on the storage system. Use this
option with care in environments in which numerous servers are configured to mount the same file
systems and Oracle is configured to mount these databases. Although this is a highly unusual
configuration, it is used by a small number of customers. If an instance is accidentally started a second
time, data corruption can occur because Oracle is unable to detect the lock files on the foreign server.
NFS locks do not otherwise offer protection; as in NFS version 3, they are advisory only.
Because the llock and forcedirectio parameters are mutually exclusive, it is important that
filesystemio_options=setall is present in the init.ora file so that directio is used. Without
this parameter, host OS buffer caching is used and performance can be adversely affected.
Table 9 lists the Solaris NFSv3 RAC mount options.
Table 9) Solaris NFSv3 mount options—RAC.
File Type
Mount Options
ADR_HOME
rw, bg, hard, vers=3, proto=tcp, timeo=600, rsize=65536, wsize=65536, noac
Control files
Data files
Redo logs
rw, bg, hard, vers=3, proto=tcp, timeo=600, rsize=65536, wsize=65536, nointr,
noac, forcedirectio
CRS/Voting
rw, bg, hard, vers=3, proto=tcp, timeo=600, rsize=65536, wsize=65536, nointr,
noac, forcedirectio
Dedicated
ORACLE_HOME
rw, bg, hard, vers=3, proto=tcp, timeo=600, rsize=65536, wsize=65536, suid
Shared ORACLE_HOME
rw, bg, hard, vers=3, proto=tcp, timeo=600, rsize=65536, wsize=65536, nointr,
noac, suid
The primary difference between single-instance and RAC mount options is the addition of noac and
forcedirectio to the mount options. This addition has the effect of disabling the host OS caching,
which enables all instances in the RAC cluster to have a consistent view of the state of the data. Although
using the init.ora parameter filesystemio_options=setall has the same effect of disabling
host caching, it is still necessary to use noac and forcedirectio.
The reason actimeo=0 is required for shared ORACLE_HOME deployments is to facilitate consistency of
files such as Oracle password files and spfiles. If each instance in a RAC cluster has a dedicated
ORACLE_HOME, this parameter is not required.
53
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
20.2 Solaris UFS Mount Options
NetApp strongly recommends using the logging mount option so that data integrity is preserved in the
case of a Solaris host crash or the interruption of FC connectivity. The logging mount option also
preserves the usability of Snapshot backups.
20.3 Solaris ZFS
Solaris ZFS must be installed and configured carefully to deliver optimum performance.
mvector
Solaris 11 included a change in how it processes large IO operations which can result in severe
performance problems on SAN storage arrays. The problem is documented in detail in the NetApp bug
report 630173, "Solaris 11 ZFS Performance Regression." The solution is to change an OS parameter
called zfs_mvector_max_size.
Run the following command as root:
echo "zfs_mvector_max_size/W 0t131072" |mdb -kw
If any unexpected problems arise from this change, it can be easily reversed by running the following
command as root:
echo "zfs_mvector_max_size/W 0t1048576" |mdb -kw
Kernel
Reliable ZFS performance requires a Solaris kernel patched against LUN alignment problems. The fix
was introduced with patch 147440-19 in Solaris 10 and with SRU 10.5 for Solaris 11. Only use Solaris 10
and later with ZFS.
LUN Configuration
To configure a LUN, complete the following steps:
1. Create a LUN of type solaris.
2. Install the appropriate Host Utility Kit (HUK) specified by the IMT.
3. Follow the instructions in the HUK exactly as described. The basic steps are outlined in this section,
but refer to the latest documentation for the proper procedure.
a. Run the host_config utility to update the sd.conf/sdd.conf file. Doing so allows the SCSI
drives to correctly discover ONTAP LUNs.
b. Follow the instructions given by the host_config utility to enable multipath input/output (MPIO).
c.
Reboot. This step is required so that any changes are recognized across the system.
4. Partition the LUNs and verify that they are properly aligned. See "Appendix B: WAFL Alignment
Verification” for instructions on how to directly test and confirm alignment.
zpools
A zpool should only be created after the steps in the section “LUN Configuration” are performed. If the
procedure is not done correctly, it can result in serious performance degradation due to the I/O alignment.
Optimum performance on ONTAP requires I/O to be aligned to a 4K boundary on disk. The file systems
created on a zpool will use an effective block size that is controlled through a parameter called ashift,
which can be viewed by running the command zdb -C.
54
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
The value of ashift defaults to 9, which means 2^9, or 512 bytes. For optimum performance, the
ashift value must be 12 (2^12=4K). This value is set at the time the zpool is created and cannot be
changed, which means that data in zpools with ashift other than 12 should be migrated by copying data
to a newly created zpool.
After creating a zpool, verify the value of ashift before proceeding. If the value is not 12, the LUNs were
not discovered correctly. Destroy the zpool, verify that all steps shown in the relevant Host Utilities
documentation were performed correctly, and recreate the zpool.
zpools and Solaris LDOMs
Solaris LDOMs create an additional requirement for making sure that I/O alignment is correct. Although a
LUN might be properly discovered as a 4K device, a virtual vdsk device on an LDOM does not inherit the
configuration from the I/O domain. The vdsk based on that LUN will default back to a 512-byte block.
An additional configuration file is required. First, the individual LDOM’s must be patched for Oracle bug
15824910 to enable the additional configuration options. This patch has been ported into all currently
used versions of Solaris. Once the LDOM is patched, it is ready for configuration of the new properly
aligned LUNs as follows:
1. Identify the LUN or LUNs to be used in the new zpool. In this example, it is the c2d1 device.
root@LDOM1 # echo | format
Searching for disks...done
AVAILABLE DISK SELECTIONS:
0. c2d0 <Unknown-Unknown-0001-100.00GB>
/virtual-devices@100/channel-devices@200/disk@0
1. c2d1 <SUN-ZFS Storage 7330-1.0 cyl 1623 alt 2 hd 254 sec 254>
/virtual-devices@100/channel-devices@200/disk@1
2. Retrieve the vdc instance of the devices to be used for a ZFS pool:
root@LDOM1 # cat /etc/path_to_inst
#
# Caution! This file contains critical kernel state
#
"/fcoe" 0 "fcoe"
"/iscsi" 0 "iscsi"
"/pseudo" 0 "pseudo"
"/scsi_vhci" 0 "scsi_vhci"
"/options" 0 "options"
"/virtual-devices@100" 0 "vnex"
"/virtual-devices@100/channel-devices@200" 0 "cnex"
"/virtual-devices@100/channel-devices@200/disk@0" 0 "vdc"
"/virtual-devices@100/channel-devices@200/pciv-communication@0" 0 "vpci"
"/virtual-devices@100/channel-devices@200/network@0" 0 "vnet"
"/virtual-devices@100/channel-devices@200/network@1" 1 "vnet"
"/virtual-devices@100/channel-devices@200/network@2" 2 "vnet"
"/virtual-devices@100/channel-devices@200/network@3" 3 "vnet"
"/virtual-devices@100/channel-devices@200/disk@1" 1 "vdc" << We want this one
3. Edit the /platform/sun4v/kernel/drv/vdc.conf:
block-size-list="1:4096";
This means that device instance 1 will be assigned a block size of 4096.
As an additional example, assume vdsk instances 1 through 6 need to be configured for a 4K block
size and /etc/path_to_inst reads as follows:
"/virtual-devices@100/channel-devices@200/disk@1"
"/virtual-devices@100/channel-devices@200/disk@2"
"/virtual-devices@100/channel-devices@200/disk@3"
"/virtual-devices@100/channel-devices@200/disk@4"
"/virtual-devices@100/channel-devices@200/disk@5"
55
Oracle Databases on ONTAP
1
2
3
4
5
"vdc"
"vdc"
"vdc"
"vdc"
"vdc"
© 2017 NetApp, Inc. All rights reserved.
"/virtual-devices@100/channel-devices@200/disk@6" 6 "vdc"
4. The final vdc.conf file should contain the following:
block-size-list="1:8192","2:8192","3:8192","4:8192","5:8192","6:8192";
Caution
The LDOM must be rebooted after vdc.conf is configured and the vdsk is created. This step cannot
be avoided. The block size change only takes effect after a reboot. Proceed with zpool configuration
and ensure that ashift is properly set to 12 as described previously.
ZIL
Generally, there is no reason to locate the ZFS Intent Log (ZIL) on a different device. The log can share
space with the main pool. The primary use of a separate ZIL is when using physical drives that lack the
write caching features in modern storage arrays.
logbias
Set the logbias parameter on ZFS file systems hosting Oracle data.
zfs set logbias=throughput <filesystem>
Using this parameter reduces overall write levels. Under the defaults, written data is committed first to the
ZIL and then to the main storage pool. This approach is appropriate for a configuration using a plain disk
configuration, which includes a SSD-based ZIL device and spinning media for the main storage pool,
because it allows a commit to occur in a single I/O transaction on the lowest latency media available.
When using a modern storage array that includes its own caching capability, this approach is not
generally necessary. Under rare circumstances, it might be desirable to commit a write with a single
transaction to the log, such as a workload that consists of highly concentrated, latency-sensitive random
writes. There are consequences in the form of write amplification because the logged data is eventually
written to the main storage pool, resulting in a doubling of the write activity.
Direct I/O
Many applications, including Oracle products, can bypass the host buffer cache by enabling direct I/O.
This strategy does not work as expected with ZFS file systems. Although the host buffer cache is
bypassed, ZFS itself continues to cache data. This action can result in misleading results when using
tools such as fio or sio to perform performance tests because it is difficult to predict whether I/O is
reaching the storage system or whether it is being cached locally within the OS. This action also makes it
very difficult to use such synthetic tests to compare ZFS performance to other file systems. As a practical
matter, there is little to no difference in file system performance under real user workloads.
Multiple zpools
Snapshot-based backups, restores, clones, and archiving of ZFS-based data must be performed at the
level of the zpool and typically requires multiple zpools. A zpool is analogous to an LVM disk pool and
should be configured using the same rules. For example, a database is probably best laid out with the
datafiles residing on zpool1 and the archive logs, control files, and redo logs residing on zpool2. This
approach permits a standard hot backup in which the database is placed in hot backup mode, followed by
a Snapshot copy of zpool1. The database is then removed from hot backup mode, the log archive is
forced, and a Snapshot copy of zpool2 is created. A restore operation requires unmounting the zfs file
systems and offlining the zpool in its entirety, following by a SnapRestore restore operation. The zpool
can then be brought online again and the database recovered.
56
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
filesystemio_options
The Oracle parameter filesystemio_options works differently with ZFS. If setall or directio is
used, write operations are synchronous and bypass the OS buffer cache, but reads are buffered by ZFS.
This action causes difficulties in performance analysis because I/O is sometimes intercepted and serviced
by the ZFS cache, making storage latency and total I/O less than it might appear to be.
21 Conclusion
As stated at the start of this document, there are few true best practices for an Oracle storage
configuration because there is so much variability between implementations. A database project could
contain one mission-critical database or it could contain 5000 legacy databases or sizes ranges from a
handful of gigabytes to hundreds of terabytes. Options such as clusterware and virtualization introduce
further variables.
A better term is design considerations or issues that must be considered while planning a storage
implementation. The right solution depends on both the technical details of the implementation and the
business requirements driving the project. NetApp and partner professional services experts are available
for assistance in complex projects. Even if assistance is not required for the duration of the project,
NetApp strongly encourages new customers to use professional services for assistance in developing a
high-level approach.
Appendix A: Stale NFS Locks
If an Oracle database server crashes, it might have problems with stale NFS locks upon restart. This
problem is avoidable by paying careful attention to the configuration of name resolution on the server.
This problem arises because creating a lock and clearing a lock use two slightly different methods of
name resolution. Two processes are involved, the Network Lock Manager (NLM) and the NFS client. The
NLM uses uname -n to determine the host name, while the rpc.statd process uses
gethostbyname(). These host names must match for the OS to properly clear stale locks. For
example, the host might be looking for locks owned by dbserver5, but the locks were registered by the
host as dbserver5.mydomain.org. If gethostbyname() does not return the same value as uname
–a, then the lock release process did not succeed.
The following sample script verifies whether name resolution is fully consistent:
#! /usr/bin/perl
$uname=`uname -n`;
chomp($uname);
($name, $aliases, $addrtype, $length, @addrs) = gethostbyname $uname;
print "uname -n yields: $uname\n";
print "gethostbyname yields: $name\n";
If gethostbyname does not match uname, stale locks are likely. For example, this result reveals a
potential problem:
uname -n yields: dbserver5
gethostbyname yields: dbserver5.mydomain.org
The solution is usually found by changing the order in which hosts appear in /etc/hosts. For example,
assume that the hosts file includes this entry:
10.156.110.201
dbserver5.mydomain.org dbserver5 loghost
To resolve this issue, change the order in which the fully qualified domain name and the short host name
appear:
57
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
10.156.110.201
dbserver5 dbserver5.mydomain.org loghost
gethostbyname() now returns the short dbserver5 host name, which matches the output of uname.
Locks are thus cleared automatically after a server crash.
Appendix B: WAFL Alignment Verification
Correct WAFL alignment is critical for good performance. Although ONTAP manages blocks in 4KB units,
this fact does not mean that ONTAP performs all operations in 4KB units. In fact, ONTAP supports block
operations of different sizes, but the underlying accounting is managed by WAFL in 4KB units.
The term “alignment” refers to how Oracle I/O corresponds to these 4KB units. Optimum performance
requires an Oracle 8KB block to reside on two 4KB WAFL physical blocks on a disk. If a block is offset by
2KB, this block resides on half of one 4KB block, a separate full 4KB block, and then half of a third 4KB
block. This arrangement causes performance degradation.
Alignment is not a concern with NAS file systems. Oracle datafiles are aligned to the start of the file based
on the size of the Oracle block. Therefore, block sizes of 8KB, 16KB, and 32KB are always aligned. All
block operations are offset from the start of the file in units of 4 kilobytes.
LUNs, in contrast, generally contain some kind of disk header or file system metadata at their start that
creates an offset. Alignment is rarely a problem in modern OSs because these OSs are designed for
physical disks that might use a native 4KB sector, which also requires I/O to be aligned to 4KB
boundaries for optimum performance.
There are, however, some exceptions. A database might have been migrated from an older OS that was
not optimized for 4KB I/O, or user error during partition creation might have led to an offset that is not in
units of 4KB in size.
The following examples are Linux-specific, but the procedure can be adapted for any OS.
Aligned
The following example shows an alignment check on a single LUN with a single partition.
First, create the partition that uses all partitions available on the disk.
[root@jfs0 iscsi]# fdisk /dev/sdb
Device contains neither a valid DOS partition table, nor Sun, SGI or OSF disklabel
Building a new DOS disklabel with disk identifier 0xb97f94c1.
Changes will remain in memory only, until you decide to write them.
After that, of course, the previous content won't be recoverable.
The device presents a logical sector size that is smaller than
the physical sector size. Aligning to a physical sector (or optimal
I/O) size boundary is recommended, or performance may be impacted.
Command (m for help): n
Command action
e
extended
p
primary partition (1-4)
p
Partition number (1-4): 1
First cylinder (1-10240, default 1):
Using default value 1
Last cylinder, +cylinders or +size{K,M,G} (1-10240, default 10240):
Using default value 10240
Command (m for help): w
The partition table has been altered!
Calling ioctl() to re-read partition table.
Syncing disks.
58
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
[root@jfs0 iscsi]#
The alignment can be checked mathematically with the following command:
[root@jfs0 iscsi]# fdisk -u -l /dev/sdb
Disk /dev/sdb: 10.7 GB, 10737418240 bytes
64 heads, 32 sectors/track, 10240 cylinders, total 20971520 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 4096 bytes
I/O size (minimum/optimal): 4096 bytes / 65536 bytes
Disk identifier: 0xb97f94c1
Device Boot
/dev/sdb1
Start
32
End
20971519
Blocks
10485744
Id
83
System
Linux
The output shows that the units are 512 bytes, and the start of the partition is 32 units. This is a total of 32
x 512 = 16,834 bytes, which is a whole multiple of 4KB WAFL blocks. This partition is correctly aligned.
To verify correct alignment, complete the following steps:
1. Identify the universally unique identifier (UUID) of the LUN.
FAS8040SAP::> lun show -v /vol/jfs_luns/lun0
Vserver Name: jfs
LUN UUID: ed95d953-1560-4f74-9006-85b352f58fcd
Mapped: mapped
2. Enter the node shell on the ONTAP controller.
FAS8040SAP::> node run -node FAS8040SAP-02
Type 'exit' or 'Ctrl-D' to return to the CLI
FAS8040SAP-02> set advanced
set not found. Type '?' for a list of commands
FAS8040SAP-02> priv set advanced
Warning: These advanced commands are potentially dangerous; use
them only when directed to do so by NetApp
personnel.
3. Start statistical collections on the target UUID identified in the first step.
FAS8040SAP-02*> stats start lun:ed95d953-1560-4f74-9006-85b352f58fcd
Stats identifier name is 'Ind0xffffff08b9536188'
FAS8040SAP-02*>
4. Perform some I/O. It is important to use the iflag argument to make sure that I/O is synchronous
and not buffered.
Note:
Be very careful with this command. Reversing the if and of arguments destroys data.
[root@jfs0 iscsi]# dd if=/dev/sdb1 of=/dev/null iflag=dsync count=1000 bs=4096
1000+0 records in
1000+0 records out
4096000 bytes (4.1 MB) copied, 0.0186706 s, 219 MB/s
5. Stop the stats and view the alignment histogram. All I/O should be in the .0 bucket, which indicates
I/O that is aligned to a 4KB block boundary.
FAS8040SAP-02*> stats stop
StatisticsID: Ind0xffffff08b9536188
lun:ed95d953-1560-4f74-9006-85b352f58fcd:instance_uuid:ed95d953-1560-4f74-9006-85b352f58fcd
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.0:186%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.1:0%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.2:0%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.3:0%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.4:0%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.5:0%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.6:0%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.7:0%
59
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
Misaligned
The following example shows misaligned I/O.
1. Create a partition that does not align to a 4KB boundary. This is not default behavior on modern OSs.
[root@jfs0 iscsi]# fdisk -u /dev/sdb
Command (m for help): n
Command action
e
extended
p
primary partition (1-4)
p
Partition number (1-4): 1
First sector (32-20971519, default 32): 33
Last sector, +sectors or +size{K,M,G} (33-20971519, default 20971519):
Using default value 20971519
Command (m for help): w
The partition table has been altered!
Calling ioctl() to re-read partition table.
Syncing disks.
2. The partition has been created with a 33-sector offset instead of the default 32. Repeat the procedure
outlined in the section “Aligned.” The histogram appears as follows:
FAS8040SAP-02*> stats stop
StatisticsID: Ind0xffffff0468242e78
lun:ed95d953-1560-4f74-9006-85b352f58fcd:instance_uuid:ed95d953-1560-4f74-9006-85b352f58fcd
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.0:0%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.1:136%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.2:4%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.3:0%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.4:0%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.5:0%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.6:0%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_align_histo.7:0%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:read_partial_blocks:31%
The misalignment is clear. The I/O mostly falls into the .1 bucket, which matches the expected offset.
When the partition was created, it was moved 512 bytes further into the device than the optimized
default, which means that the histogram is offset by 512 bytes.
Additionally, the read_partial_blocks statistic is nonzero, which means I/O was performed that
did not fill up an entire 4KB block.
Redo Logging
The procedures explained here are applicable to datafiles. Oracle redo logs and archive logs have
different I/O patterns. For example, redo logging is a circular overwrite of a single file. If the default 512byte block size is used, the write statistics look something like this:
FAS8040SAP-02*> stats stop
StatisticsID: Ind0xffffff0468242e78
lun:ed95d953-1560-4f74-9006-85b352f58fcd:instance_uuid:ed95d953-1560-4f74-9006-85b352f58fcd
lun:ed95d953-1560-4f74-9006-85b352f58fcd:write_align_histo.0:12%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:write_align_histo.1:8%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:write_align_histo.2:4%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:write_align_histo.3:10%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:write_align_histo.4:13%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:write_align_histo.5:6%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:write_align_histo.6:8%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:write_align_histo.7:10%
lun:ed95d953-1560-4f74-9006-85b352f58fcd:write_partial_blocks:85%
The I/O would be distributed across all histogram buckets, but this is not a performance concern.
Extremely high redo-logging rates might, however, benefit from the use of a 4KB block size. In this case,
60
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
it is desirable to make sure that the redo-logging LUNs are properly aligned. However, this is not as
critical to good performance as datafile alignment.
61
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
Refer to the Interoperability Matrix Tool (IMT) on the NetApp Support site to validate that the exact
product and feature versions described in this document are supported for your specific environment. The
NetApp IMT defines the product components and versions that can be used to construct configurations
that are supported by NetApp. Specific results depend on each customer’s installation in accordance with
published specifications.
Copyright Information
Copyright © 1994–2017 NetApp, Inc. All rights reserved. Printed in the U.S. No part of this document
covered by copyright may be reproduced in any form or by any means—graphic, electronic, or
mechanical, including photocopying, recording, taping, or storage in an electronic retrieval system—
without prior written permission of the copyright owner.
Software derived from copyrighted NetApp material is subject to the following license and disclaimer:
THIS SOFTWARE IS PROVIDED BY NETAPP “AS IS” AND WITHOUT ANY EXPRESS OR IMPLIED
WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, WHICH ARE HEREBY
DISCLAIMED. IN NO EVENT SHALL NETAPP BE LIABLE FOR ANY DIRECT, INDIRECT,
INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
THE POSSIBILITY OF SUCH DAMAGE.
NetApp reserves the right to change any products described herein at any time, and without notice.
NetApp assumes no responsibility or liability arising from the use of products described herein, except as
expressly agreed to in writing by NetApp. The use or purchase of this product does not convey a license
under any patent rights, trademark rights, or any other intellectual property rights of NetApp.
The product described in this manual may be protected by one or more U.S. patents, foreign patents, or
pending applications.
RESTRICTED RIGHTS LEGEND: Use, duplication, or disclosure by the government is subject to
restrictions as set forth in subparagraph (c)(1)(ii) of the Rights in Technical Data and Computer Software
clause at DFARS 252.277-7103 (October 1988) and FAR 52-227-19 (June 1987).
Trademark Information
NETAPP, the NETAPP logo, and the marks listed at http://www.netapp.com/TM are trademarks of
NetApp, Inc. Other company and product names may be trademarks of their respective owners.
TR-3633-0717
62
Oracle Databases on ONTAP
© 2017 NetApp, Inc. All rights reserved.
Download PDF

advertising