SCSI - VISER-u
SCSI
 Parallel SCSI
 Serial SCSI = SAS (Serial Attached SCSI)
SCSI
Slide 1 of 25
Parallel SCSI
 Parallel 16 bit disk interface
 High Performance
 BUS based interfaces-up to 16 devices per bus
 Disconnect / reconnect capability
SCSI
Slide 2 of 25
SCSI BUS
 Arbitration phase — Initiator gains control of the SCSI bus.
 Selection phase — Initiator selects a device for communication.
 Message phase — Messages are exchanged that identify specific
devices and specify parameters for data-transfer transactions.
 Command phase — Initiator sends out a request for action.
 Data phase — Data is transferred.
 Status phase — The status of the last operation is posted.
SCSI
Slide 3 of 25
SCSI
Slide 4 of 25
Overview of SCSI-3 Standards
Protocol
Description
Abbreviation
and
Status
Generation
Document
Standard
Project
or
Withdrawn;
now
incorporated
into
later versions of the -SCSI-3
Parallel
Interface
Interlocked
(Parallel
Bus)
Defines the protocol for SCSI-3 Interlocked
SIP
"regular" parallel SCSI
Protocol
Fibre
Channel
Defines the protocol for SCSI-3
Fibre FCP
running SCSI on the Fibre Channel Protocol
FCP-2
Channel interface
Published
X3.269-1996
Pending Publication
T10 1144-D
SBP
Withdrawn
--
Serial Bus
Defines the protocol for
transporting
commands Serial
over
the
IEEE-1394 Protocol
(serial) interface
SBP-2
Published
NCITS.3251998
Serial
Storage
Architecture SCSI- SSA-S3P
3 Protocol
Published
NCITS.3091998
Serial
Storage
Architecture
SSA-TL2
Transport Layer
Published (replaced NCITS.308SSA-TL1)
1998
Defines the transport layer
for
Serial
Storage
Serial
Architecture, an advanced
Storage
interface used in servers
Architecture and enterprise hardware;
there are two documents
that specify the protocol
SCSI
Bus
Slide 5 of 25
SCSI Bus Speed
StandardDefined
Bus Speed
Common
Clock
Signaling Speed Speed
Name
(MHz)
Clocking
Throughput (MB/s)
Transfer Rate
(Mtransfers/s)
Narrow (8- Wide (16bit)
bit)
SCSI-1
"Regular"
5
Single
5
5
--
Fast
"Fast"
10
Single
10
10
20
Fast-20
"Ultra"
20
Single
20
20
40
Fast-40
"Ultra2"
40
Single
40
40
80
Fast80(DT)
"Ultra3"
"Ultra160"
40
Double
80
--
160
Fast160(DT)
"Ultra320"
80
Double
160
--
320
or
SCSI
Slide 6 of 25
Command Queuing and Reordering




SCSI is often described as being "advanced", or is called an "intelligent interface". One of
the reasons for these descriptions is that SCSI hardware incorporates features that
improve overall system performance, where simpler interfaces such as IDE/ATA do not.
One of these techniques is a special feature that allows for concurrent, multiple requests
to devices on the SCSI bus. This feature is called command queuing and reordering;
sometimes the name is given as tagged command queuing. It was first introduced in the
SCSI-2 standard.
Traditionally, a simple interface like SCSI-1 or IDE/ATA will allow only a single command
to be outstanding at a time to any device. This means that once a particular command is
sent to a device, any other commands must wait for the first one to be completed, which
slows down performance. Command queuing allows a device to accept as many as 64
or even 256 concurrent commands. The commands can also come from different
originating devices. Command reordering allows a device that has multiple commands
outstanding to fill them "out of order", meaning, not necessarily in the order that they were
received.
For a very simple SCSI bus, such as a single hard disk on a host adapter in a desktop PC,
command queuing and reordering may not make a particularly huge difference in
performance. The reason is simply that there aren't that many concurrent processes
running, and not a great deal of activity on the bus. This feature really comes into its own
in a multiple-device, multitasking environment, such as that experienced by a shared
server. In that environment, command queuing and reordering will improve performance
significantly, by allowing devices to accept multiple simultaneous requests from different
users, and fill them in the most efficient manner.
This is very important for devices like hard disks, which are very slow compared to the rest
of the system. If commands are processed only as they are received, a great deal of time
may be wasted while the hard disk's mechanical components move past a physically close
piece of data that will be needed one or two requests "down the road". For a more
thorough explanation of how drives can improve performance by reordering commands,
see this discussion.
SCSI
Slide 7 of 25
Negotiation and Domain Validation
 SCSI hardware supports many different speeds, and newer, faster hardware is
generally backwards-compatible with older, slower devices. You can use a
host adapter capable of 160 MB/s throughput with drives that can only support
20 MB/s transfers, or vice-versa. This leaves an obvious question: how does
each device determine what speeds the others on the bus are capable of?
Without knowing this, senders can't figure out how fast receivers can handle
data being sent.
 Since this is so important, the SCSI protocols build in support for a method by
which the host adapter can interrogate all devices on the bus to find out what
speeds they support. This process is called negotiation, and is one of
the first tasks performed by the SCSI host adapter when the system power is
applied. Under conventional SCSI rules, this negotiation is done with each
device; the host adapter records the maximum transfer speed that each device
claims to support, and then uses that information when the device is accessed.
 This works great in theory, but there's a problem with it: theory doesn't always
translate into practice, especially when the technology "pushes the envelope"
with high-speed signaling. For example, even if the host adapter can support
Ultra160 transfers and the device says it can as well, this doesn't mean that
160 MB/s signaling is actually possible on the bus. Perhaps the cabling being
used is inferior or too long, or there's a problem with a terminator, or the
system is in a particularly electrically noisy environment. Regular negotiation
just "trusts" that everything will work at the speed the hardware decides is
possible, but it may not actually work. If there are difficulties, they may
manifest themselves in the form of data errors or reliability problems.
SCSI
Slide 8 of 25
Negotiation and Domain Validation
 To improve negotiation, the SPI-3 standard introduced a new feature called
domain validation, sometimes abbreviated DV. This feature basically adds a
verification step to the normal negotiation procedure (note that "domain" is
another word for a SCSI channel or bus). After a device tells the host adapter
that it is capable of transfers at a particular speed, the host adapter tests the
device by sending write requests to the device's internal buffer at that speed.
The data just written is then read back and compared. If the data is different, or
if parity or CRC errors occur during either the read or the write, the host
adapter knows that communication at that speed is not reliable. It will then
retry at the next lower speed, and continue until reliable operation is
established. (If this sounds similar to the way that two regular analog modems
determine a communications speed, that's because it is!)
 Domain validation is one of the five "optional" features of Ultra3 SCSI, and is
a required feature for hardware meeting the Ultra160 or Ultra160+
specifications. This feature may be expanded in the future to include more
frequent validation during the operation of the system, since over time errors
may occur on a channel that worked fine when the system was first powered
up.
SCSI
Slide 9 of 25
Quick Arbitration and Selection (QAS)




During the time when the system is running, the SCSI bus is generally either active
or idle. If active, the bus is busy transmitting data from one device to another; if idle,
it is available for a device to begin sending a command or data. When a device
decides it wants to use the bus, it "bids" for control of the bus. It is also possible that
other devices on the bus will want to use it at the same time, so they too may "bid"
for control. A specific method is used to resolve these requests and decide which
device gets to use the bus first; this is based to some extent on the devices'
respective priority levels. This process is called arbitration.
While arbitration works fine in regular SCSI configurations, it introduces overhead.
During the time that arbitration is going on, no data is being transferred on the bus,
so it makes sense that doing this faster will allow improved performance of the entire
SCSI subsystem. To this end, the SPI-3 standard defined a feature that reduces the
overhead required for arbitration. This feature is called quick arbitration and selection
or QAS. You may also see it called by the name it carried during development, quick
arbitration and select; IBM calls it quick arbitration select and Adaptec, simply quick
arbitrate. These are all different names for the same feature.
In a nutshell, QAS works by reducing the number of times arbitration must occur on
the bus. When the feature is used, a device waiting for the bus can grab it more
quickly after the last device on the bus sends the signal that it is done, without having
to begin a new arbitration process. Provision is made in the specification to ensure
that one device does not "dominate" the bus by "unfairly" blocking out other devices
that may be of a lower priority or may not implement QAS.
Quick arbitration and selection is one of the five "optional" features of Ultra3 SCSI. It
was not included as one of the required features for hardware meeting the Ultra160
specification, but is present in Ultra160+ devices.
SCSI
Slide 10 of 25
Packetization
 While the SCSI interface is widely implemented on high-end hardware due
to its flexibility and high performance, its complexity does mean that some
of its potential performance is lost to overhead. In an effort to improve SCSI
bus performance by reducing overhead, the SPI-3 SCSI standard describes
a new feature that is generally called packetization or packetized SCSI.
 Packetization is a technique whereby some of the phases that are involved
in setting up a command request and data transfer are combined. For
example, under traditional SCSI interfacing, several different types of
information are sent over the bus separately: commands, data, status
messages and so on. With packetization, these are grouped together into
packets (also called information units) and sent as a single entity. This
reduces some of the wasted bus cycles normally sent on managing all the
individual transfers in regular SCSI.
 Packetization is one of the five "optional" features of Ultra3 SCSI. It was not
included as one of the required features for hardware meeting the Ultra160
specification, but is present in Ultra160+ devices. It may also be part of the
requirements for Ultra320 SCSI when that specification is complete.
SCSI
Slide 11 of 25
SCSI kontroleri
SCSI
Slide 12 of 25
SCSI Host Adapters

SCSI
Slide 13 of 25
Low-Voltage Differential (LVD) Signaling
 Icons for hardware using LVD SCSI (left) and multimode LVD/SE
SCSI (right).
SCSI
Slide 14 of 25
SCSI Cable Types
 An assortment of different internal ribbon cables used for connecting SCSI
hardware. Note that some are strictly flat cables, but the one on the far left
and the one third from the right are partially flat and partially twisted pair
cable.
SCSI
Slide 15 of 25
Low Voltage Differential (LVD) Cables, Connectors
and Signals
 An internal, 68-wire, 5-connector LVD cable. Note the distinctive
"loose" twisted pair wiring between the connectors. The circuit
board at bottom right is an integrated LVD/SE terminator.
(Incidentally, one of the connectors is hidden behind the terminator
and hard to see.)
SCSI
Slide 16 of 25
SCSI kablovi
SCSI
Slide 17 of 25
Single Connector Attachment (SCA)

Single Connector Attachment (SCA): "Alternative 4" in the SCSI
standards for unshielded connectors doesn't actually refer to cable
connectors, but the connector used for the single connector
attachment system for backplane-connection of SCSI drives.
 A female 80-pin SCA connector. This is the connector that would be
found on a backplane designed for SCA SCSI
SCSI
Slide 18 of 25
Tipičan SCSI disk
Quantum Atlas 10K II
SCSI
maximum Capacity
73.0 GB
average seek time
4.7ms
rotational speed
10.000 rpm
media speed
54MB/s
disk interface
ultra 160
maximum disk buffer throughput
160 MB/sec
disk buffer
8 MB
Slide 19 of 25
Primeri za razumevanje
 PRIMER ZA VEŽBU BR. 1
 SCSI diskovi
 PRIMER ZA VEŽBU BR. 2
 SCSI taktovi i brzine prenosa
SCSI
Slide 20 of 25
PRIMER BR. 1
< SCSI diskovi >
 Date su karakteristike jednog realnog SCSI diska.
Quantum Atlas 10K II
maximum Capacity
73.0 GB
average seek time
4.7ms
rotational speed
10.000 rpm
media speed
54MB/s
disk interface
ultra 160
maximum disk buffer throughput
160 MB/sec
disk buffer
8 MB

U konfiguraciji računara imate 2 takva diska.

Pristigla su 2 zahteva, za svaki disk po jedan



disk #1: sa adrese (cilindar 15, head 0, sektor 13), pročitati 10 sektora
disk #2: sa adrese (cilindar 940, head 2, sektor 1), pročitati 40 sektora
odrediti ukupno vreme trajanja oba disk pristupa bez preklapanja media i interface faze
SCSI
Slide 21 of 25
PRIMER BR. 1
< SCSI diskovi >
 Rešenje:
 Podjimo od šeme SCSI magistrale
 SCSI magistrala je multi-task orijentisana, dozvoljava više od jedne disk
operaciju u jednom trenutku.
 U našem slučaju, 2 SCSI disk su multi-tasksing orijentisana, svaki radi
nezavisno, tako da se 2 disk pristupa mogu istovremeno odvijati.
SCSI
Slide 22 of 25
PRIMER BR. 1
< SATA diskovi >
 Oba diska su SCSI
 i ovde se disk operacije preklapaju, obe se rade istovremeno
 Ttotal=Max(Taccess1, Taccess2)






disk #1: sa adrese (cilindar 15, head 0, sektor 13), pročitati 10 sektora
Taccess1 = Tseek1 + Trotate1 + Tmedia1 + Tinterface1
Tseek1 = 1 x average seek time = 4.7 msec
Trotate1 = Trotate_avg(10000 rpm)
= 3 msec
Tmedia1 = Q/Vmedia=10x0.5KB/54MB/s= 0.093 msec
Tinteface1=Q/Vinterface=10x0.5KB/160MB/s= 0.03125msec
 Taccess1= 7.824 msec
SCSI
Slide 23 of 25
PRIMER BR. 1
< SCSI diskovi >
 disk #2: sa adrese (cilindar 940, head 2, sektor 1), pročitati 40 sektora
 Taccess2 = Tseek2 + Trotate2 + Tmedia2 + Tinterface2
 Tseek2 = 1 x average seek time = 4.7 msec
 Trotate1 = Trotate_avg(10000 rpm)
= 3 msec
 Tmedia2 = Q/Vmedia=40x0.5KB/54MB/s= 0.37msec
 Tinteface2=Q/Vinterface=40x0.5KB/160MB/s= 0.125msec
 Taccess2= 8.195msec
 Ttotal= Max( 7.824, 8.195) = 8.195msec
SCSI
Slide 24 of 25
PRIMER BR. 2
<SCSI taktovi i brzine prenosa >
 Zadatak 1.
 Odrediti osnovni UDMA takt za SCSI ultra 320?









Rešenje:
Prvo ćemo poći od formule:
V=(Frequency x 2 x 16)/8
Ultra 320 transfer rate=
X MHz strobe
x
2 for double data rate clocking
x
16 for bits per edge
/
8 bits per byte
=
320 Mbytes/sec
 Frequency= 8xV / (2x16) = 8x320/32=80MHz
SCSI
Slide 25 of 25
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising