Topic 5 - Disk Subsystems
Ultra 160/m (Ultra3 SCSI)
Emerging SCSI standard replacing Ultra2 SCSI
Ultra 160/m is a subset of Ultra3
Ultra 160/m: double-edge clocking, CRC, domain validation
Ultra3 SCSI
Ultra3 SCSI
SCSI disk
PCI bus
Wide Ultra2 Ultra2
SCSI disk SCSI disk
16-bit = 160MB/sec = Ultra3 SCSI
(no 8-bit narrow Ultra3)
Requires low-voltage differential (LVD) disks
Slower speed if use Ultra2 disks
Ultra2 cabling, terminators, and connectors fully compatible
Ultra2 SCSI
In late 1998, the SCSI Trade Association (STA) ratified Ultra3 SCSI. Ultra 160/m is a
subset of Ultra3 SCSI that was introduced in 1999 by major industry vendors. Ultra
160/m is not a STA recognized term. The "m" portion of the name stands for
manageability due to the new Cyclical Redundancy Checking and domain validation
The Ultra3 specification describes five new features: double-edge clocking, Cyclical
Redundancy Checking (CRC), domain validation, packetization, and Quick Arbitration
and Select (QAS). Ultra3 calls for the use of any, or none, of the features listed in the
specification. Double-edge clocking, CRC, and domain validation are necessary
components of the Ultra 160/m specification. Packetization and QAS are not included in
Ultra 160/m because these features would add cost but give little benefit to most
Double-Edge Clocking (Double Transition (DT) clocking or Fast-80DT) - Before the
development of the Ultra3 SCSI specification, data was transferred over the SCSI bus
using a single edged clock. With a single edged clock, the maximum data transfer rate is
half of the clock speed — thus, a 160MHz data clock would only support an 80MB/sec
maximum transfer rate. With clocking performed at each edge of the clock signal (lead
edge and trailing edge), two data pulses can be handled during each clock cycle, enabling
data to be transferred at data clock speeds. A 160MHz signal supports a 160MB/sec data
transfer rate.
IBM PC Institute 04/06/99
Topic 5 1-2
Topic 5 - Disk Subsystems
Ultra2 SCSI (continued)
Cyclic-Redundancy Checking (CRC) - CRC is, by now, a time-honored method of
verifying that transferred data is received correctly. CRC has been used in
telecommunications for more than a decade. CRC has also been used to provide error-free
data transmission over ethernet, FDDI and Fibre Channel. By applying CRC to the SCSI
signal, the accuracy of data transfers has been made almost perfect. With data streaming
at rates as high as 160MB/sec, such checking can be an important addition to the data
transfer protocol. The rate of undetected random errors drops to approximately 2-32 when
CRC is used. Without CRC, noise, signal flutter, and other phenomena can render data
transfer at such high rates to be unreliable and, in the worst case, unusable.
Domain Validation - Domain validation tests the SCSI network before it completes
negotiation for network bandwidth. If portions of the network are unable to operate at
maximum speed, the network will reduce the data speeds for that portion of the network to
levels that will provide error-free data transfer. In the event that the network develops a
problem, the Ultra 160/m can automatically fall back to a lower data rate. Implementation
of the drop back feature is an area where there may be differences among Ultra 160/m
providers. In some cases, once the speed is dropped back to a lower speed, the SCSI bus
will continue operating at the lower data rate. Regular polling of the SCSI domain may
result in further drops, if the domain does not appear to be able to support higher rates.
Other implementations, however, may enable the bus to return to higher transfer rates if
polling indicates that the bus will support the faster rate.
Ultra2 SCSI (continued)
As with previous upgrades of SCSI standards, Ultra 160/m also has backward compatibility.
Existing Ultra2 cabling, cable lengths, terminators, connectors, connector spacing,
terminators, and back plane designs are fully compatible with Ultra 160/m.
Because Ultra 160/m SCSI supports LVD signaling, Ultra 160/m controllers can support
Ultra2 SCSI devices in Ultra2 SCSI modes. Moreover, Ultra2 controllers can support Ultra
160/m SCSI devices in Ultra2 SCSI modes. This allows mixing Ultra 160/m SCSI and
Ultra2 SCSI devices on the same bus. Either an Ultra 160/m SCSI or an Ultra2 SCSI host
controller can support such a bus. When an Ultra 160/m SCSI controller operates such a
bus, each attached device operates independently at the maximum allowed speed — Ultra
SCSI devices at up to 80MB/sec and Ultra 160/m SCSI devices at up to 160MB/sec.
The SCSI protocol allows each device to separately determine the data transfer rate between
itself and the adapter. For example, a SCSI-2 bus could have some devices running at
5MB/sec, some running Fast/Wide at 20MB/s, and some running at Wide Ultra SCSI at
40MB/sec all sharing the same bus. However, since Ultra2 and Ultra 160/m busses rely on
Low Voltage Differential (LVD) type bus driver circuits to support longer cable lengths and
faster data rates, if a device that is not capable of operating in LVD mode (i.e., SCSI-2
device with single-ended drivers) is connected to an Ultra2 or Ultra 160/m capable bus, then
the entire bus reverts back to single-ended mode which will limit the cable length, number of
devices, and maximum data rate for all devices to Ultra SCSI limits (40MB/sec for Ultra
Wide devices). If only LVD devices are attached to a common bus, then the Ultra2 devices
can run at up to 80MB/sec and the Ultra 160/m devices can run at up to 160MB/sec.
IBM PC Institute 04/06/99
Topic 5 3-4
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