Ultra320 SCSI Meets The Challenges.

Improvements in recording technologies continue to increase the rate at which data storage devices can read data from their media and host adapters increase their data transfer capabilities. Standardized definitions for new SCSI transfer rates must be devised to prevent the interface from becoming a performance bottleneck for systems. The ANSI NCITS T10 Technical Committee (often called, "the ANSI SCSI committee" or simply "T10") with input from the SCSI Trade Association (STA) develops standards that define the requirements for the parallel SCSI interface.

Ultra320 SCSI, with a maximum transfer rate of 320MB/sec, is the latest development for the SCSI parallel interface. Additional new features to enhance performance, improve data reliability, and increase ease of use have also been developed by T10. This article will describe the mandatory and optional features defined for Ultra320 SCSI devices. Additional details about these features are available in the ANSI draft standard document SCSI Parallel Interface-4 (SPI-4). The latest revision of this draft is available for review at http://www.tl0.org/.

Mandatory Features

DT (Double-transition) data transfers: For "single-transition" (ST) transfers, the maximum frequency of the clock signals is twice the maximum frequency of the data signals because data is only transferred on one edge of the clock. DT transfers provide a method for the ACK and REQ clock signals to have the same maximum frequency as the data signals by using both asserting and negating transitions of ACK and REQ for clocking data. Each transition of the DT clock signal transfers two bytes of data as DT transfers can only be used with wide (16-bit) buses.

Free-running clock (FRC): A free-running clock is used to improve integrity of the clock signal by removing inter-symbol interference (ISI). ISI is the effect that a transition on a signal line has on transitions immediately before or after it on the same line. A pulse (or symbol) will cause a nearby preceding pulse to shift forward in time, and it will cause a nearby subsequent pulse to shift backward in time (i.e., a pulse will "interfere" with the placement in time of its adjacent pulses). By having a clock running at a constant frequency and a separate lower-speed signal for qualification of data, the ISI effect is neutralized. The free-running clock is restricted for use with DT information unit transfers at 320MB/sec.

Training pattern: The training pattern is a pre-determined pattern that is transmitted from the sender to the receiver at a specified time. The receiver can use portions of this pattern to perform skew compensation because it knows what the pattern will be (i.e., exactly when data transitions should occur). Other portions of this pattern may be used to perform other signal adjustments such as tuning adaptive active filtering (AAF). The training pattern may be sent before each data transmission or after some period of time or an event such as a bus reset.

Skew compensation of data signals relative to the clock signal: Skew is the difference in time from when one transition launched by a sender arrives at a given point (e.g., a recipient's connector) to when a second transition launched by the sender arrives at the same point. The arrival time difference is caused by several factors including differences in length and electrical characteristics of the two signal paths. If a data transition is skewed so much relative to the clock that it falls outside of the qualifying clock window, the device will not accurately detect data. One of the largest losses in the error budget for Ultra320 SCSI is skew. With the training pattern specified, an Ultra320 SCSI device can establish skew compensation simultaneously for each of the received transitions on the data lines so that they occur at the correct time relative to the clock. Once established, this compensation is used for all subsequent transfers until the next training pattern sequence is initiated.

CRC (Cyclic redundancy check): CRC is an algorithm that provides improved data reliability for the parallel SCSI bus. A sending device uses the algorithm to generate check bytes for transferred information. These check bytes are transmitted immediately following the information. The recipient uses the same algorithm to calculate check bytes from the received information and compares the result to the received check bytes. If the two sets of check bytes match, the information is correct. CRC is defined for use only with DT transfers.

Domain Validation (also known as "Physical layer integrity checking"): Simple domain validation describes how an initiator can use the INQUIRY command to query targets to determine their capabilities (e.g., maximum transfer rate), the system configuration (e.g., the width of the bus), basic functionality of the system components, and how the initiator can use the READ and WRITE BUFFER commands to send and receive known data patterns from the targets for rudimentary data integrity checking.

Backward compatibility: Backward compatibility means that a device supporting a new feature set can be used with legacy devices which only support transfer rates and protocols previously defined for the SCSI interface. Examples include: the ability for transceivers to operate in "single-ended" (SE) mode (as opposed to "low-voltage differential", or LVD, mode required by the higher transfer rates), the ability to tolerate five volt singleended signaling from other devices, and the ability to function properly with the current cable plant specifications (i.e., a 25-meter cable in a point-to-point configuration with only two devices attached or a 12-meter cable with up to 16 devices attached).

Information unit transfers (or "IU transfers", also known as "packetized transfers"): IU transfers provide a protocol to increase overall system performance. Some of the elements of the protocol include:

* A method to encapsulate non-data information (like commands sent from the initiator to the target and status sent from the target to the initiator) into packets and transfer those packets at the maximum negotiated data rate of up to 320MB/sec--as opposed to those same transfers occurring in asynchronous mode at five megabytes per second or less;

* A method to transfer packets for a number of I/O processes without an intervening physical disconnection (e.g., an initiator could send several packets each containing a queued command to the target during a single physical connection);

* A method to minimize overhead by eliminating several bus phase changes per I/O process, for example: a typical WRITE operation using normal data group transfers would require ARBITRATION, SELECTION, COMMAND, DATA OUT, STATUS, and MESSAGE IN phases. The same WRITE operation using IU transfers would only require ARBITRATION, SELECTION, DATA OUT, and DATA IN phases. The command and data would be transferred during the DATA OUT phase, and the STATUS and COMMAND COMPLETE message information would be transferred during the DATA IN phase, all at the maximum transfer rate.

Transmitter pre-compensation with cutback: Transmitter pre-compensation with cutback is an open-loop method of compensating for some of the signal loss that is most severe on the first part of a signal's transition. The transmitting device boosts the amplitude of the first part of the transition, or cuts back the signal for the remainder of the transition. This provides additional signal amplitude where it is most needed and then decreases the amplitude to decrease the negative effects of cross-talk and reflections.

Optional Features

QAS (Quick Arbitration and Selection): QAS provides increased overall system performance by allowing arbitration to occur without incurring the overhead of intervening BUS FREE phases, saving up to microseconds per operation. This is significant when compared to the 1.6 microseconds it takes to transfer each 512-byte sector of data at 320MB/sec. Rather than waiting for a BUS FREE phase, a target may initiate arbitration by issuing a QAS REQUEST message. The devices on the bus may "snoop" the message and participate in the arbitration. QAS can only be enabled if information unit transfers are enabled.

SCSI bus fairness (fairness): Fairness prevents a device from "hogging" the bus by guaranteeing that all devices have an opportunity to arbitrate. Fairness must be enabled when QAS is enabled, as "hogging" could potentially be more of an issue with the QAS protocol. The standard method of arbitration for parallel SCSI is that the highest SCSI ID on the bus always wins arbitration. For fairness arbitration, the SCSI devices monitor arbitration attempts in a fairness register. If a device has won an arbitration during which other devices with lower SCSI IDs had also participated, the device checks its fairness register during the next arbitration. If a device with a lower SCSI ID had participated in the previous arbitration, lost the arbitration, and is participating in the current arbitration, the device that had won the previous arbitration will drop out, allowing the device with the lower SCSI ID to win the arbitration. The device with the higher SCSI ID will not participate in an arbitration until all devi ces with lower SCSI IDs have had an opportunity to win an arbitration.

AAF (Adaptive Active Filter, also known as "receiver equalization with filtering"): AAF is a closed-loop method of improving received signal quality by amplifying the fundamental frequency of the signal while filtering noise and other undesirable components. Devices implementing AAF establish the gain of its amplifiers by setting the amplitude of the high frequency portion of the training pattern to be the same as the low frequency portion at the beginning of the training pattern. Using the training pattern to perform this adjustment of signal amplitude provides for an inherent closed-loop system that can adjust signal quality for different cable plants and changes in system conditions (e.g., when a new device is added to a system causing the electrical characteristics of the cable plant to change). AAF settings may be adjusted as often as necessary because either the initiator or target may initiate the training pattern sequence. A receiver may disable transmitter pre-compensation in a transmitter as AAF pe rforms better in the configuration.

Flow Control: Flow control provides a method during a data streaming operation for a target to give an indication to an initiator that the current SPI data stream information unit is the last of the current stream. This early warning provides a way for the initiator to more effectively manage other.

AIP (Asynchronous Information Protection): AIP provides an enhanced error detection method for the COMMAND, MESSAGE, and STATUS asynchronous transfer phases by transferring error detection information (a BCH Hamming code) on the upper eight data bits of the data bus simultaneous with the information transfer. The recipient uses the same Hamming algorithm to check the received information. If the result of the calculation is correct, the information transferred is correct.

Parallel SCSI continues to develop methods to meet challenges such as increased performance, increased reliability, increased data integrity, and increased ease of use. VARs and system integrators who continue to supply SCSI provide theircustomers with these benefits and connections to their legacy systems. Mark Evans is Maxtor Corp.'s (Milpitas, CA) principal member on the ANSI NCITS T1O Technical Committee.

Year  Media Rate
1999  28 MB/sec
2000  42 MB/sec
2001  56 MB/sec
2002  73 MB/sec
(*)Typical bus bandwidth capability should be 4
times the maximum sustained drive throughput.
(*)Example: In 2001, one 15,000 RPM drive can
sustain about 56MB/sec throughput. Therefore,
as few as three drivers can saturate an
Ultra 160 bus.
(*)Assumed HDD Media Rates

[Graph omitted]