Date: October 31, 1989 To: X3T9.2 Membership From: Lawrence J. Lamers, X3T9.2 Secretary John B. Lohmeyer, X3T9.2 Chairman Subject: October 30-31, 1989 X3T9.2 Working Group Meeting John Lohmeyer opened the meeting by thanking Ken Post of Future Domain for hosting the meeting in Santa Ana, CA. John noted that the Embassy Suites hotel had excellent facilities both for the meeting and the guests. John explained that the meeting is informal and does not count toward X3T9.2 attendance requirements. The people in attendance introduced themselves. John requested that everyone sign in on the attendance roster. Dal Allan stated that American Airlines and United Airlines will not renew the discount meeting fares next year due to the low usage during 1989. John Lohmeyer said he will attempt to obtain the discount fares during 1990 and he encouraged people to use the discount numbers. It is necessary to call the meetings desk in order to use these discounts, but the extra effort means that the ticket can be changed like a full-coach ticket at 40% off. The final agenda was as follows: 1. Review of X3B7.1 Diagnostic Command Set 2. SEARCH Command modifications [Snively] 3. Alternate Physical Protocol Layer Preparation [Stephens] 4. Single-Cable 16-bit Wide SCSI [Penokie, Lohmeyer, Lamers] 5. Cable configurations in wide SCSI [Penokie, Cornaby] 6. Autosense [Snively] 7. Arbitration fairness (89-61) [Penokie] 8. Arbitration unfairness (89-65) [Buesing] 9. New SCSI transceivers [Murdock] 10. Signal quality issues on differential (89-124) [Curry] 11. Multiple Port Function and Path Control (89-133R0) [Stephens] 12. SCSI Working Group Schedule The following people attended the meeting: Name Stat Organization ------------------------------ ---- ------------------------------ Mr. George Canevit V ACT Technology Mr. Timothy M. Christianson V Adaptec, Inc. Mr. Bob Whiteman A AMP, Inc. Mr. Joe Lawlor P AT&T Mr. David R. Krent V Cambrian Systems/JCS Mr. Paul Hanmann P Emulex Corp. Ms. Jean Kodama S Emulex Corp. Mr. Scott Hopkinson V Emulex Corp. Mr. I. Dal Allan P ENDL Mr. Jerry Edwards V Ficomp Electronics Mr. Ed Johnson V Ficomp Electronics Mr. Robert Liu P Fujitsu America, Inc. Mr. Kenneth Post P Future Domain Mr. Kurt Chan P Hewlett Packard Co. Mr. Eric Tausheck V Hewlett Packard Co. Mr. Oscar Kornblum A Hirose Electric U.S.A. Mr. George Penokie P IBM Corp. Mr. Gary R. Stephens A IBM Corp. Mr. David A. Buesing O IBM Corp. Mr. Lawrence J. Lamers P Maxtor Corp. Mr. Stephen Cornaby P Micropolis Corp. Mr. John Spongr A Mitsubishi Electronics Amer Mr. Gary Murdock S National Semiconductor Mr. John Lohmeyer P NCR Corp. Mr. David Steele V NCR Corp. Mr. Gene Milligan O Seagate Technology Mr. Gerald Houlder O Seagate Technology Mr. Fred Burgess S Seagate Technology Mr. Robert N. Snively P Sun Microsystems, Inc. Mr. D. W. Spence P Texas Instruments Mr. Ricardo Dominguez S Texas Instruments Mr. Reuben Yomtov V UNISYS Mr. Jeff Stai P Western Digital Mr. Doug Pickford A Western Digital 34 People Present Status Key: P Principal A Alternate O Observer S Special Interest (frequent visitor) V Visitor The following new documents were distributed at the meeting: Document Doc Date Author Description of Document ------------- -------- --------------- --------------------------------------- X3T9.2/89-94 10/25/89 G. Penokie 16 Bit Data Path on a Single 68-pin Rev 4 Connector X3T9.2/89-130 10/27/89 G. Stephens Proposed Parallel Packet Structure for SCSI-3 [This document will be re-formatted and included in the January 1990 mailing] X3T9.2/89-131 10/27/89 G. Murdock Short Distance Differential SCSI Transceiver X3T9.2/89-132 10/27/89 G. Murdock Long Distance, Small Swing Differential SCSI Transceiver X3T9.2/89-133 10/27/89 G. Stephens Multi-ported SCSI X3T9.2/89-134 10/30/89 R. Dominguez Proposed SCSI Diagnostic Enhancements X3T9.2/89-135 10/27/89 D. Pickford Draft Proposal for a SCSI Diagnostic Command Set (DCS) X3T9.2/89-136 D. Pickford Slides from Santa Ana Presentation on 89-135 X3T9.2/89-137 10/27/89 R. Yomtov Testing Recommendation to the SCSI subcommittee Results of meeting 1. Review of X3B7.1 Diagnostic Command Set The following documents are relevant to the discussion: X3T9.2/89-119 Diagnostic Command set for SCSI-2 devices. [Hospodor] X3T9.2/89-134 Proposed SCSI Diagnostic Enhancements [Dominguez] X3T9.2/89-135 Proposal for an SCSI Diagnostic Command Set [Pickford] X3T9.2/89-136 Overheads - SCSI Diagnostic Command Set Presentation [Pickford] X3T9.2/89-137 Recommendations to SCSI committee [Yomtov] George Canevit, co-chairman of X3B7.1, Ricardo Dominguez and Reuben Yomtov, members of X3B7.1, attended the X3T9.2 working group meeting. George Canevit presented an overview of the X3B7.1 activities. The Diagnostic Command Set (DCS) ad-hoc committee has developed a working paper for SCSI. The goal was to communicate what was necessary to test an SCSI device. At issue is the the cost of evaluating a disk drive, the inability to test early SCSI drives, and lack of common test points among drive manufacturers. Low-level testing of embedded SCSI disk drives is difficult: The interface masks the drive electronics. Often people have had to buy the HDA with another interface to evaluate the HDA, but this means that a second test is necessary to evaluate the controller electronics. Also, SCSI drives tend to have very large capacities, increasing the test time. There are five core commands in the DCS: Format, Format Track, Erase, Diagnostic Read, and Diagnostic Write. These commands are defined as pages of the SEND DIAGNOSTIC and RECEIVE DIAGNOSTIC RESULTS commands. Four additional commands are used that already exist, but some of the fields are re-defined. These are: READ BUFFER, WRITE BUFFER, WRITE LONG, and READ LONG. The existing SCSI commands need additional functionality to provide the features required for DCS. There is a need to be able to do more than one sector at a time, to have a block count instead of a byte count, and to be able to do all sectors in one revolution instead of one sector per revolution. Steve Cornaby countered that one can write a whole track with the WRITE LONG command. There are five HDA tester manufacturers in the industry, but there is no common method to test the SCSI drives. It is George Canevit's opinion that the promotion of the DCS will lead to better competition through simplified testing. Some low-level testers will require secondary connection to a zero cross- over detector in the read channel or a differentiated analog output from the HDA. Ricardo Dominguez stated that the DCS is useful even if this secondary connection is not available. There was an extensive discussion over whether this level of testing is required. Gene Milligan requested the parent document on head and media test be distributed also. George Canevit stated that the test signal document is complete, and the test method document should be completed within the next two weeks. The DCS applies to a subset of SCSI direct-access devices, specifically rigid disk drives. The remaining direct-access devices and all the other device types are not addressed by the DCS. George Penokie asked how the DCS addresses the internal testing of the drives. He stated that this functionality appears to be missing. The major points of Ricardo Dominguez's presentation were: * Objective - Quantify disk drive performance and reliability throughout the product life, drive development, drive manufacturer production, system integrator evaluation and production, and drive repair. * Disk drive testing frequently involves assessment of: HDA mechanics, media parameters, head parameters, and drive electronics. * Access to critical drive signals has been necessary. Some these include: read analog, position error, index, write fault, and head select signals. * Some form of physical addressability has been inherent to testing of key drive attributes: media integrity, servo analysis (forced head offset), and read channel. * Test data correlation, test methods, and test equipment have frequently been areas of inefficiency. * Often the system integrator is left to deal with drive deficiencies that could have addressed earlier in the product life. * Continuing drive evolutions make it increasingly difficult to test drive attributes. * As a systems integrator, he desires provisions for a diagnostic vehicle to better accommodate testability. Bob Snively questioned that what was meaningful before, may no longer be meaningful, because the technology is eliminating the need for analog analysis. Furthermore the causes of loss of data have more to do with environmental conditions, power supply problems, and noise in the system. Gary Stephens asked what is it that X3B7 wants to measure? He stated that the results desired should be defined for each function and not the particular command(s) that do it. Ricardo Dominguez appears to want to be able to test the SCSI drives in a similar fashion to the way lower-level interfaces are tested (e.g., ESDI, ST-506). There seems to be a three-way stalemate, with the needs of drive vendors, drive tester vendors, and systems integrators all being slightly different. There is general agreement that there is a need to have ways to obtain testability results. The conflict is over which tools and methods should be used to get these results. Several drive manufacturers expressed concern about users getting into the drives and causing problems with the drive's functionality and integrity. Doug Pickford presented an alternative proposal for diagnostic commands (see 89-135). Is the physical mode of operation allowed to impact the logical mode? Doug raised a problem that exists with the page structure is that a DATA IN and DATA OUT phase are not allowed within one command. Steve Cornaby suggested that the diagnostic functions be put in the CAM level and driver translation tables could be licensed by manufacturers to customers. All manufacturers provide similar vendor-unique testing functions, but they all are slightly different and are enabled in different ways. He also requested a special subject working group be established to review 89-135. There was some consensus on this approach: Doug, Steve, and anyone else who is interested plan to meet Monday morning December 4, before the next plenary meeting. Reuben Yomtov presented his recommendations on diagnostic functions. He has a preference to perform these functions at the operating system level. Several utilities have been developed in the UNIX environment for SCSI devices to test those devices. Some of his recommendations are already addressed in the SCSI-2 draft standard and several overlap with other proposals received at the meeting. X3B7.1 planned to meeting on November 6-7 in Phoenix. The working group asked that they review the proposals and attempt to consolidate them into one document. John Lohmeyer thanked the X3B7.1 members for participating in the working group meeting and asked that they consider attending another joint working group meeting. The afternoon of the January 10 meeting was tentatively selected. 2. SEARCH Command modifications [Snively] Bob reported that to date he has found little interest in the SEARCH command among the user community. He will keep on searching. 3. Alternate Physical Protocol Layer Preparation (89-130) [Stephens] Gary Stephens had prepared a proposal on a parallel packet SCSI protocol. The document showed the changes needed in sections zero through six to allow for packetized SCSI. Gary made a presentation of the salient features of packetized SCSI. There was debate on whether or not the packetization should be applied to a parallel interface. Several people felt that the efforts should concentrate on a serial interface implementation where packetization is necessary. An implementation on a parallel interface permits device firmware to be written that could be used in either environment. Thus the device would have a choice of interfaces. Due to formatting problems, 89-130 will not be included in the November mailing. Gary plans to bring cleanly formatted copies to San Diego and the re-formatted copy will be included in the January mailing. Gary accepted an action item to develop a presentation for the next plenary meeting. 4. Single-Cable 16-bit Wide SCSI (89-094R4) [Penokie, Lohmeyer, Lamers] George illustrated the A, B, and P cable combinations and the problem that can occur with a P/B combination. These devices, capable of 8-bit, 16-bit single connector, 16-bit dual connector and 32-bit transfers, need to determine how they are going to arbitrate and communicate. The proposal evolved during the meeting, with the first approach being: Cable Arbitration Data Transfers ----- ----------- -------------- A 0-7 0-7 P 0-15 0-15 A/B 0-7/16-31 0-7/8-31 P/B 0-15/16-31 0-7/8-31 Note that A/B cable setups do not arbitrate on bits 8-15 in this proposal. While this approach might work, it is complicated. End users may be confused by the configuration rules. In keeping with the KISS principal (i.e., Keep It Simple Stupid), the group reduced the table to: Cable Arbitration Data Transfers ----- ----------- -------------- A 0-7 0-7 P 0-15 0-15 A/B 0-7 0-7/8-31 Discussion then turned to looking at a way to upgrade P cable systems to 32 bits. A second 16-bit cable was defined, call the Q cable. The consensus was to use a 68-pin connector for the Q cable, the same as the P cable except omitting the control lines. George Penokie agreed to revise 89-094 to incorporated these changes. Discussion then went off on whether the P cable proposal should be included in SCSI-2 or SCSI-3. The same points were made as in the last several meetings with the same results: no consensus. Two new compromises were discussed: 1) add an implementors note to SCSI-2 saying that the committee is investigating an alternative solution in SCSI-3 and 2) add an informative appendix to SCSI-2 giving early information on the P cable approach. Gary Stephens requested the chair take a straw poll on the willingness of the working group to delay SCSI-2 to include 16-bit single connector. There was so much discussion on how to phrase the straw poll and on how much delay might result, that the straw poll was never taken. 5. Cable configurations in wide SCSI [Penokie, Cornaby] This item was dropped since we decided that A/B and P cable systems are mutually exclusive. 6. Autosense [Snively] Dropped at Bob Snively's request. This is included in Gary Stephen's 89-130 proposal. 7. Arbitration fairness (89-61) [Penokie] Deferred. 8. Arbitration unfairness (89-65) [Buesing] Deferred. 9. New SCSI transceivers (89-131, 89-132) [Murdock] Gary Murdock gave a presentation on two proposals, one for short distance differential SCSI transceivers and one for long distance, small swing differential SCSI transceivers. Gary said that the presentation was for purpose of determining whether a market exists for such transceivers. The short distance differential SCSI (89-131) allows higher transfer rates on a 6 meter cable with improved noise margins. Devices could be configured to work on old-style single-ended cables. These parts could work on longer cables, but at reduced transfer rates. The long distance, small swing differential SCSI (89-132) supports up to 20 Megabyte/sec on an 8-bit cable. Can drive long cables (up to 50 m) although arbitration timing would have to be changed for anything over 25 meters. Both proposals support high integration levels, possibly including all the transceivers on the protocol chip. 10. Signal quality issues on differential (89-124) [Curry] Deferred. 11. Multiple Port Function and Path Control (89-133R0) [Stephens] Gary Stephens briefly presented a proposal for handling multiple ports for SCSI with multiple busses. This allows multiple paths for initiators to communicate with targets. Gary accepted an action item to develop a presentation for the next plenary meeting. 12. SCSI Working Group Schedule Bob Snively is hosting the next set of working group meetings in San Jose on January 8-12, 1990. While the contract had not been signed yet, the hotel will likely be the Red Lion Inn near the San Jose airport. A CAM meeting is scheduled for January 8, 1989 in the same hotel. The SCSI working group meeting is scheduled for January 9-10, 1989, with the DCS scheduled for 1:30 p.m. on January 10, 1990. The Fiber Channel meeting is January 11-12, 1990. March 5-9, 1990 - tentatively hosted by Jeff Stai in Irvine. May 7-11, 1990 - tentatively hosted by Joe Lawlor in Chicago. July 9-13, 1990 - no host or location selected yet. September 4-7, 1990 - no host or location selected yet (no CAM meeting). October 29-November 2, 1990 - no host or location selected yet. The 1990 working group meetings are scheduled for five days as follows: CAM Meeting Monday SCSI Working Group Meeting Tuesday & Wednesday Fiber Channel Working Group Meeting Thursday & Friday The working group meeting was adjourned at 1:00 p.m. on October 31. Following a lunch break, Gary Stephens gave a presentation on the Fiber Channel project that had been given by Joe Mathis at the previous Fiber Channel meeting.