August 31, 1989 TO: X3T9.3 Fiber Optic Study Group Members FROM: Roger Cummings SUBJECT: FIBER CHANNEL WORKING GROUP MINUTES Please find attached a draft of the minutes of the ANSI X3T9.3 Fiber Channel Working Group of August 23, 1989. Note that there are also nine Attachments to the minutes that relate to presentations at the meeting. Note that this package of minutes and attachments will be directly mailed only to those persons who have attended at least one of the two most recent working group meetings. The full package of minutes and attachments will continue to be included in the regular X3T9.3 mailing that results from the plenary meeting following the working group (in this case the October mailing). Thus interested parties that are unable to attend the working group meetings are strongly advised to subscribe to that mailing. The next Fiber Channel Working Group is being hosted by Terry Anderson of Ancor Communications at the Embassy Suites Hotel in Bloomington, Minnesota on September 11 and 12, 1989. The meeting will conclude at 3pm on Tuesday to allow people to head home. The hotel is located close to the Minneapolis-St. Paul airport, and its phone number is (612) 854-1000. For further info, contact Terry or Kathy Woodford at (612) 932-4000. Note that at this meeting it is intended to make a decision regarding a coding scheme that will be proposed for adoption at the October X3T9.3 plenary. A schedule of X3T9.3 meetings (both plenaries and working groups) is attached, along with a meeting notice for the October plenary to be held in Raleigh, NC. If there are any corrections required to, or omissions noted from, the minutes I can be reached as follows: Phone: (303) 673-6357 (Business) (303) 665-0761 (Home) Telex/MCI Mail: (650) 289-5060 Fax: (303) 673-5891 Regards Roger Cummings Senior Engineer Subsystems Controller Development #nbj0/rc MINUTES OF THE NINTH FIBER OPTIC WORKING GROUP MEETING The Ninth meeting of the ANSI X3T9.3 Fiber Optic Working Group was held on August 23 at the Sheraton Hotel in Colorado Springs, CO. The plenary week during which the meeting was held was hosted by Del Shoemaker of Digital Equipment. A total of 51 people attended, as follows: AMD Jim Kubinec AMDAHL Rich Taborek AMP Charles Brill ANCOR COMMUNICATION Ron Benton ANCOT CORP. Jan V. Dedek ARTEL COMMUNICATIONS CORP. Jan Helbers AT&T James E Morris BULL HN Duane Barney CANSTAR Kumar Malavalli CDC Wayne Sanderson CONVEX COMPUTER CORPORATION Thomas W. McClendon Gary Stager CORNING INC. Kevin Able CRAY COMPUTER CORP. Rod Ruesch Dale Smith CRAY RESEARCH INC. Wayne Roiger DIGITAL EQUIPMENT CORPORATION Chris Baldwin Del Shoemaker DUPONT Mike Kaplit ENDL I Dal Allan FORD AEROSPACE Gary Waldeck FUJITSU AMERICA Jim Luttrull GIGABIT LOGIC Carl Deierling HARRIS SPACE SYSTEMS CORP. Allen Williams HUGHES AIRCRAFT CO. Douglas Cook IBM Henry Brandt Horst L Truestedt IMPRIMIS Gene Milligan IPITEK Inc. Michael Pugh LAWRENCE LIVERMORE NATIONAL LABS. Paul Rupert LOS ALAMOS NATIONAL LAB Don Tolmie MADISON CABLE Bob Bellino MAXTOR CORP. Larry Lamers MOTOROLA INC. Ernest Johnson NATIONAL SEMICONDUCTOR Sam Laymoun NCAR Gil Green NCR John Lohmeyer NETWORK SYSTEMS Ken Drewlo PCO Jim Goell PRISMA Mike Baird SCIENTIFIC COMPUTER SYSTEMS Don Deel SHELL DEVELOPMENT CO. Patric Savage SIEMENS Schelto Van Doorn STORAGE TECHNOLOGY CORP. Roger Cummings Floyd Paurus Steve Zanowick STRATUS COMPUTER Steve Kohaluni ULAN CORP. Edouard Rocher UNISYS Doug Morrissey UNIVERSITY OF COLORADO Mark Fink US SPRINT Michael Sobek The meeting began with a greeting by the Chairman of the Working Group, Dal Allan of ENDL Consulting. Dal noted that this meeting had been intended as a vehicle to attract people that attend the Plenary Week but not separate working groups. He reminded the Group that at the last meeting it had been agreed that the next (September) Working Group meeting would be devoted to the subjects of coding and topology. He also noted that the intention was to produce a recommendation from that meeting to take to the October plenary meeting to allow a formal vote on the adoption of a coding scheme. The first presentation was made by John Lohmeyer of NCR on the subject of SCSI requirements for the Fiber Channel. A copy of John's slide is Attachment 1. John began by introducing SCSI as an interface capable of 5 Megabytes/s today which has the capability of reaching 40 MB/s with a 32 bit parallel bus in the near future. He noted that the Fiber Channel requirements that he was presenting had been derived from within NCR as well as from the X3T9.2 plenary. He noted that the Cost/Node was for all of the hardware from a microprocessor bus through to the connectors, and contrasted the $100 - $250 figure given with a $50 - $60 MLB-type cost for a set of differential transceivers and a protocol chip for today's SCSI. Jim Goell of PCO asked if there was a low end application for fiber ($40/node @ less than 200 Mb/s), and John stated in reply that today's SCSI marketplace was 70%+ single-ended and thus limited to a 25 feet operating distance by poor transmission characteristics. Therefore if a fiber channel version could be specified where the cable cost in the area of $25 for a 50 meter length, with a mated connector cost of less than $10 and a protocol chip cost of $5, John said that he could foresee fiber being competitive with todays implementations, providing better signal quality and thus having a huge potential market. However nobody at the meeting was confident enough to assure John that this was possible! Wayne Sanderson of Control Data asked if there was a market for a serial copper version of the Fiber Channel and John, on being told that the copper version would have lower performance, noted that low performance SCSI applications are a thing of the past. John emphasized that a small connector footprint was a key requirement, given the trend to smaller disk diameters and peripheral packages. Therefore he believed that there would be a market for a serial copper version if it was cost competitive, flexible, and lead to a smaller footprint. Schelto Van Doorn of Siemens pointed out that a connector that integrated both power and logic functions would be possible with fiber, and a short discussion concluded that this might well be attractive as a means of reducing the overall connector footprint. Dal Allan then briefly discussed possible applications for a SCSI Fiber Channel that had been discussed in the X3T9.2 committee. The main application that was seen was not one of a single workstation with many local peripherals, but rather the interconnection of peripherals at existing locations throughout a building. Some comments were made after John's presentation to the effect that nobody had yet been found to present a system based on plastic fiber to the Working Group. Jim Kubinec of AMD stated that he was trying to arrange for Codeknoll to present on that subject at the September Working Group meeting. The next presentation was made by Roger Cummings of Storagetek on the subject of the FC-2 protocol requirements to support the IPI Level 3 protocols. A copy of Roger's slides is Attachment 2. Roger's requirements were at a very detailed level and he emphasized that a major advantage of the current IPI Physical interface that must be retained in the Fiber Channel is the ability to tune the performance of the channel from a host computer. Wayne Sanderson asked if all of the requirements were intended to be met at the lowest level in the Fiber Channel, and was told that they were not specific to any level within the Fiber Channel and were only intended to be features of the service to the IPI protocols. Roger emphasized that, in talking to existing IPI users, nobody had been interested in a scheme such as dual counter-rotating rings for redundancy. There was a strong preference for redundancy being achieved by the same methods as today - namely by multiple separate paths with no common point of failure. Dal Allan then lead an effort to determine which of Roger's requirements are features of the existing IPI system, and which are specifically for the Fiber Channel (the attachment shows this distinction). Dal also noted in passing that there is a proposal to move SCSI towards a Master/Slave philosophy and thus SCSI and IPI are converging. Paul Rupert of Lawrence Livermore Labs questioned the requirement for a burst level acknowledgement scheme and was told that, while this may only be a feature of the interface to the IPI protocols, some sort of acknowledgement may be required at a lower level within the Fiber Channel to provide for source buffer control etc. Roger was followed by Mike Pugh of Ipitek (formerly Integrated Photonics Inc.) who presented a more detailed discussion of the architecture for the Fiber Channel interface hardware that he had originated at a previous Working Group meeting. A copy of Mike's presentation is Attachment 3. Mike began by defining two interface types using multimode and single mode components respectively. His multimode type used 62.5 um core fiber and Schelto Van Doorn noted that 50 um fiber was the standard for the world outside the USA (except for some installations in Europe which use 62.5 fiber to be common with the USA). Mike agreed that this was true, but he did not think that the difference was significant as there is much equipment that works with both standards. Mike had also defined a connector (the ST type) that is popular in the USA but not elsewhere. He noted that the connector exhibits high reflections and inconsistent laser light launch capability, but he stated that many people are working on solving these problems. Dal Allan asked Mike to project the speed of a multimode fiber system five years into the future. Mike replied that three or four years ago fiber was indifferent quality due to manufacturing process problems, but that these problems have largely been resolved and fiber today routinely achieves 800 MHz.Km. He emphasized that this figure is for laser light only, as the spectral width of an led source limits most led-based systems to 300 MHz.Km. Given that there are a number of physical problems with increasing the frequency of leds in addition to the above, Mike concluded that he did not expect the speed of a multimode system to increase significantly in the future. Mike then moved on to describe a single mode interface type. He quoted a quantity price of $270 for lasers in receptacles and noted that a Class 1 laser fits this application best regardless of the fact that it implies greater receiver sensitivity. He stated that it is possible to limit the output of a Class 3B laser such that it meets the Class 1 restrictions, and noted that AT&T do this already. The definition of the Parallel Interface Fiber Modem was then reviewed, with Mike stating that he was presently visualizing a chip having in the order of 30 pins. He noted that the tradeoff is to minimize the number of pins to allow for future scaling down while having enough pins to reduce the frequency to something that can be handled by standard processes. He estimated that 120 MHz and 240 MHz are the state of the art for CMOS and BiMOS processes respectively. Two people immediately took issue with this statement. Gary Waldeck of Ford Aerospace said that they have CMOS technology operating at 320 MHz. Jim Morris of AT@T stated that rates of 1.2 GHz are not going to cause a technology change, noted that SONET components that operate at 2488 MHz may be of interest, and thus asked that the parallel interface width not be fixed at the present. The impact on the modem design of the synchronization method was also briefly discussed. A coding scheme such as 8B/10B uses a coding violation to establish sync, and this is simple to include in the modem because it is instantaneous. If a telco-type scheme using one bit is used, however, the received data has to be shifted until the right boundaries are established. Mike's solution to this was to make the method of synchronization a programmable option. Mike emphasized that he did not intend to make the modem interface mandatory - in that if a user wished to employ a single chip solution it would still be regarded as Fiber Channel compatible if it met the optical interface requirements. He did, however, express the wish that the modem interface would be stable enough such that new technology would not cause a re-design. Dal Allan strongly concurred with this approach, stating that interoperatibility would be defined at the "plug" interface and multiple sources defined at the silicon (modem) interface. Chris Baldwin of Digital Equipment Corp. noted that the definition of a modem interface would support the notion of having different coding schemes for different FC-0 classes. The following discussion reaffirmed that a single coding scheme is the defined goal, but noted that mode than one scheme may be adopted if the tradeoff at the next Working Group meeting produces good reasons for doing so. Mike noted in closing that the timing and coding requirements contained in Chapter 5 of his presentation were derived from experience rather than by algorithmic methods. The next presentation was given by Jim Goell of PCO on the subject of optical interface parameters. A copy of Jim's presentation is Attachment 4. He proposed use of the format being developed for SONET, and both the table structure and the accompanying notes defined by this effort are contained in the Attachment. Jim also provided a sample set of parameter values for a single mode system. He stated that the variation in mean launched power shown (+/- 3dB) is the smallest possible given the uncertainty of the back facet monitor, and noted he has seen hot setup (trimming resistors etc.) being used even at this level. He emphasized that making a smaller variation a requirement would mean that a cooling element would be required, and after discussion it was again agreed that the requirement is that a cooling element not be used. Jim noted that the European requirements for fiber optic systems are more strict than those in the USA, and thus for a product to be saleable worldwide these European requirements must be met. Mike Pugh noted that FDDI uses a different set of parameters, but Chris Baldwin reassured him that the differences are minor. Don Tolmie of Los Alamos National Labs. then presented a view of the different components of the Fiber Channel in terms of a set of layers. A copy of Don's presentation is Attachment 5. It was generally agreed that this view was complementary to Mike Pugh's architecture, and that the layer diagram had some benefit. However some controversy arose when Don moved on to consider if the SCSI, IPI and HSC protocols for the Fiber Channel should have any compatibility other than in a single byte at the head of a frame. This lead to an effort to identify the common features of the three protocol types, which resulted in the following matrix representing present features, with new features thought to be required for the Fiber Channel shown in parentheses: FEATURE HSC IPI SCSI Destination Address Y Y Y Source Address Y N (Y) Y Connection Control Flag Set up Destination IDs N (Y) N (Y) N (Y) Do not break connection Y N (Y) N (Y) Circuit Switch Topology Setup Next Destination N (?) N (?) N (?) Application Identifier Y N (Y) N (Y) No. of Bursts in Packet N N N Sequential Burst Number N N N Identify Last Burst Y Y Y Error Control CRC Y N (Y) N (Y) Parity Y Y Y Error Detection and Correction N N N As can be seen from the matrix, enough requirements for each protocol are the same to make it worthwhile to invent a single Fiber Channel service to satisfy all three requirements. It was agreed that the primary need for commonality is in routing, and resolving local addresses. As at least some of the topologies under discussion will involve the use of routing and/or switching elements and unless a common scheme is adopted by all three interfaces, then such devices will have to be unique for each protocol. If there is common format for every header, however, then common routers and switchers could be developed that would work with all configurations. Wayne Sanderson repeated his comment of an earlier meeting to the effect that today system manufacturers have to support a large number of different interfaces types on their systems at considerable cost. He therefore sees the Fiber Channel as fulfilling a desperate need to simplify this problem by allowing a single interface design to be used to connect to a multiplicity of peripheral types. Chris Baldwin then attempted to specify the error rates required for the Fiber Channel. Clearly the transfer rate should have an impact on the specification, because the time between recovery events would be dramatically different. Jim Morris calculated that at 200 Mbs with a BER of 10E-12, the total number of errors requiring retry would "waste" about 35 usec per year, and it was agreed that the undetected error rate should be specified as 10E-18. Chris objected to a detected BER of 10E-12 on 1 Gbs links, and proposed that it be made 10E-16, and that the undetected error rate be 10E-36. It was noted that this is longer than the life of the universe and although Chris could not justify these numbers at the time, he did promise a presentation on the subject at the next Working Group meeting. Don Tolmie then distributed copies of the four skeleton Fiber Channel standards that had been updated at the last working meeting. These documents are Attachments 6 thru 9 inclusive. Jim Morris of AT&T then gave a short verbal presentation which disputed some of the claims made earlier that, for frequencies above 1 Gigahertz, gallium arsenide was the only possible technology. Jim stated that within AT@T he knows of two silicon bipolar processes that can operate at 1 GHz or greater. He noted that there are also some cmos processes in the industry with one micron geometry that are fast enough to consider for Fiber Channel functions, but he cautioned that these speeds are achieved at a cost of higher power dissipation. In fact he used an example of a SONET cmos transmitter at 622 MHz that is being redesigned using mostly nmos to operate in the Gigabit range. He referenced a chip set made by Avantek which has similar performance, and noted that the Japanese have both GaAs and silicon processes at these speeds. Jim stated that AT&T has internally two gallium arsenide processes, and that they are in the process of migration. He emphasized that gallium arsenide is the highest speed process that is in place and in manufacture, and that it offers the greatest receiver sensitivity, but he questioned if such performance was really required. He was strongly of the opinion that with a reasonable specification at the frequencies under discussion a choice among a number of technologies will be able to be made. It was then reaffirmed that the coding scheme decision would be made at the next Working Group meeting. Persons that had proposed coding schemes were asked to contact John Severyn at (415) 422-5383 (fax 423-8567) to agree on parameters. Henry Brandt of IBM asked if the actual error characteristics of the media had been established and this lead to a discussion on the interpretation of the various bit error rate numbers. Patric Savage of Shell Development was of the opinion that the error rates should represent end to end performance (i.e. via intermediate switches), but there seemed to be no consensus on how this could be specified. Chris Baldwin volunteered a presentation by Digital at the next meeting on bit error rate requirements for links. It was also agreed that the October Working Group meeting would include discussion of component parameters. ACTION ITEMS 1) John Severyn of Lawrence Livermore Labs. to produce a matrix comparing the features of the FDDI 4B/5B, IBM 8B/10B, and Naked coding schemes. 2) Mike Pugh to further define the parallel transceiver interface in his Low Level Architecture. CLOSED 3) Paul Scott to produce a definition of Bit Error Rate. 4) Roger Cummings to identify the requirements placed on the FC-2 protocol by the IPI Device Generic protocols. CLOSED 5) Bob Snively to identify the requirements placed on the FC- 2 protocol by the SCSI command protocol and message structure. CLOSED (by John Lohmeyer) 6) Don Tolmie to produce updates of his draft skeleton standards. CLOSED 7) Henry Brandt of IBM to respond to the comparison of the error handling characteristics of the 4B/5B and 8B/10B codes. 8) Kevin Able of Corning to define the parameters for the media of both the Multimode classes (both 50 and 62.5) and the Single Mode class. 9) Chuck Brill of AMP to define the parameters for the media- end connectors of the Coaxial, Multimode and Single Mode classes. 10) Bob Carter to define the parameters for the patch panels of both the Multimode classes and the Single Mode class. 11) Paul Scott to define the parameters for the media and transceivers of the Coaxial class. 12) Schelto Van Doorn of Siemens to define the parameters for the transceivers for the Multimode LED class. 13) Ron Soderstrom of IBM to define the parameters for the transceivers for the Multimode Laser class. 14) Jim Goell to define the parameters for the transceivers for the Single Mode class. CLOSED 15) Gary Labelle of Avantek to try to get a representative of Hewlett-Packard experienced in their plastic fiber system to take responsibility for defining the components for the Plastic Class.