MINUTES IEEE 802.3z ENSCHEDE, THE NETHERLANDS JULY 9-10, 1996 As corrected and approved by 802.3z on 11-Sept-1996 1. WELCOME AND INTRODUCTIONS Mr. Howard Frazier opened the meeting and welcomed all to the organizational meeting for the 802.3z task force. This meeting is an important milestone in the development of gigabit CSMA/CD, even though the technical work of the Task Force will be a continuation of the Higher Speed Study Group. 2. AGENDA Mr. Frazier cajoled Mr. Robert Grow into taking these minutes with the promise that he "would never have to do them again". Mr. Frazier asked for approval of the proposed agenda and scheduled the technical presentations. 3. P802.3z STATUS REPORT The PAR for the development of a gigabit CSMA/CD standard was approved on: 14 March 1996 by the 802.3 WG 14 March 1996 by the 802 Executive Committee 19 June 1996 by NESCOM 20 June 1996 by the IEEE Standards Board With the authorization of a Gigabit Task Force by 802.3 on 8 July 1996, Mr. Frazier indicated it was now appropriate to organize the Task Force. 4. ELECTION OF OFFICERS Mr. Geoff Thompson opened the floor for nominations for Chairperson of the 802.3z Task Force. Mr. Howard Frazier was nominated and elected to the position by acclamation. The floor was then opened for nominations for Editor-in-Chief of the standard. Mr. Howard Johnson was nominated and elected to the position by acclamation. 5. VOTING RIGHTS & RULES Mr. Frazier outlined two options for voting rules for the working group, and left open for suggestion any alternatives. After discussion, voting was held on the two alternatives: A. All individuals present at the time a vote is taken may vote, if they feel qualified to do so. B. Only individuals with voting rights in the 802.3 WG have the right to vote in the 802.3z Task Force. Option A was selected over option B by a vote of 62 to 11. The voting rules adopted without objection are that technical decisions require a >=75% approval and procedural decisions require a >50% approval. 6. EMAIL REFLECTOR AND WEB/FTP SITE The Task Force will continue to use the reflector set up for the high speed study group hosted by IEEE. stds-802-3-hssg@mail.ieee.org The reflector is for announcements comments discussions and dissemination of information related to the work of the task force. It should not be used for recruiting, advertising, soliciting, flaming, or whining. To be added to the reflector, send email containing the following line: subscribe stds-802-3-hssg to either majordomo@ieee.org or howard.frazier@sun.com. Subscriptions are on an individual basis, no proxies or reflectors will be subscribed. A web/ftp site is also hosted by IEEE for the Task Force. ftp://stdsbbs.ieee.org/pub/802_main/802.3/gigabit The site contains drafts (when there is one), minutes, future meetings and technical presentations. Presenters are requested to submit ASCII or if necessary PDF versions of their presentations to howard.frazier@eng.sun.com for upload to this site. A straw poll favored posting most presentations to the reflector. Large presentations will have an email pointer posted to the reflector indicating the ftp location of the presentation file. 7. PRELIMINARY OUTLINE FOR P802.3z Mr. Frazier presented a possible outline for a 802.3z supplement to 802.3. The structure is based on the work done in development of 100Base-T. As part of this task, naming conventions were included using the root 1000Base-. Some discussion ensued about clause numbers, splitting the example clauses, etc. The group was asked to think about the outline and naming for future refinement. 8. STANDARDS DEVELOPMENT TIMELINE Mr. Frazier showed the development timeline for 100 Mbps standards and a proposed timeline for a 1 Gbps standard. This timeline included a proposal cut-off date of November 1996, a first draft in January 1997, last feature added and Task Force review of a second draft in March 1997, a working group ballot in July 1997, last technical change in September 1997, LMSC ballot in November 1997 and the standard approved in March 1998. There was some discussion and some disagreement as to the ability to achieve the proposed dates for a copper PHY. 9. REVIEW HSSG OBJECTIVES Mr. Frazier indicated to the group that refinement of objectives 11b and 12 was appropriate. He indicated that he wanted the group to discuss and decide later this week on objective modifications. Questions about link lengths and collision diameter evoked some explanations. Link lengths longer than the collision domain diameter are appropriate for full duplex operation. The 25 meter link length objective for copper is not intended for horizontal wiring, but rather for machine room wiring. A 100 meter copper link length would be a candidate for horizontal wiring applications. 10. TECHNICAL PRESENTATIONS Ahmad Nouri -- Fiber Length Survey Mr. Nouri presented a summary of a survey performed by Compaq In response to questions, Mr. Nouri clarified that a respondent to the survey would have links included in multiple categories (e.g., 3 links between closets, 2 links form closets to basements, etc.) The tables in the presentation are on the number of fiber runs (not necessarily currently in use). The survey did not ask about jumper lengths, so respondents may or may not have included jumper length in responses. Mr. Nouri will post the survey questions with his presentation. Alan Flatman -- Fiber Usage in Europe Mr. Flatman gave an informal presentation about the installed fiber in Europe. Preliminary results from a limited survey of big users and cable supplier indicates that fiber usage is somewhat limited. There is a mixture of 50 and 62.5 micrometer cables installed. There is hardly any single mode installed. Most buildings would be covered with runs less than 300 meters. Low rise installations (approximately 5% of installations) require greater than 500m links. Mr. Flatman indicate that the data are not sufficient to draw hard conclusion and additional work is appropriate. He will not be able to continue the study without funding. Igor Zhovnirovsky Mr. Zhovnirovsky was reporting for Chris Deminico of Digital. They are initiating a survey. The survey will help to understand the value of characterizing shorter link lengths (less the 1 km) since optical performance is non-linear. Questions posted to the reflector will be considered for inclusion in the survey. Steve Swanson -- Fiber Bandwidth Requirements for 802.3 Gigabit Ethernet Mr. Swanson summarized some information presented at the Wakefield meeting and the group response to that information. The presentation includes table showing distances for various fiber bandwidths, source rise times and spectral widths. He recommended allowing both short wavelength and long wavelength devices. 50 micrometer fiber will support 500 meter links and 62.5 fiber 200 meter links. Additional study is required to extend link lengths. Michael Hackert -- TIA FO-2.2 Task Group on Modal Dependence of Bandwidth Mr. Hackert's presentation describes the objectives of the Task Group and its status and future activity. The group is developing recommendations for short haul interconnect applications, including test that are more representative of installed cable performance. No conclusion can yet be drawn from the work of the group, nor does the Task Group have conclusion dates that can be used by 802.3 at this time. He will convey the schedule of 802.3z to the Task Group. Anyone desiring to participate can contact him at hackert_mj@corning.com. Paul Kolesar -- Issues and Opportunities for the Installed 62.5 um Fiber Base Mr. Kolesar provided background on building cable distances, 62.5 distance verses bandwidth, and techniques for improving distance. Techniques discussed include lower baud encoding methods, equalization, use of longer wavelenth transceivers, multi-wavelength, sub-rate operation, restricted launch conditions and parallel transmission. Mr. Kolesar requested feedback on the alternatives, and was told sub-rate operation would not be considered by most participants. It was also noted that use of multilevel coding would create a power penalty problem. Del Hanson -- IEEE 802.3 1.25 GBd MMF Link Specification Development Issues Mr. Hanson presented additional work resulting from previous presentations on multimode fiber link lengths. This presentation explored the use of 1300 nm single mode transceivers on both single mode fiber and multimode fiber. 62 MMF dominates the installed base, but 850 nm technology doesn't support the required 550 m building backbone link length. 1300 nm devices will support the desired longer link lengths with smaller optical power budgets. They also provide higher eye safety limits and better reliability, but they are more expensive than 850 devices. Mr. Hanson reiterated that 850 and 1300 devices are not compatible. He also indicated that he believed that 1300 nm VCSEL devices will be available, but probably not for three years as standard products. Mr. Hanson recommends specifying 850 nm for horizontal wiring (up to 500 meters with 50MMF and 250 meters with 62MMF). 1300 nm single mode transceivers should be used on MMF for backbone links to 550 meters, and on SMF for longer links. He also advocated continued efforts to extend 1300 nm link lengths on 62MMF to the 2 km campus backbone length. David Cunningham -- 125 GBaud, 550 m, Building Backbone Links on Installed 62.5 Multimode Fibre for IEEE 802.3: leveraging existing long wavelength transceiver specifications Mr. Cunningham presented a paper on the above topic. He reinforced many of the points made by Mr. Hanson, with additional information on European standards, installed base, and requirements. He indicated that 62.5 MMF is dominant in building backbone with a small population of SMF. He also said that 300 meter links extending home run from the horizontal wiring will be important for future installations. He described the effects of modal noise and efforts to improve test procedures for characterization of backbone links, and modeled and measured link performance including restricted launch conditions. He supported the use of 1300 nm single mode transceivers on MMF and indicated that restricted launch looked very promising for 2 km links on 62.5 MMF. John Bowerman -- Gigabit High Speed Serial Modules Mr. Bowerman's presentation covered the use of long wavelength lasers, selective mode launch and use of short wavelength lasers at 200 meters. His conclusion was that selective mode launch with 850 nm devices should eliminate modal bandwidth problems for 500 meter links, increase BER performance and reduce connector losses, thus achieving the goals of 802.3z. Howard Johnson -- PCS Update to Protocol Proposal Mr. Johnson provided a brief update to his PCS proposal. A number of slides were included for continuity and skipped during the presentation because there has been no change since the previous presentation of the material. The current version reflects changes to the propagation of code violations. Rich Taborek -- Gigabit Ethernet Serial Link Codes Mr. Taborek provided an updated version of the 8b10b code proposal, reflecting direction changes since its last presentation. He is considering swapping his recommendation for the T and R codes. Mike Salzman -- 1. Redundancy and Fault Reporting, 2. Test Modes, 3. Transparent Signaling Channel Mr. Salzman presented a proposal to implement enhanced redundancy and test capabilities for 802.3z. He suggested that the use of redundant links is a market requirement, and that reconfiguration between the links must occur quickly (10-100 msec.), and therefore, enhancements are needed for remote fault cases. He also advocated an independent (not related to redundancy) capability for adding a test mode similar to 100BaseT2. This test mode should support graceful take down of a link. The third part of the presentation was for a transparent PHY layer signaling channel to support protocol extension. This would be accomplished by defining additional ordered sets which logically represent idle in the interpacket gap but convey other signaling information. This capability would be negotiated during link configuration. The discussion included questions about multiplexing this traffic with MAC data, and what functions this would provide that would not be available through MAC frames similar to the 802.3x pause frame. Charlie Hochstedler -- Gigabit Ethernet Cat 5 Transceiver Mr. Hochstedler gave a brief presentation on coding alternatives for gigabit operation on four pairs of category 5 twisted pair. Simulation results for 2D 3x3 indicate operation at 10E-10 BER is possible. Robert Grow -- Consolidated Interface for Fiber or Copper PHY Mr. Grow presented an outline for a logical Media Independent Interface. The presentation evoked significant discussion on how similar the interface can be to the 100Base MII, while maintaining a "content-less" reconciliation sublayer. David Fifield -- Gigabit Ethernet MAC/PHY Interface Mr. Fifield proposed that a MII definition including an electrical interface is appropriate for inclusion in the standard. Comments varied from only wanting a logical interface, to wanting a full MII with connector. Moshe De-Leon -- flow Control For Gigabit Ethernet Mr. De-Leon's presentation analyzed the XOFF(time) flow control of 802.3x and PHY based flow control signaling as proposed in earlier meetings (Buffered Repeater). The analysis included latency calculations for the two alternatives and the buffering requirements. The overhead of the two methods was also evaluated as was their robustness. He also described extension of frame based flow control with credit and rate based schemes. His conclusion was that PHY based signaling provides no real advantage to XOFF flow control. He also recommended that more work is appropriate for review of alternate frame based schemes. Howard Frazier -- Scaling CSMA/CD to 1000Mb/s An Update Mr. Frazier reviewed the effects of gigabit operation on the physical and equipment topologies used in lower speed version of CSMA/CD, and how carrier extension solves many of the problems created in the scaling. The presentation includes modifications to Clause 4 that are required to implement carrier extension. He also reviewed progress in simulation and evaluation of alternative changes to MAC operation. He concludes that performance with the baseline proposal is adequate though refinements can be evaluated. Stephen Haddock -- Carrier Extension Issues Mr. Haddock reviewed the status of proposals made at the HSSG Wakefield meeting. The two basic proposals are to allow multiple small packets to be packed into a single carrier event, and to modify the backoff algorithm to reduce capture effect. Packet packing is viable, with simulation showing that packet packing largely compensates for throughput degradation introduced by carrier extension. It has also been validated since the Wakefield meeting that 8b10b coding can support a comma character after a run of R characters (the interpacket characters when extending carrier). Mr. Haddock also briefly reviewed his second proposal to establish context for the simulation results to be presented next. Based on the simulations, the modification is not a good idea. Jayant Kadambi -- Packet Packing and mTBEB Simulation Results Mr. Mohan Kalkunte and Mr. Kadambi performed additional simulations for MAC changes proposed by Mr. Haddock. The simulations verify that a significant performance boost is realized for 64 byte packets. With the workgroup distribution of packet size, improvement only occurs at high offered loads. Packet packing has no effect on large packet traffic. Packet packing adds minimally to capture effect. The modifications to the backoff algorithm did reduce capture effect but at significant cost. All other measures of performance were degraded. In general, there is an inverse relationship between throughput and capture effect as the maximum backoff time is varied. Moti Weizman -- HSSG CSMA/VCD(tm) proposal Mr. Weizman presented a proposal to improve system operation through the use of virtual collisions and frame batch. The proposal support operation over a collision domain of 200 meters with performance effected little by frame sizes, twice the network diameter before carrier extension is required, no loss of bandwidth when a collision occurs, and guaranteed fairness. The discussion included concern about fairness in access. 11. AOB Mr. Frazier asked the group to revisit the objectives for the Task Group as discussed earlier. A series of discussions and motions resulted on both the objectives raised by Mr. Frazier and others raised in the technical presentations. Motion #1 (Technical) Adopt H. Johnson proposal for application of long wavelength and short wavelength transceivers. M: Szostak, S: Hanson Passed: Y=77, N=0, A=13 H Johnson proposal: Use LWL transceivers for: Single-mode fiber (2 km) 11.c Multi-mode (50 and 62.5 micrometer) fiber (500 m) 11.a Use SWL transceiver for : Multi-mode (50 and 62.5 micrometer) fiber in horizontal fiber (100 m) and some building backbone. (Low Cost) Characterize each transceiver over its full operating range on each type of fiber. Do not rely on controlled mode launch assumptions until TIA FO 2.2 completes its work. Straw Poll Collision Domain Diameter of: a) 50 meters (small) 6 b) 200 meters (large) 65 c) 50 or 200 meters (small or large) 5 d) no CSMA/CD 3 Motion #2 (Technical) Modify objective 12 to read: Support maximum collision domain diameter of 200m. M: Grow, S: Thomson Passed: by acclamation Motion #3 (Technical) Add an objective to include a specification for an optional gigabit MII. M: Muller, S: Fifield Passed: Y=87, N=2, A=8 (A motion to amend to read "Add an objective to include a specification for an optional electrical gigabit MII" failed by a large enough margin that no count was performed.) Motion #4 (Technical) Modify objective 14 to read: Adopt flow control based on 802.3x. M: De-Leon, S: Salzman Passed: Y=66, N=2, A=30 Motion #5 (Technical) Modify objective 9 to read: Support fiber and copper media. M: Hochstedler, S: Andersson Motion to table Motion #5 M. Muller, S: unrecorded Motion to table passed: Y=36, N=28, A=16 12. NEXT MEETING An interim meeting for 802.3z will be hosted by Packet Engines Inc. in Coeur d'Alene, Idaho, September 9-11, 1996. Meeting information is posted at: ftp://stdsbbs.ieee.org/pub/802_main/802.3/gigabit/MEETING.TXT The location is Coeur d'Alene Inn and Conference center (800-251-7829), ask for the IEEE 802.3z room rate. Rooms are being held until August 8th. The nearest airport is Spokane, Washington. 13. APPROVE MAY MINUTES The minutes of the Wakefield meeting were approved without objection. Mr. Frazier thanked the attendees for a successful meeting and the meeting was then adjourned.