Minutes of the Ad Hoc group on Gigabit Ethernet Wednesday, November 8, 1995 Hotel Queen Elizabeth Montreal, Province of Quebec The meeting was called to order by Brian MacLeod at 2:00 pm The first order of business was the agenda for the meeting. The agenda proposed by Brian MacLeod was approved with an updated list of technical presentations. Brian reported that a tutorial session entitled "Emerging Physical Layer Standards for Gigabit Data Transmission had been held the previous evening. The tutorial was organized by Howie Johnson, and included presentations on the subject of physical layer specifications for ATM, Fiber Channel, and HIPPI. There was also a presentation on the characteristics of "Category 6" UTP cable. Colin Mick observed that the attendance at the tutorial session was in excess of 200 individuals. The next item of business was the election of a chairman. Nominations were opened. Howard Frazier was nominated by Howie Johnson and seconded by Bernard Daines. Howard was elected chair by acclamation Howard Frazier took an informal headcount that indicated that the attendance at this meeting of the ad hoc group on Gigabit Ethernet was in excess of 100 individuals. Next on the agenda were the technical presentations. The first presentation was entitled "Gigabit Ethernet: PHY Issues" and was given Dr. Howard Johnson. Applications for gigabit ethernet were discussed. The reuse of existing standards by adopting some existing LAN as a backbone was considered. The reuse of an existing PHY was proposed as an even better way to minimize development time and expense. And the Fiber Channel PHYs were proposed as the most promising. Discussion followed the presentation. Larry Birenbaum suggested using 4b/5b encoding used by 100BASE-TX even if 8b/10b encoding is a few dB better in performance. Bernard Daines pointed out that there is a difference between using 8/10b encoding just for data encoding and using some or all of the Fiber Channel protocol. Tom Gandy observed that 8b/10b encoding is a much better code than 4b/5b is terms of run length. John Renwick observed that you need large packets to achieve a high effective data rate and flow control to prevent overrunning the the target. Schelto van Doorn suggested that one of the Fiber Channel comma codes could be used for packet/data framing and the others used to support the Ethernet protocol. Shimon Muller asked what support will be needed for VLANs. Geoff Thompson said that one VLAN support being considered is an increase in frame size. The second presentation was entitled "SeriaLink, SeriaLink Core Technology Overview" and was given by Sanjay of LSI. SeriaLink is designed to be protocol independent and to be compatible with Fiber Channel. It currently uses 8b/10b encoding. It supports both copper and fiber media. BER for 10 meters of copper is < 10^-12. Transceivers are implemented in CMOS. Input data width is 20 bits at 53 MHz. Output data width is 10 or 20 bits. In response to a question, Sanjay did not know whether stepup transformers were used to achieve the signaling levels on copper media. The third presentation was on a Fiber Channel physical layer transceiver and was given by David Allen of TI. The transceiver has a 10 bit wide parallel interface, which is currently being standardized by ANSI, and uses differential PECL on the lineside. The transceiver is in a QFP package. The current version of the part uses a 5 Volt power supply. A 3.3 Volt version of the part is coming. The development of the transceiver is supported by HP, TI, IBM, Vitesse and NSC. The part is currently being used in "arbitrated loops". In response to a question, it was not known whether the part would work at 1.2 Gb/s. Schelto van Doorn offered to bring additional information on the 10 bit wide interface parts to the group. There is a technical report available on the 10 bit transceiver from ftp://fission.dt.wdc.com/standards/10bit/postscript 10 bit information is also available at http://www.wco.com/~schelto The fourth presentation was entitled "Gigabit Ethernet" and given by Ahmod Nouri. The need and requirements for gigabit ethernet were discussed. A review of the Fiber Channel protocol and PHYs was presented. And wording was proposed for a PAR. Gideon Pratt raised the question of what is the distinctive identity of gigabit ethernet? Geoff Thompson observed that "higher speed" in the third of the five criteria is relative. 1 Gb/s is not greater than speeds at which 802.12 might operate. Howard Frazier observed that the responses to the five criteria must be responsive to the actual questions in the criteria. The fifth presentation was an overview of recent presentations to 802.12 on technology for their higher speed work and was given by David Cunningham. Some initial observations were given on how 802.12 LANS might employ Fiber Channel PHYs and media in higher speed LANS. And presentations to 802.12 on the use of UTP as a physical media were noted. The sixth presentation was entitled "gigabit Ethernet" and was given by Brian MacLeod. The marriage of the 802.3 MAC with an existing PHY, specifically the Fiber Channel PHY, was proposed. The likely major applications for gigabit ethernet were presented and some projected trends were quoted. Geoff Thompson reviewed the 802 and the 802.3 five criteria. Howie Johnson asked whether 802.3 could adopt the 802 five criteria. Geoff Thompson responded that that it could. Bernard Daines asked whether 802.12 is defined as half-duplex and whether there might be a conflict between full-duplex demand priority and one MAC per 802 working group. Geoff Thompson responded that 802.12 already had two MACs, the end-station MAC and the repeater MAC. Howie Johnson said he would make ASCI text of the 5 criteria available on Brian MacLeod's reflector. (BMACLEOD@NETCOM.COM) Howard Frazier, Geoff Thompson and Brian MacLeod will get the ieee.com reflector set up. Brian MacLeod moved and Shimon Muller seconded the following motion. The ad hoc group on Gigabit Ethernet requests permission to form a Study Group under the IEEE 802.3 Working Group. The group believes that sufficient interest in the subject has been demonstrated during this weeks Plenary. The scope of work is to investigate the possibility of extending the 802.3 standard and approved projects to higher speed operation and reporting recommendations to the 802.3 working group. The objective of the Study Group is to complete recommendations by the end of the March, 1996 La Jolla Plenary, however the anticipated time frame for the study group is no more than two Plenary cycles to allow adequate time for study. Brian MacLeod spoke in favor on the motion and pointed out how the motion meets the 802 requirements for study groups. Bernard Daines offered a friendly amendment that "reporting" in paragraph 3, line 3 of the motion be changed to "report". The amendment was accepted by mover and seconder. Thompson stated that paragraph 4 in particular needed to be more specific. He offered a friendly amendment adding ", including PAR material, if appropriate" after "recommendations" in line 2 of paragraph 4. The amendment was accepted by mover and seconder. Howie Johnson questioned the lifetime of the study group. Bernard Daines asked if the wording would be a problem if the study group needed more than two Plenary cycles? Geoff Thompson responded that it would not be a problem. The group has to be reaffirmed each Plenary anyway. John Paine questioned whether "gigabit" in line 1, paragraph 1 was too limiting?. Brian MacLeod responded that "Gigabit" has been used in all of the preparatory work leading up to this meeting. Geoff Thompson observed that no name for the group in proposed in the motion. John Paine offered a friendly amendment that "gigabit ethernet" in line 1, paragraph 1 should be in quotes. The amendment was accepted by the mover and seconder. Larry Birenbaum expressed concern about questions that might be raised by the 802 Executive Committee at the Thursday night meeting. He recommended that material be prepared to respond to those questions. Geoff Thompson read the current 802 rules with respect to study group formation and pointed out that interest was the primary criteria. Bill Bunch asked whether this was a new study group. Geoff Thompson responded that it was a new study group. Bob Campbell observed that the first paragraph does not say why the group is to be formed. He proposed a friendly amendment that replaced paragraph 1 with the following paragraph and reversed the order of paragraphs 2 and 3. The ad hoc group on Gigabit Ethernet requests IEEE 802.3 form a study group to investigate 1 Gigabit applications. The amendment was accepted by the mover and seconder. Brian MacLeod proposed changing "one gigabit applications" to "one gigabit extensions to the 802.3 standard". PJ Singh proposed changing "one gigabit" to "higher speed". The proposed changes were accepted by the mover and seconder. Howard Frazier stated that all present in the room could vote. Howard Frazier read the final text of the motion. The ad hoc group on "Gigabit Ethernet" requests IEEE 802.3 form a study group to investigate higher speed extensions to the 802.3 standard. The scope of work is to investigate the possibility of extending the 802.3 standard and approved projects to higher speed operation and report recommendations to the 802.3 working group. The group believes that sufficient interest in the subject has been demonstrated during this weeks Plenary. The objective of the Study Group is to complete recommendations, including PAR material, if appropriate, by the end of the March, 1996 La Jolla Plenary, however the anticipated time frame for the study group is no more than two Plenary cycles to allow adequate time for study. The motion passed with 69 yes, 0 no and 2 abstain. Howard Frazier announced that there will be an interim meeting. At least one and a half days will be needed. The preferred option for meeting location was the San Francisco Bay area. The meeting was adjourned at 4:30 pm. Bill Quackenbush Secretary du jour