1. Higher Speed Ethernet Update Greg Hankins Global Peering Forum 2.0 GPF 2.0 2007 2007/03/29

2. 2 Per IEEE-SA Standards Board Operations Manual, January 2005 At lectures, symposia, seminars, or educational courses, an individual presenting information on IEEE standards shall make it clear that his or her views should be considered the personal views of that individual rather than the formal position, explanation, or interpretation of the IEEE.

3. 3 Higher Speed Ethernet - Technology Pull, Not Push Many different networks are asking for 100 GbE today –IXs –ISPs –Content providers –Financial –R&E and HPCC Already getting media attention and mention by C-level execs Participation by end users has never been this high in the IEEE before –Presentations at meetings given by individuals from: AMS-IX, Comcast, DT, EDS, Equinix, Google, LBNL, NTT America, NYSE, Sprint, Time Warner, T-Systems, Yahoo! –Over 30 individuals contributing and supporting presentations –Thanks for your support

4. 4 Higher Speeds Drive Density (or, Why Should I Care?) 100 GbE will benefit everyone –Even if you don’t need it –Drives a fundamental advance in technology Drives 10 GbE port density up and cost down Possible line-rate combinations –1 x 100 GbE port –10 x 10 GbE ports –100 x 1 GbE ports –And even more oversubscribed port density… Your bandwidth requirements and port densities are growing, not shrinking

5. 5 Call for Interest Study Group Task Force Working Group Ballot Sponsor Ballot Standards Board Approval Publication Feasibility and Research Ideas From Industry Industry Pioneering 1 Year HSSG is Here Ad Hoc Efforts CFI July 18, 2006 Q4 2007? 2009 – 2010? Birth of an IEEE Standard: It Takes About 5 Years IEEE ~4 Years

6. 6 HSSG Focus MDI - Medium Dependant Interface PCS - Physical Coding Sublayer PHY - Physical Layer Device PMA - Physical Medium Attachment PMD - Physical Medium Dependent WIS - WAN Interface Sublayer XGMII - 10 Gigabit Media Independent Interface MAC data rate objective and architectural issues PHY objectives Types Reaches IEEE Std 802.3-2005

7. 7 Next Step: Becoming a Task Force (The Good) Write a Project Authorization Request (PAR) –PAR A working draft created at January meeting –Vote for chair to present PAR A to IEEE 802.3WG at the July Plenary Answer the 5 Criteria for PAR A –Broad Market Potential: √ (59 / 5 / 12) –Compatibility: √ –Distinct Identity: √ –Technical Feasibility: √ –Economic Feasibility: √ HSSG request to extend SG lifetime again –SGs only exist for 6 months –2 nd extension granted after March Plenary

8. 8 PAR A Working Draft Objectives All: All people in the room, 802.3: Registered 802.3 voters Objective Votes (Yes / No / Abstain) Support full-duplex operation onlyAll 73 / 0 / 4 Preserve the 802.3 / Ethernet frame format at the MAC Client service interface All 76 / 0 / 6 Preserve minimum and maximum FrameSize of current 802.3 StdAll 74 / 0 / 4 Support a speed of 100 Gb/s at the MAC/PLS interface All 67 / 9 / 14 802.3 26 / 6 / 11 Support at least 10km on SMF All 86 / 0 / 4 802.3 40 / 0 / 4 Support at least 100 meters on OM3 MMF All 61 / 3 / 27 802.3 33 / 2 / 13 Support a BER better than or equal to 10 -12 at the MAC/PLS service interface All 68 / 0 / 4 Under Consideration Support at least 40km on SMF All 38 / 10 / 32 802.3 12 / 6 / 16

9. 9 More Objectives: Copper? (The Bad?) Several presentations have been given supporting a copper objective –Technical and economic feasibility –5m – 10m reach Some interest in adding copper as an objective –Straw poll: 34 / 15 / 36 –Currently does not meet >= 75% criteria May be under consideration for PAR A or a separate PAR

10. 10 More Objectives: 40 Gb/s? (The Ugly) Growing support for a 40 Gb/s rate for server applications –January Interim: 22 / 33 / 21 –March Interim: 35 / 33 / 20 Positioned as a server interconnect technology –Servers do not need 100 GbE today –Aligns with 16 x PCIe2 bus speed –100m MMF, copper and backplane reaches Supported by individuals from component vendors, Intel, Sun, IBM –Others at Dell, HP and IBM only want 100 GbE –Little end user support May be under consideration for PAR A or a separate PAR

11. 11 Impact of 40 GbE on HSSG and 100 GbE Some delay is anticipated –Time needed to accommodate 40 GbE objective –This also includes figuring out co-existence of the two speeds –Increased scope and change for slippage –Expected that 40 GbE and 100 GbE standards would be available at about the same times Options –Let HSSG voters decide –Compromise and allow market to decide –40 Gb/s MSA outside of IEEE Politics are in play, this is not a logical argument

12. 12 Impact of 40 GbE on System Vendors Board design is a lengthy and expensive process –FPGA: 6 – 12 months –$2.5M –$5M development costs –ASIC: 9 – 18 months (it takes 3 months just to make a chip) –$7.5M – $10M development costs We have to stop work on 100 GbE or work on it in parallel –Everyone has limited resources –Delays 100 GbE –We’d rather put effort into 100 GbE and deliver something faster in about the same time Dual-rate line cards –Costs everyone more –You pay for a 100 GbE line card and run it at a lower speed –Assumes there is some auto-negotiation of speeds Component vendors face similar constraints –Impacts components available to us to build stuff for you

13. 13 Impact of 40 GbE on the Gear 40 GbE switches need an uplink technology –This would likely be 100 GbE Multiple port combinations will be required –10 GbE and 40 GbE –GbE, 10 GbE and 40 GbE –40 GbE and 100 GbE –10 GbE, 40 GbE and 100 GbE –Several possible fiber and copper combinations for each speed

14. 14 Impact of 40 GbE on the Market Will make both technologies more expensive because of volume and market split Confusion in the marketplace about mass adoption –Delay buying to see who wins and what turns out cheaper 40 GbE would need to be –Cheaper than 4 x 10 GbE LAG –Available much sooner than 100 GbE –But data from Intel indicates two years between server needs for 40 GbE and something faster Are people going to deploy 10 GbE, then 40 GbE, then 100 GbE to servers? –10GBaseT standard in 2006 –Just now starting to see lots of 10 GbE NICs on the market –Probably would go straight to 100 GbE instead of spending money on a slower technology

15. 15 What’s Next? April Meeting Agenda 40km 100 GbE reach objective –Technical feasibility? –Economic feasibility? 40 Gb/s MAC rate –Add as an objective? –Economic feasibility? –Broad Market Potential? –What reaches? Copper –Add as an objective? Motions have to be made in April Finalize PARs

16. 16 IEEE 802.3 HSSG Reflector and Web Page To subscribe to the HSSG reflector, send mail to with the following in the body of the message:ListServ@ieee.org subscribe stds-802-3-hssg end (over 460 people have subscribed to the list) HSSG web page has links to all presentations: http://grouper.ieee.org/groups/802/3/hssg/index.html

17. 17 Future HSSG Meetings IEEE 802.3 Interims –http://grouper.ieee.org/groups/802/3/interims/index.htmlhttp://grouper.ieee.org/groups/802/3/interims/index.html –April 17 – 19, 2007 –Ottawa, Canada –May 28 – 31, 2007 –Geneva, Switzerland IEEE 802 Plenary –July 16 – 19, 2007 –San Francisco, CA, USA

18. 18 Thank You

19. Overview of IEEE 802.3 Standards Process (1/5)- Study Group Phase Idea Call for Interest 802.3 Form SG 802 EC Form SG Study Group Meetings 802.3 Approve No Yes 802 EC Approve NesCom Approve SASB Approve Approved PAR Yes No PAR 5 Criteria Objectives Check Point Check Point Check Point Check Point RIP Check Point Note: At "Check Point", either the activity is ended, or there may be various options that would allow reconsideration of the approval.

20. Task Force Meetings Proposals Selected Task Force Review TF Review Done Yes Objectives Approved PAR No D1.0 D1.(n+1) No Yes A D2.0 Overview of IEEE 802.3 Standards Process (2/5) - Task Force Comment Phase To 802.3 WG Ballot

21. 802.3 WG BALLOT Yes No D3.0 No A A Yes D2.(n+1) Yes B A No 802 EC Forward to Sponsor Ballot 802.3 Forward to Sponsor Ballot No TF Resolves Comments Substantive Changes > 75% Yes No Overview of IEEE 802.3 Standards Process (3/5) - Working Group Ballot Phase Notes:At "Check Point", either the activity is ended, or there may be various options that would allow reconsideration of the approval. See 802.3 Operating Rules 7.1.4 and listed references for complete description In Scope New Negatives Check Point

22. LMSC Sponsor BALLOT Yes No B B Yes D3.(n+1) Yes C B No 802 EC Forward to RevCom 802.3 Forward to RevCom No TF Resolves Comments Substantive Changes > 75% Yes No Overview of IEEE 802.3 Standards Process (4/5)- Sponsor Ballot Phase Notes: At "Check Point", either the activity is ended, or there may be various options that would allow reconsideration of the approval. See 802.3 Operating Rules 7.1.5 and listed references for complete description In Scope New Negatives Check Point

23. RevCom Review SASB Approval RevCom Approval Yes No B Yes Standard Check Point C Overview of IEEE 802.3 Standards Process (5/5) - Final Approvals / Standard Release Publication Preparation Approved Draft Notes:At "Check Point", either the activity is ended, or there may be various options that would allow reconsideration of the approval.