Paul, I’m glad you capture my point. Yes, CFP4 is longer. For a slot restricted to copper and VCSEL optics, the heat sink might be lessened in some manner since higher power modules would not need to be cooled that CFP4 form factor in general is slated to support. In regards to upgrade, I’m a little confused. A dual rate slot could be devised for QSFP, but that is a new implementation of the swtich. Re-use of existing 40GBASE-SR4 QSFP modules or need for interoperation with a legacy box at the other end will prompt in part then the use of QSFP for the slot type. I do not know what if any support for 40GBASE-SR4 might occur in CFP4 form factor. If supported, then interopation with a legacy box would be supported by a dual rate CFP4 slot. Jeff From: Kolesar, Paul [mailto:PKOLESAR@xxxxxxxxxxxxx]
Sent: Friday, January 13, 2012 6:56 PM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] - MMF Ad Hoc - 17th Jan meeting timing Jeff, I did not appreciate the near-parity faceplate density between the two forms factors. Thanks for pointing it out. Since CFP4 handles considerably more power dissipation, I suppose it must be larger in the depth dimension then. What is your opinion of the advantage of the QSFP to enable an upgrade path from 40G-SR4 to 100G-SR4 from the same switch port? Is that path significantly closer to reality via the QSFP, or will it not happen regardless of form factor commonality? Paul
From: Jeffery Maki [mailto:jmaki@xxxxxxxxxxx]
Sent: Friday, January 13, 2012 8:41 PM
To: Kolesar, Paul; STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: RE: [802.3_100GNGOPTX] - MMF Ad Hoc - 17th Jan meeting timing Paul, I agree with your statements. I wanted to point out, though, that integer count of interfaces is what matters when the form factors are similar in size. It will take high port count to leverage the QSFP advantage such as 16 CFP4 versus 18 QSFP. From: Kolesar, Paul [mailto:PKOLESAR@xxxxxxxxxxxxx]
Sent: Friday, January 13, 2012 6:17 PM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] - MMF Ad Hoc - 17th Jan meeting timing Jeff, While I agree that universal port slots are very attractive, density may be the overriding factor for server access-optimized switches, where singe-mode optics would only be needed for aggregated uplinks. If thermal management is the limiting issue, as Chris points out in another reply to this thread, then density must be sacrificed. My main point in drawing parallels between copper and fiber is for purposes of commonality with FEC. Given copper has the lead, I am trying to determine what impact following that lead has on fiber capability. Paul
From: Jeffery Maki [mailto:jmaki@xxxxxxxxxxx]
Sent: Friday, January 13, 2012 7:53 PM
To: Kolesar, Paul; STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: RE: [802.3_100GNGOPTX] - MMF Ad Hoc - 17th Jan meeting timing Paul, This seems like a good chance to reiterate why CXP is not most useful to the industry. It is good for copper, just ok for parallel VCSEL optics, and a dead end for anything more. QSFP will be good for copper and good for parallel VCSEL optics, but a dead end for certain longer reach optics. Initially, it will be a dead end for 100GBASE-LR4 but most are thinking implementation will eventually reach the QSFP form factor. How long is the question. My point is that CFP4 is a form factor that achieves almost the same density as QSFP, and may achieve the same integer number of ports for system companies on their line cards, while at the same time being far more enabling. It will be interesting to see how system companies prioritize things. Regardless, QSFP will have to interoperate with CFP4 when both ends are built according to the same standard. Jeff From: Kolesar, Paul [mailto:PKOLESAR@xxxxxxxxxxxxx]
Sent: Friday, January 13, 2012 5:44 PM
To: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] - MMF Ad Hoc - 17th Jan meeting timing If copper fits in a QSFP+ then of course it will also fit into larger forms, and it is not restricted. But for access switch ports, which is where copper would be deployed, then higher density is a big driver due to blade server density trends.
From: Chris Cole [mailto:chris.cole@xxxxxxxxxxx]
Sent: Friday, January 13, 2012 7:18 PM
To: Kolesar, Paul
Cc: STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GNGOPTX] - MMF Ad Hoc - 17th Jan meeting timing What is the basis for your statement that 100G copper is restricted to QSFP+? John, Thanks so much for this exploration, which addresses just the sort of questions that have been puzzling me. In fact I was wishing for a good comparison to SR10 just today. I see some good parallels to Ali’s work emerging. While it is not clear to me how much support exists for FEC behind optical interfaces, it would be quite helpful if your material could address how the addition or substitution of FEC affects the numbers. If the outcome is that FEC allows QSFP+ form factor, then that seems to enable common forms between fiber and copper media, which to me look like a big advantage to the user and vendor, if not for near-term aggregation switch ports at least for future server and access switch ports. I also request the same clarification as Jeff. If you need assistance with cabling cost particulars, I can help. I have been analyzing cabling costs in my efforts to add such analysis to the solution analyzer spreadsheet. A reasonable assumption for the cost of SR4 cabling is ~40% of SR10 cabling cost. Regards, Paul
From: Jeffery Maki [mailto:jmaki@xxxxxxxxxxx]
Sent: Friday, January 13, 2012 6:12 PM
To: John Petrilla; Jonathan King; STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx; a_flatman@xxxxxxxxxxxxx; Amezcua, A. (Adrian); Anslow, Peter; Anthony Torza; Bernstein, Gary; Brad Booth; Daniel Dove; Ephrem Wu; Gary Nicholl (gnicholl); Harry Fu; Jack Jewell; Keith Nellis; Kolesar, Paul; Lian Zhao; Martin Gilpatric; mike.dudek@xxxxxxxxxx; mnowell@xxxxxxxxx; Oren Sela; Piers Dawe; Rick.Pimpinella@xxxxxxxxxxx; Robert Coenen; ryan.latchman@xxxxxxxxxxxxx; Scott Kipp; Shmuel Levy; Sudeep Bhoja; Vipul Bhatt
Subject: RE: [802.3_100GNGOPTX] - MMF Ad Hoc - 17th Jan meeting timing John, Great effort. On your slide 2, can you make clear whether your conclusion includes the MMF cabling cost? Jeff From: John Petrilla [mailto:john.petrilla@xxxxxxxxxxxxx]
Sent: Friday, January 13, 2012 3:03 PM
To: Jonathan King; STDS-802-3-100GNGOPTX@xxxxxxxxxxxxxxxxx; a_flatman@xxxxxxxxxxxxx; Amezcua, A. (Adrian); Anslow, Peter; Anthony Torza; Bernstein, Gary; Brad Booth; Daniel Dove; Ephrem Wu; Gary Nicholl (gnicholl); Harry Fu; Jack Jewell; Jeffery Maki; Keith Nellis; Kolesar, Paul; Lian Zhao; Martin Gilpatric; mike.dudek@xxxxxxxxxx; mnowell@xxxxxxxxx; Oren Sela; Piers Dawe; Rick.Pimpinella@xxxxxxxxxxx; Robert Coenen; ryan.latchman@xxxxxxxxxxxxx; Scott Kipp; Shmuel Levy; Sudeep Bhoja; Vipul Bhatt
Subject: RE: [802.3_100GNGOPTX] - MMF Ad Hoc - 17th Jan meeting timing Hello Jonathan Attached is a draft presentation prepared for the Newport Beach meeting. Although I’m not sure of my schedule/availability for Tuesday morning, I would appreciate feedback from any who care to offer some. Have MMF ad hoc conference contact details for Tuesday been distributed? Regards, John Dear all A reminder that there will be an MMF ad hoc meeting on Tuesday 17th Jan 2012, from 8am to 10.30 am PCT, in order to minimize the overlap with a concurrent OIF meeting. If you have material to present, please let me know before hand. Thanks Jonathan +1 408 368 3071
|