10G DAC solutions are not standardized by 802.3, do not use auto-negotiation, and as far as I know don't work with clause 74 FEC (SFP+ ports also support 10G
optics which is incompatible with FEC).
However, clause 74 FEC can be used with 40GBASE-CR4 cables, at 10G/lane, so it has the same latency discussed as the scenario Ali points to.
(As far as I know, enabling clause 74 FEC is not very popular among 10GBASE-KR users. Most links operate without it. I don't have data for 40GBASE-CR4 but I
would assume the same.)
I think the numbers Ali mentioned are somewhat excessive… see
ran_081214_25GE_adhoc which compares FEC options including latencies. A short summary:
-
The minimum time required for clause 74 FEC error correction at 10G/lane,
without marking uncorrectable errors, is 205 ns – one "codeword time".
-
Since the clause 74 codeword is x2.5 shorter than the RS-FEC codeword, this is the same codeword time as in single lane clause
91 RS-FEC at 25G rate.
-
RS-FEC error correction requires more than a full codeword time. this is implementation dependent, but a common estimate is
an additional 50 ns, so a total of 250 ns for a single-lane 25G.
-
RS-FEC with
correction bypassed (only error checking and marking, assuming the raw BER is sufficiently low) would require 205 ns for a single-lane 25G.
</Adee>
From: Kapil Shrikhande [mailto:Kapil_Shrikhande@xxxxxxxx]
Sent: Saturday, August 30, 2014 7:34 AM
To: STDS-802-3-25G@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-3-25G] Announcing 25Gb/s Ethernet Architecture Ad-hoc conference calls
Ali,
That’s only if you assume 10G DAC cabled solutions are using KR FEC…
But interesting data point if 10G KR FEC has similar latency to single lane RS-FEC.
Kapil
From: Ali Ghiasi [mailto:aghiasi@xxxxxxxxx]
Sent: Thursday, August 28, 2014 12:10 PM
To: Shrikhande, Kapil
Cc: STDS-802-3-25G@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-3-25G] Announcing 25Gb/s Ethernet Architecture Ad-hoc conference calls
Kapil
Actually the latency for the 10G-KR FEC is also around 350-400 ns, so when you upgrade from 10G to 25G the latency will be the same just 2.5x the BW.
On Aug 28, 2014, at 11:35 AM, <Kapil_Shrikhande@xxxxxxxx> <Kapil_Shrikhande@xxxxxxxx> wrote:
Ali,
You bring up a good point … that is a significant addition to the latency for a few meters of cable.
Also, a change in latency for users upgrading from 10G to 25G for server interconnect.
Kapil
From: Ali
Ghiasi [mailto:aghiasi@xxxxxxxxx]
Sent: Thursday, August 28, 2014 9:16 AM
To: STDS-802-3-25G@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-3-25G] Announcing 25Gb/s Ethernet Architecture Ad-hoc conference calls
Rich
In 32 FC the same FEC is used with the gain but latency did increased to 4x to about 370 ns.
My suggestion is to use the same FEC unless absolutely 370 ns latency is not acceptable.
<image001.png>
On Aug 28, 2014, at 8:45 AM, Mellitz, Richard <richard.mellitz@xxxxxxxxx> wrote:
Hi Chris,
Going from 4 lanes to 1 lane mandates a changing FEC in some way. The consequences of change FEC may change the channel budget allocation. We should leave
those discussions to the WG. To me, at least three meters allows the discussions to be relevant. With five meters the choices are limited and debates mooted. So I would support the objective with at least 3 meter for now. This does not abdicate 5 meters
but in my opinion the debate should not happen until we go to WG.
… Rich
From: Christopher
T. Diminico [mailto:00000025925d7602-dmarc-request@xxxxxxxx]
Sent: Thursday, August 28, 2014 1:21 AM
To: STDS-802-3-25G@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-3-25G] Announcing 25Gb/s Ethernet Architecture Ad-hoc conference calls
Colleagues,
My ad hoc presentation(s) to date are primarily focused on supporting the technical feasibility of (25G) simply arguing that if 802.3bj 4x25G is technically feasible then
so is 1x25G.
I've changed my presentation from "reuse 802.3bj" to "consistent with"; the original language was not intended to preclude considering RS-FEC options in Task Force, just
to exit Study Group demonstrating feasibility. I'm agreeing with others suggesting we are debating details better left for Task Force and not necessarily helping us exit Study Group.
I believe we already have consensus to consider RS-FEC options in Task Force. Given that, we still need to reach consensus on the objectives. I suggest we move
to a straw poll during the ad hoc tomorrow on the reach/channel obective language candidates to help us close the debate.
First question we might consider is;
A>>objective for reach (i.e., 802.3bj twin-axial with lengths up to at least 5m)
B>>objective for channel
Given the results of that question then choose from options under reach or channel (below) others are possible.
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
802.3 25G objective - options based on 802.3bj objective with reach
+++++++++++++++++++++++++++++++++++++++++++++++++++
A>>Define a 1-lane 25 Gb/s PHY for operation over links consistent with copper twin-axial cables with lengths up to at least 5m.
comment>>>does not preclude 3m without FEC
B>>Define a 1-lane 25 Gb/s PHY for operation over links consistent with copper twin-axial cables with lengths up to at least 3m.
comment>>>does not preclude 5m with FEC
+++++++++++++++++++++++++++++++++++++++++++++++++++
802.3 25G objective - options based on 802.3bj channel usage
++++++++++++++++++++++++++++++++++++++++++++
A>>Define a single-lane 25Gb/s PHY for operation over copper twin-axial cables consistent with channels (TP0-TP5) specified in IEEEStd802.3bj-2014
Clause92
B>>Define a single-lane 25Gb/s PHY for operation over copper twin-axial cables based on channels (TP0-TP5) specified in IEEEStd802.3bj-2014 Clause92
C>>Define a single-lane 25 Gb/s PHY for operation over copper twin-axial cables consistent with the overall channel budget specified in IEEE Std 802.3bj-2014 Clause 92
Regards, Chris
-----Original Message-----
From: Matt Brown (APM) <mbrown@xxxxxxx>
To: STDS-802-3-25G <STDS-802-3-25G@xxxxxxxxxxxxxxxxx>
Sent: Fri, Aug 22, 2014 10:06 pm
Subject: Re: [STDS-802-3-25G] Announcing 25Gb/s Ethernet Architecture Ad-hoc conference calls
*** correction ***
The next architecture ad hoc meeting has been re-scheduled from Tuesday Aug 26 to Thursday Aug 28 9AM-10:30AM PDT. Meeting details are posted here:
From: Matt
Brown (APM) [mailto:mbrown@xxxxxxx]
Sent: Friday, August 22, 2014 6:49 PM
To: 'STDS-802-3-25G@xxxxxxxxxxxxxxxxx'
Subject: RE: Announcing 25Gb/s Ethernet Architecture Ad-hoc conference calls
The next architecture ad hoc meeting has been re-scheduled from Tuesday Aug 26 to Thursday Aug 29 9AM-10:30AM PDT. Meeting details are posted here:
Please send requests for presentation time to me on or before Tuesday August 26.
Minutes from the August 19 meeting are posted here:
Appoved minutes from the August 12 meeting have been updated in response to a request at the August 19 meeting and posted here:.
From: Matt
Brown (APM) [mailto:mbrown@xxxxxxx]
Sent: Monday, August 18, 2014 10:27 PM
To: STDS-802-3-25G@xxxxxxxxxxxxxxxxx
Subject: RE: Announcing 25Gb/s Ethernet Architecture Ad-hoc conference calls
Minutes from the August 12 meeting are posted here:
From: Matt
Brown (APM) [mailto:mbrown@xxxxxxx]
Sent: Monday, August 18, 2014 9:49 PM
To: STDS-802-3-25G@xxxxxxxxxxxxxxxxx
Subject: RE: Announcing 25Gb/s Ethernet Architecture Ad-hoc conference calls
This is a reminder that the second 25 Gb/s Ethernet Architecture ad hoc is scheduled for tomorrow Tuesday Aug 19 at 9AM PDT.
Meeting access information may be found on the 25 Gb/s Ethernet Study Group ad hoc web page:
The draft agenda for the meeting is posted here:
Matt Brown
AppliedMicro
From: Matt
Brown (APM) [mailto:mbrown@xxxxxxx]
Sent: Thursday, August 07, 2014 9:27 PM
To: 'STDS-802-3-25G@xxxxxxxxxxxxxxxxx'
Subject: RE: Announcing 25Gb/s Ethernet Architecture Ad-hoc conference calls
This is a reminder that the first 25 Gb/s Ethernet Architecture ad hoc is scheduled for Tuesday Aug 12 at 9AM PDT. Meeting access information may be found on the 25 Gb/s Ethernet
Study Group ad hoc web page:
As indicated in the announcement, requests for presentation time are appreciated by the end of day tomorrow (Friday Aug 8) so that we can plan the agenda for the ad hoc meeting.
Note that this is a study group ad hoc. The purpose is to provide appropriate materials supporting and building consensus toward the 25 Gb/s Ethernet CSD (5 criteria), objectives,
and PAR.
Please consider the following guidelines for ad hoc presentations and discussions:
Contribute towards support of and/or design of the objectives, PAR, and CSD.
Lead towards achieving and demonstrating consensus in the objectives/PAR/CSD in September.
Point out one or more options and discuss pros, cons, and relevance to the objectives/PAR/CSD.
Presentations including multiple authors, contributors, and supporters are highly encouraged.
Avoid promoting or picking a particular solution. This is for task force.
Matt Brown
AppliedMicro
From: Matt
Brown (APM) [mailto:mbrown@xxxxxxx]
Sent: Monday, July 28, 2014 4:35 PM
To: 'STDS-802-3-25G@xxxxxxxxxxxxxxxxx'
Subject: Announcing 25Gb/s Ethernet Architecture Ad-hoc conference calls
This in an announcement of the 25GE Architecture ad hoc conference calls.
“The purpose of the ad-hoc will be to discuss the different areas of work (RS/PCS/FEC/PMDs/AN/EEE) that will fall under the work of the group and make sure we are prepared
with content and contributions to be able to make progress in September’s meeting.”
The first conference call is scheduled for Tuesday August 12 @ noon EDT (9AM PDT) for two hours. Subsequent meetings will occur at the same time/day each week up to the September
interim meeting.
If you are interested in presenting, please request a timeslot by Friday preceding each meeting.
Agenda and webex information will be provided prior to each meeting.
This is an official study group ad hoc meeting. As such, please familiarize yourself with the meeting guidelines and patent policy before the meeting.
Patent policy: www.ieee802.org/3/patent.html
Matt Brown
AppliedMicro
From: Mark
Nowell (mnowell) [mailto:mnowell@xxxxxxxxx]
Sent: Monday, July 28, 2014 12:11 PM
To: STDS-802-3-25G@xxxxxxxxxxxxxxxxx
Cc: 802.3 Dialog; Matt Brown
Subject: Announcing 25Gb/s Ethernet Architecture Ad-hoc
Hi,
I’d like to announce the formation of an Architecture ad-hoc to support the 25Gb/s Ethernet study group. Matt Brown from APM has graciously offered to lead this effort and
he will be sending out meeting announcements shortly.
We will be using the 25Gb/s Study group reflector for all communication so if you have not already joined please do so - instructions are on the 25Gb/s Study Group page http://www.ieee802.org/3/25GSG/index.html
[This email has been copied to the dialog reflector this time in case people haven’t had a chance to join the 25GE reflector yet]
The purpose of the ad-hoc will be to discuss the different areas of work (RS/PCS/FEC/PMDs/AN/EEE) that will fall under the work of the group and make sure we are prepared with
content and contributions to be able to make progress in September’s meeting.
I look forward to the discussions at the meetings.
Regards,
Mark