Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [STDS-802-3-400G] [802.3_400G] Presentation for next week



Ian

 

That’s the exactly the point. Speeds like 800G and 1.6T allow 2x 400G or 4x 400G solutions, respectively.

 

1T requires 2.5x 400G which fits my definition of contortion.

 

Chris

 

From: Dedic, Ian [mailto:Ian.Dedic@xxxxxxxxxxxxxx]
Sent: Thursday, July 17, 2014 6:11 AM
To: STDS-802-3-400G@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-3-400G] [802.3_400G] Presentation for next week

 

Is there really much point spending all the time and effort needed to define a new 800G standard which is only a 2x speed increment? Why not just use 2 400G interfaces?

Just because we *can* define another standard doesn't mean we *should*...

Ian
----------------------
Ian Dedic
Chief Engineer
Fujitsu Semiconductor Europe
3 Concorde Park, Concorde Road, Maidenhead, Berkshire SL6 4FJ, UK
Email : Ian.Dedic@xxxxxxxxxxxxxx
Tel : +44(1628)50-4711
www : http://emea.fujitsu.com/semiconductor

 

From: Fred Rabouw (frabouw) [mailto:frabouw@xxxxxxxxx]
Sent: Thursday, July 17, 2014 04:30 AM
To: STDS-802-3-400G@xxxxxxxxxxxxxxxxx <STDS-802-3-400G@xxxxxxxxxxxxxxxxx>
Subject: Re: [STDS-802-3-400G] [802.3_400G] Presentation for next week
 

I presume 1.6T and 800G for switches and 800G for servers, for 800G on servers and no 800G on switches would be challenging to connect….

 

Regards

Fred

 

 

 

From: Chris Cole
Sent: Monday, July 14, 2014 2:25 PM
To: 'Joel Goergen (jgoergen)'; STDS-802-3-400G@xxxxxxxxxxxxxxxxx
Subject: RE: [STDS-802-3-400G] [802.3_400G] Presentation for next week

 

Joel,

 

1)      Thank you for giving me the benefit of the doubt that I wasn’t advocating something silly. It is critical to have roadmaps, project future technology, and put up straw men. This makes discussion specific and substantive rather than generic. While doing that, it is important to keep in mind that we rarely get it right when making projections so should maintain a healthy dose of skepticism about how we use it.

To make our exchange specific, my concern about adopting a 100G Serial architecture now is exactly the same as my concern about adopting 40G Serial in .ba for 40GBASE duplex SMF PMD. It’s too soon.

http://www.ieee802.org/3/ba/public/sep08/cole_02_0908.pdf#page=5

Agreeing that 40G Serial is the right long-term high-volume solution doesn’t mean that it is right when adopted too early. 40G Serial was 15 years too early as 40G OC768 Telecom client, and it would have been 8 years too early if we adopted it as 40Gb/s Ethernet in .ba. 100G Serial will be the right answer >2020. It’s too soon to standardize it now and lock in today’s understanding and technology as the right long term solution.

40G Serial offers another example why we should be cautious. In .ba we thought the ultimate 40G Serial architecture is NRZ. However, we are now considering PAM-4 as a possible alternative, which at the time wasn’t even on anyone’s radar screen as a PMD alternative.  

Dave Chalupsky wonderfully qualifies his projections:  “Current Forecast … surely accurate J “. Great attitude to have.

2)      My view of the right next rate steps after 400Gb/s Ethernet are 1.6T for switches and 800G for servers, but not 1T.

 

Chris