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

[STDS-802-3-BWA] Comments about Why 400G Now ?


I had some comments regarding the last section in the Deck – the Why 400G Now ?  The deck seems to solely rely on two pieces of evidence to conclude 400G is the right choice now.  1) 400G is easier to standardize and build today, and 2) Bandwidth is increasing rapidly.  What does not seem to have effectively been answered is "Why 400G now and not 1T in 2 (N?) years ?"  I believe I heard someone at the mic pose an almost identical question, though I didn't hear a response.

I also heard a lot of conversation about LAG – operators dislike LAG for a variety of reasons.  As an operator I can say that's true, however, as soon as 100G became generally available, we installed it and we introduced LAG and multilink capability (ECMP) to turn multiple 100Gs into bigger Nx100G transport links.  100G didn't allow us to avoid LAG/ECMP and neither will 400G nor will 1T.  To me, the argument for why 400G now should be based on some evidence that for operators (whether datacenter or service provider) scaling 100G with the big pipe tools that exist for us – TRILL, ECMP, LAG, etc – will become ineffective before 1T is generally available.