|Thread Links||Date Links|
|Thread Prev||Thread Next||Thread Index||Date Prev||Date Next||Date Index|
With availability of QSFP-DD800/OSFP and 802.3ck 100G/lane coupled with Ethernet Technology Consortium (ETC) 800G MAC the cost of providing 800 GbE practically speaking is free, but we all know the primary applications are 100/200/400 GbE breakouts.
As long the plumbing exist for 1600 GbE and adding 1600 GbE MAC/PCS/FEC is about free then 51.2 Tb switches may include 1600 GbE MAC where the real usage will be 200/400/800 GbE.
According to Mr. Rob Stone (https://www.ieee802.org/3/B400G/public/21_03/stone_b400g_01_210301.pdf), Facebook will be operating the fabric links at 400 GbE with 51.2 Tb switch generation.
In my opinion we are facing some serious headwinds:
- the pluming for 1600 GbE may end up to be much more complex and costly
- some of the proposed FEC implementations see Mr. Xinyuan Wang (https://www.ieee802.org/3/B400G/public/21_02/wang_b400g_01_210208.pdf ) could make it too costly for an ASIC implementing x256 or x512 lanes to incorporate and 1600 GbE end-end FEC These in effect likely will push out 1600 GbE availability and may not be included on 51.2Tb switches due to cost premium until there exist real 1600 GbE applications.
To unsubscribe from the STDS-802-3-B400G list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-B400G&A=1