Yepp, I'm quite familiar with the work since a good while. M$ evolved their DC since 2008 or so in very interesting way around the ecosystem they bred around Azure (always a _big_ advantage if you can strong-arm your kernel to do things that makes your DC better or even understand what traffic characteristics you run & how you can mash them to shape ;-). Earlier papers (very insightful) are centered around VL2 & Monsoon DC architectures. Newest output I saw from there points however to yet a different data point (I think it's hinted @ in this draft) where the computation is being actually largely moved out to something akin to a PCS and very impressive DC sizes. Peter may leak more once he warms up showing the work ;-)
-- tony > | > | Again, looking fwd' to presentation and discussion on the floor. > > i found the nanog preso on this matter pretty interesting ... > https://www.nanog.org/meetings/nanog55/presentations/Monday/Lapukho > v.pdf _______________________________________________ rtgwg mailing list [email protected] https://www.ietf.org/mailman/listinfo/rtgwg
