Hi Lily, I think coordinated congestion control in AI clusters is a good topic for the AIDC/HPC side meeting.
Hesham On Sun, Oct 22, 2023, 9:10 PM Lvyunping (Lily) <lvyunping= [email protected]> wrote: > Hi, > > I’ve posted a new draft, ‘Coordinated congestion management’: > https://datatracker.ietf.org/doc/draft-lyu-rtgwg-coordinated-cm/ > > > > This is mainly for AI fabric which is sensitive to bandwidth. Adaptive > routing is a main approach to improve path utilization and avoid path > congestion caused by asymmetric traffic, while congestion control is > another important mechanism in congestion management to alleviate > congestion. Currently they work independently, that results in unnecessary > throughput reduction. This draft talks about the issue and proposes a > coordinated scheme for AR and CC . It identifies the causes of congestion, > and uses ‘coordination tag’ in data packets to instruct network switches to > perform appropriate congestion management mechanism. > > > > We encourage you to read the draft and provide your feedback and comments. > > > > Thanks! > > Lily > _______________________________________________ > rtgwg mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/rtgwg >
_______________________________________________ rtgwg mailing list [email protected] https://www.ietf.org/mailman/listinfo/rtgwg
