Hi, What are the implications of this multicast proposal with RDMA transports like infiniband / UE / RoCE? Thanks --- Troy
On Tue, Oct 21, 2025 at 6:41 PM <[email protected]> wrote: > > Hi, > The draft LLM MoE multicast use case proposed at the last IETF side > meeting has been updated based on the feedback received. > Welcome more reviews and comments. > Thank you! > Best regards, > Sandy > > > <http://www.zte.com.cn/> > > > <http://www.zte.com.cn/> > Original > *From: *[email protected] <[email protected]> > *To: *段威10036319;Xiaohu Xu <[email protected]>;张征00007940; > *Date: * 2025年10月20日 15:22 > *Subject: **New Version Notification for > draft-zhang-rtgwg-llmmoe-multicast-01.txt* > > A new version of Internet-Draft draft-zhang-rtgwg-llmmoe-multicast-01.txt has > been successfully submitted by Zheng Zhang and posted to the > IETF repository. > > Name: draft-zhang-rtgwg-llmmoe-multicast > Revision: 01 > Title: Multicast usage in LLM MoE > Date: 2025-10-20 > Group: Individual Submission > Pages: 7 > URL: > https://www.ietf.org/archive/id/draft-zhang-rtgwg-llmmoe-multicast-01.txt > Status: > https://datatracker.ietf.org/doc/draft-zhang-rtgwg-llmmoe-multicast/ > HTML: > https://www.ietf.org/archive/id/draft-zhang-rtgwg-llmmoe-multicast-01.html > HTMLized: > https://datatracker.ietf.org/doc/html/draft-zhang-rtgwg-llmmoe-multicast > Diff: > https://author-tools.ietf.org/iddiff?url2=draft-zhang-rtgwg-llmmoe-multicast-01 > > Abstract: > > Large Language Models (LLMs) have been widely used in recent years. > The Mixture of Experts (MoE) architecture is one of the features of > LLMs that enables efficient inference and cost-effective training. > With the MoE architecture, there are potential multicast use cases > such as tokens dispatching. This draft attempts to analyze these use > cases. > > > > The IETF Secretariat > > > > > _______________________________________________ > rtgwg mailing list -- [email protected] > To unsubscribe send an email to [email protected] >
_______________________________________________ rtgwg mailing list -- [email protected] To unsubscribe send an email to [email protected]
