just need to workout how to build large fat-trees with ethernet. With 100G we get by with mlag for the spines (upto 640 hosts on 10g).
with our 40G/10G network, we are using ospf/mpls/vpls to give a static fat-tree... 30k lines of config applied to ~20 switches. Nasty. On Thu, Jun 2, 2016 at 11:57 AM, Greg Lindahl <lind...@pbm.com> wrote: > On Thu, Jun 02, 2016 at 07:48:10AM +0800, Stu Midgley wrote: > > That's it. As I said, I haven't used it since the early 00's. > > > > With 100Gb becoming common it might be time for these sort of MPI's to > come > > back. > > You could do better than these old interfaces, too... modern ethernet > chips have multiple send/receive queues, so you could hack up an > interface which looks a lot more like InfiniPath's one-queue-per-core > model -- the OpenIB code for InfiniPath also has code that maps M > cores:N queues if M>N. > > -- greg > > -- Dr Stuart Midgley sdm...@sdm900.com
_______________________________________________ Beowulf mailing list, Beowulf@beowulf.org sponsored by Penguin Computing To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
