On Thu, 2008-06-12 at 15:16 +0300, Sasha Khapyorsky wrote: > On 04:50 Thu 12 Jun , Hal Rosenstock wrote: > > On Thu, 2008-06-12 at 14:46 +0300, Sasha Khapyorsky wrote: > > > On 04:45 Thu 12 Jun , Hal Rosenstock wrote: > > > > > > > > > > LMC > 0 is mainly used by MPI, all other protocols / > > > > > applications will use only one LID (PATH), > > > > > > > > That's today; what happens when more ULPs are able to take advantage of > > > > alternate LIDs ? > > > > > > Will be even better since those LIDs will be better balanced now. > > > > Right but why is OK to disrupt non base LID traffic but not base LID > > traffic ? Is it only because it's only being used by one ULP > > currrently ? > > How this disrupt non base LID traffic more than it was now?
I didn't say that. I was asking why it's OK for non base LID traffic but not base LID traffic. Isn't it that's it's many ULPs rather than just MPI ? > Potentially it could be affected by only base LIDs balancing it is against > all LIDs > balancing today. I can't parse that sentence. -- Hal > Sasha _______________________________________________ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
