[openib-general] scaling issues, was: uDAPL cma: add support for address and route retries, call disconnect when recving dreq

[Michael S. Tsirkin]

Quoting r. Roland Dreier <rdreier at cisco.com>:
> Subject: Re: [openib-general] scaling issues, was: uDAPL cma: add support for address and route retries, call disconnect when recving dreq
> 
> > > With 256 nodes all to all already gives you 65000 requests
> > > which is the same order of magnitude as the reported 130000.
> 
> I think the only advantage of caching is if you start your app twice.
> 
> But maybe you can fill the cache more efficiently by doing a single
> get table to find all the paths at once, rather than having to let the
> cma query each path after arping for the GID...

Actually, this sounds like an excellent idea - this immediately makes the number
of queries linear with cluster size.

-- 
MST