Testbed Support for GridPP member institutes
> [mailto:[log in to unmask]] On Behalf Of Rob Fay said:
> Out of curiosity, what if you don't run 1 job per thread?
> E.g. if you limit the
> box to, say, 12 job slots (for reasons of disk or network
> contention perhaps), what do you publish then?
As I said in a mail a few weeks ago, in general I don't think you should
publish resources that nothing can in practice use. This is somewhat
similar to the "deployable in 45 minutes" argument, if you have machines
sitting in a cupboard then they also shouldn't really count as being
part of your capacity, even though they exist. On the other hand, if
it's a partial restriction by VO then you can take account of it via the
VO shares. In the end this is a management decision, and I'm not aware
that the GridPP PMB has expressed an opinion on what it would like to
see - Jeremy would be the one to answer that. LCG only care that the
resources advertised for the LHC VOs correspond to something that those
VOs can use, i.e. LogicalCPUs*HEPSPEC*(sum of LHC VO shares) should
correspond to resources that could be used by those VOs if the system
were full.
Stephen
--
Scanned by iCritical.
|