Installed capacity for CPU and Storage is described within
WLCG_GlueSchemaUsage-1.8.pdf
A recent presentation on the matter.
gstat is being changed to conform GGUS:47466. We expect some progress
here before the end of the week.
Computing sites have been able to fix their total installed capacity for ever basically
with the current YAIM. In particular installed capcity is the sum of SI2000 * #logicalCPUS.
All tier1s are publishing non-zero values for this now.
For details by region the strict view of gridmap is currently the best tool.
The following publish 0 as a value.
- DE/CH - 1 site SCAI
- Central - 1 site irb.hr
- Northern - 6 sites.
- UK/I - 2 sites, Liverpool, Manchester.
- CERN - 1 site.
- Russia - 4 sites.
- SouthWest - 2 sites.
- AP - 5 sites.
- SouthWest - 7 sites, ticket opened yesterday.
- Italy - Lots of sites. Ticket open. They are working on it clearly.
Once gstat is corrected and starts to develop warnings if for instance a site publishes
0 capacity I will be following up more actively.
Within GGUS there is now a problem type concerning installed capacity.
I am happy to take questions directly particularly at a regional level.
The next YAIM release has now been released to production. This includes the extra
variables as described in the document. In particular the following can now be published.
- The HEPSpec2006 benchmark.
- The VOShare string. For sites using wall/cputime sharing this can and should
be published.
- A reference CPU benchmark can be published. i.e the processor on which you
guarantee the default queue lengths.
YAIM variables are all described in the usual place with other
install
instructions.