LHC Computer Grid - Rollout [mailto:[log in to unmask]] said:
> I thought the idea behind glue-cluster was to allow publishing (indeed
> via YAIM and not by-hand hacks) the actual subclusters as glue
> subclusters.
It does help with that, but as you say you would still need to create more queues. What it does let you do is have all the queues hosted from a single CE node - with the standard yaim functions you can only have one subcluster published by a given node, so splitting subclusters implies multiplying nodes as well as queues.
> So you are telling me that we could use glite-cluster to
> publish the single "subcluster" we have now, only once instead of once
> per CE host that we provide? I guess that use case is useful!
Yes. It makes the configuration easier to manage, it reduces the data volume and it will hopefully help when glue 2 publication arrives in EMI 2.
Stephen
|