LHC Computer Grid - Rollout
> [mailto:[log in to unmask]] On Behalf Of
> Maarten Litmaath, CERN said:
> > Do I understand correctly that this has been corrected for
> the newer SEs?
>
> Yes to all. An SRM SE is self-sufficient, as it should.
With the new schema the classic SEs are also self-sufficient, and
apparently lcg-* is now fixed:
http://savannah.cern.ch/bugs/?func=detailitem&item_id=12222
Will that version be in 2.7?
> Another reason is that direct
> access to file systems is in contradiction with the SE being
> in charge of the file system management.
I think read-only access could make sense, but it seems that no-one
wants it.
> Not really: it still makes sense for a site to indicate a set
> of preferred SEs,
> as far as the network connection is concerned. If the site
> does not have a
> reasonable SE physically nearby, it can refer to someone else's SE.
> Whether applications care about close SEs is yet another matter.
In some cases I think they care too much, with fast networks you could
often use remote SEs quite easily, but people still usually seem to
assume they have to use the local one. (Or else they do everything to
Castor!) In EDG the close SE was supposed to be temporary and we were
expecting to get a dynamic measure of the "cost" of accessing any SE,
but like many things it never arrived.
Stephen
|