Thanks John & Stephen.
I would rather provide external network access from my WNs (in fact I
think I already do) than go to the bother of a VO box. In fact, I had
understood this to be a requirement.
Digging around a bit more I came up with some more concrete points on
slide 12 of this talk:
http://agenda.cern.ch/askArchive.php?base=agenda&categ=a051784&id=a051784s4t1%2Fmoreinfo%2FRequirements-summary.pdf
The only thing I understand from this is the need for local database
replicas on each site.
Taking this example: whilst this would clearly improve the robustness of
production, I do not see the absolute need for one at each Tier2. As long
as the network is up, could it not equally be cited centrally for each
Tier2 or even at the nearest Tier1? I don't see any need for it to be
physically located on my network. It is easier for me to maintain some
firewall holes than a VO box. Performance might drop a bit without it, but
if the threshold for being an LCG site is kept low, guaranteeing the
widest possible participation, surely that would more than make up for it.
For ATLAS (my experiment) I would be willing to maintain a local DB
replica myself as it would benefit local users, but that is beside the
point.
I'm willing to be convinced but I am perplexed that no one seems to want
to do that. I realise that such people may not read tb-support, but could
it be arranged for some of them to talk to us before we try to decide our
position on this?
Cheers,
Simon
On Fri, 16 Sep 2005, Gordon, JC (John) wrote:
> See the Baseline Services Working Group Report in
> http://lcg.web.cern.ch/LCG/peb/BS/
>
> Section 2.8. The text there probably isn't correct though. The best use
> case I have heard is as a proxy to contact central databases. eg the VOs
> database of jobs cannot be updated directly by running jobs as WNs may
> not have external network access. The VObox will provide this link.
>
> One problem is the lack of a coherent description. Since this document
> was written we have seen subsequent requests for remote root access
> (later withdrawn) and for inbound network access.
>
> John
>
> > -----Original Message-----
> > From: Testbed Support for GridPP member institutes
> > [mailto:[log in to unmask]] On Behalf Of Simon George
> > Sent: 16 September 2005 15:33
> > To: [log in to unmask]
> > Subject: Re: VO Boxes - there will be a GridPP position on
> > this discussion
> >
> > Can you clarify what is meant by the first point on p.7? Are
> > you saying that the VOs should back up their box, or I should
> > back up the VO box? I would not be able to offer the latter service.
> >
> > I agree with page 8, but I still feel uniformed about the
> > experiments'
> > point of view.
> >
> > I repeat my request of yesterday: could someone *please*
> > explain or point me to an explanation of what the problem is
> > that the VO box is necessary to solve.
> >
> > Cheers,
> > Simon
> >
>
|