Hi,
The Tier 1 is constrained by the allocations made by the GridPP User Board, therfore we have no sharing of pools so we can show that we have allocated the space to a vo/experiment and if they're not using it its the vo/experiment's problem :-).
There does appear to be a similar document for Tier 2s - http://www.gridpp.ac.uk/tier2/Tier-2_Experiment_Shares_v1.1.doc
But I have no idea how binding it is.
One possible problem with a free-for-all approach is that if one pool system has a castrophic failure then all users are potentially affected rather than a subset.
Derek
> -----Original Message-----
> From: GRIDPP2: Deployment and support of SRM and local storage
> management [mailto:[log in to unmask]]On Behalf Of Greig A
> Cowan
> Sent: 26 October 2005 15:18
> To: [log in to unmask]
> Subject: Re: Mapping pools to VOs
>
>
> Hi Matt,
>
> I've tried to get some information from all the sites a while
> ago, without
> any success.
>
> http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0507&L=gridpp
> -storage&T=0&O=A&P=8830
>
> We currently have just the default assignment, where any VO
> can go into
> any pool, but I will probably change this when I reinstall
> the next dCache
> release. I think it really depends on what VOs you support. You'll
> probably find that the LHC experiments will use more space
> than someone
> like biomed, so it might be a 'waste' giving them a few TBs
> when no-one
> else can use it.
>
> I guess in the true Grid model you shouldn't really assign VOs to
> particular pools. Instead, you should just let them get on with using
> what storage you have and if it fills up then they can go
> elsewhere. I
> don't know what peoples thoughts are about this. I presume
> that there will
> be local site policies regarding these sorts of issues.
>
> Cheers,
> Greig
>
>
>
> On Wed, 26 Oct 2005, Matt Doidge wrote:
>
> > Hello,
> > Seeing as we're on a fresh install of dcache here at lancaster I'm
> > having to go through the pool-VO mapping process again. Seeing as we
> > have twice as many pools as we did last time I decided to look over
> > our "pool assignment policy" (it's a simple text file) and
> update it.
> > After discussion with a collegue we're wondering: Is VO-pool mapping
> > an absolute nessicity, or could we just keep everything as `default'
> > and hope no VO starts to get greedy with the disk space.
> >
> > Our general site policy otherwise is to ensure that each VO
> has one or
> > more `reserved' pools for its own use, then we have a large
> amount of
> > `default' space availiable to all. So no matter how greedy someone
> > gets with the disk space, a VO always has at least a couple of TB to
> > call their own.
> >
> > What do people think? Is the free-for-all idea just plain folly? And
> > what are everyone else's policies?
> >
> > cheers,
> > matt
> >
> > ps my apolgies if this topic has been bought up before, I had the
> > nagging feeling we've already has this debate but can't find the
> > thread, so I thought it worth bringing up again
> >
>
> --
> ==============================================================
> ==========
> Dr Greig A Cowan
http://www.ph.ed.ac.uk/~gcowan1
School of Physics, University of Edinburgh, James Clerk Maxwell Building
TIER-2 STORAGE SUPPORT PAGES: http://wiki.gridpp.ac.uk/wiki/Grid_Storage
========================================================================
|