If this is necessary, then I think we need quotas implemented in DPM.
Allocating disk pools exclusively to a VO when you have less pools than
VOs isn't a viable approach. Even when you have more pools than VOs it
is too rigid.
I'd also find it useful to know how much spaced each LHC VOs are asking
for at Tier2s (the UK share of it).
We then need to look at how the UK Tier2 resources match up to this and
decide what to allocate in response.
I note that so far I have seen a marked lack of use of the storage at my
site by any VO. I can understand why they might not want to use it. But
until that changes fixed allocations are pointless. I can guarantee my 2
TB to everyone while none of them want to use it.
Cheers,
Simon
Jensen, J (Jens) wrote:
> Folks,
>
> This is just a warning that LCG may require that all Tier 2
> sites can account for space used and available to the larger
> LHC VOs.
>
> "Used" is relatively easy, but "available" may require you
> to allocate dedicated storage resources (pool groups or whatever)
> to at least the major LHC VOs to which you provide resources.
>
> So as an example, if site X is providing 10 TB storage for
> Atlas, then those 10 TB will likely have to be dedicated fully
> to Atlas, so when the SE publishes "8 TB are available" (if
> they have used two), then those 8 TB really are available and
> cannot be used by people from another VO.
>
> Non-LHC VOs can share allocations, LCG doesn't care about them
> (in this context :-) I am thinking that even if site X primarily
> supports Atlas and provides smaller but shared allocations to CMS
> and LHCb, that may be OK.
>
> We shall discuss this at the next storage phone conference.
> The details are still up for debate, and we can as a group provide
> feedback into LCG.
>
> Thanks,
> --jens
>
|