For my twopenny's worth, I would look at the functional aspects (hack it
and see if it works) separately from performance (use the core stuff
which all things support and stress test them.)
My interest in this is different again from the others mentioned, in
interoperation with other service providers. Sooner or later we have
users with data in the EGI cloud and/or a public cloud and will want to
move data to our stuff, or out of it. Probably not LHC(?), but very
likely non-HEP VOs which we'd still try to support. Also as an NGI, we
are increasingly likely to have to support both flavours - or at the
very least have a position on it!
EGI have a CDMI implementation. There is also the reference
implementation from SNIA which I've got on my laptop (doesn't quite
compile yet, I wanted to run it in a IDE but I am not really used to
IDEs more advanced than Emacs ;-)
Incidentally, I am at CNR (Italian research council) this week, in a
meeting at the relevant time, so Wahid, could you chair the discussion
today please? If you could discuss the cloud stuff and the proposed
agenda for GridPP, that'd be great.
Thanks!
--jens
On 26/02/2013 09:21, Wahid Bhimji wrote:
> Ah yes Ewan ,
> Sorry I missed your question. The answer was at the bottom of my original email I think.
> Adam has mentioned the Imperial plans .
> But also RAL have set up S3 atop of Ceph
>
> Also CERN have both a Huwai appliance and an OpenStack one then they are willing to collaborate on..
> But I would start with the UK ones I think
>
> Wahid
>
> On 26 Feb 2013, at 08:18, Adam Huffman <[log in to unmask]> wrote:
>
>> On Mon, Feb 25, 2013 at 5:44 PM, Ewan MacMahon
>> <[log in to unmask]> wrote:
>>
>>> On a practical note, where are you planning to find a 'cloud' storage
>>> system to test on?
>>>
>> I'm hoping that OpenStack Swift will be available for Wahid (and
>> anyone else who's interested) on the embryonic GridPP setup in a few
>> days' time.
>>
>> Adam
>>
>
--
Scanned by iCritical.
|