It was mentioned that Wigner is "23ms away" so latency to remote files would have an effect.
--
----------------------------------------------------------------------
Peter Gronbech GridPP Project Manager Tel No. : 01865 273389
Department of Particle Physics,
University of Oxford,
Keble Road, Oxford OX1 3RH, UK E-mail : [log in to unmask]
----------------------------------------------------------------------
> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of David Colling
> Sent: 24 October 2014 16:49
> To: [log in to unmask]
> Subject: Re: Apologies and a question ...
>
> Hi Jeremy,
>
> > Did you mean specifically in the cloud context, or more generally (the direct)
> use of WNs?
>
> Good point. I believe that CMS is planning to use nearly all CERN
> resources as cloud resources. Specifically I am interested with job
> efficiencies and data location. EOS offers a unified namespace across
> both the Wigner and Meyrin instances so you don't know where your file
> physically resides. Even if there are replicas at both sites you will
> only pick the local copy **most** of the time not all the time. Now with
> a 100Gb/s pipe you might think that this doesn't matter, however for
> certain sorts of workflow and hence data access it matters a great deal.
> For example CMS merge jobs take about 10 times longer if accessing
> geographically remote data. So we are finding ways around these things
> and I was wondering what other experiments were doing and what they were
> finding.
>
>
> Best,
> david
|