ATM in its simplest form you need to setup an MCORE queue with dedicated
resources because if you mix
jobs requiring different number of cores you end up with a lot of idle
time. I wouldn't put much resources
behind it but as you kindly made us notice today about IPv6 it would be
better to start staring at it to
understand what it needs to be done. :)
cheers
alessandra
On 27/11/2013 15:59, Ewan MacMahon wrote:
>> -----Original Message-----
>> From: Testbed Support for GridPP member institutes [mailto:TB-
>> [log in to unmask]] On Behalf Of Alessandra Forti
>> Sent: 27 November 2013 14:48
>>
>> while I still have to keep this "exploratory" until the TF is formally
>> approved next week, Atlas has requested this morning multicore resources
>> to process 10M events and this is a good kickstart motivation.
>>
> So,
>
> Let's say one has a normal ATLAS site, with a bunch of assorted
> worker nodes, a couple of cream CEs, and a torque batch server,
> and it doesn't support multi-core jobs - what, in a nutshell,
> would have to be done to it to get it into a state which would
> be useful for this?
>
> Ewan
--
Facts aren't facts if they come from the wrong people. (Paul Krugman)
|