> -----Original Message-----
> From: LHC Computer Grid - Rollout
> [mailto:[log in to unmask]] On Behalf Of Burke,
> S (Stephen)
> Sent: 22 September 2005 11:50
> To: [log in to unmask]
> Subject: Re: [LCG-ROLLOUT] RB output sandbox size limit
>
> LHC Computer Grid - Rollout
> > [mailto:[log in to unmask]] On Behalf Of
> Andrea Sciaba'
> said:
> > What would happen if the sandboxes smaller than 100 GB and
> > newer than 3
> > weeks were already more than enough to fill the RB disk?
>
> You need to buy a bigger disk? ;)
In reality, computer centres do not buy bespoke systems for jobs like an
RB - we have either:
A) batch worker type machines with maybe a 200GB single disk (ours are
120GB or lower, depending on generation),
B) high capacity disk servers with 1TB+ disk arrays.
Typically, if a system is needed for an RB or CE or VO box (spit) you
get a re-tasked batch worker box. In some cases (very few) we harden
the system against certain failure modes (extra raid1 disks etc) but in
general, all these services are running on batch worker class boxes with
no hotswap disk or PSU. And post-procuremnt upgrading of systems has
ramifications for maintenance etc.
It's unrealistic to expect anything else for a service unless it was
specified years ago as a requirement.
Martin
RAL Tier1 Systems Team.
|