 |
 |
hi Winnie, It's not unusual to have that much space used by atlas - look at the software tags you're publishing to see just how many releases of the ATLAS software you have installed! Most of this is for legacy support, I believe - there are still people using surprisingly old releases of the software - although I'm actually not sure that ATLAS cleans up "ancient" releases as quickly as they could. As for other, non-LHC VOs - there's barely any use at all (apart from, as a special case, NGS...), mainly, I suspect, due to there not being enough members to justify the whole sgm installer kaboodle. Software space is volatile, yeah - and the big four VOs can reinstall releases pretty quickly (<< 1 week for the entire ATLAS release set, IIRC). On the other hand, do you want to encourage the VOs to install even more ancient release than they already do? ;) RAID0 might be better than 1 for performance reasons as well, though - despite others' claims to the contrary, we've definitely seen our software area get high load when a lot of jobs start up at once. Sam 2009/10/26 Winnie Lacesso <[log in to unmask]>: > Dear All, > > For new SL5 WN we'll have a new experiment software space. > It would suit us to have it as 500GB. Current SL4 software space > is hugely dominated by atlas: > > 2G alice > 263G atlas > 82G cms > 49G lhcb > Rest minimal. > > Is it normal to have so much atlas software? Is it possible to > restrict atlas software space to a limit so that 500GB will fit all > main LCG VOs into future? > > Re: other VOs: Bristol doesn't support many yet but hopefully will in > future. What kind of experiment software space are they requiring - is > anyone seeing large experiment software space use by non-LCG VOs? > > An option other than 1x500GB (RAID1) is 1TB RAID0 (2 x 500GB). > Could all go kablooey. Software space is volatile anyway, innit? > > Grateful for advice >