 |
 |
> -----Original Message----- > From: Testbed Support for GridPP member institutes [mailto:TB- > [log in to unmask]] On Behalf Of Stuart Purdie > Sent: 03 April 2012 11:39 > To: [log in to unmask] > Subject: Memory on Linux / Atlas memory survey. > > There's a number of different types of memory that we can discuss. > It might also be worth backing off a little from the undoubtedly interesting discussion about memory accounting, and asking what it is that ATLAS are actually trying to achieve. If what we're really talking about here is some jobs that usually use less than 2GB of memory, but occasionally and briefly expand to about 3.5GB, and we want to know whether sites can run them or not, then we've got something concrete to think about. Running jobs like that certainly doesn't need swap space equal to twice RAM, and in practice, doesn't even need swap space equal to RAM. Indeed, this sort of usage pattern seems to me to be more or less what we already see, and on a mixed worker node individual jobs going over 2GB for a bit usually just fits in memory that's unused by other jobs on the node. So; is that what lies behind all this, or is there something else (like a workload that would use 2Gb of memory and 2GB of swap as a steady state)? Ewan