[p4] P4d running under VMWare
gb.perforce at googlemail.com
Mon Aug 4 07:57:37 PDT 2008
> I'm curious as to what you guys have defined for your resources for
> your perforce VMs that handle 500 + users. How much RAM have you
> assigned? And how are your disks setup, ie. are you running the
> database on a VMFS volume, RAW LUN, or NFS, likewise for the depot
I have 2 CPUs (I am not convinced 2 are necessary, based on ESX's
performance graphs, only maybe when doing the weekly p4 verify's MD5
computation). 12GB of RAM, which is still more than the size of the
db.rev table and we don't have 600 users on the system at the same
time. Anyway we are not swapping yet, but we are watching this closely
and will probably up this soon as we grow.
Storage-wise, a NetApp aggregate used to be presented to the ESX host
over iSCSI, on which 3 VMDKs were created (one for the journal, one
for the DB, one for the RCS store). Recently, we have switched from
iSCSI to NFS for ESX's storage.
Our next move will be to have the VM handle the storage (rather than
the ESX host), using snapdrive.
> Guillaume, I've read recently that version 3.5 of ESX now handles
> multiple CPU VMs a lot better since they've updated their algorithms,
> and no longer need to wait for all cores to free up to give the VM the
> resources it needs. So I was just wondering which version of ESX that
> you were running?
Indeed, we were running 3.0.2 at the time. We are on 3.5 now, but
anyway I don't think Perforce needs more than 2 (or even 1) CPU in our
More information about the perforce-user