Jan 31 2008

More VMware Server in Production

Published by Brian at 4:08 pm under Clustering, Linux, Tlf, vmware

I happened to run across this fella today, who also runs a very large VMware Server farm in a production environment. He makes a few mentions of his architecture which is slightly different than our approach, however it could prove handy to someone else building out such a thing.

2 Responses to “More VMware Server in Production”

  1. Ben Ruseton 02 Feb 2008 at 6:28 am

    We had originally attempted to put our vmdk’s on a PowerEdge 2900 with 8×500gb SATA drives and a PERC controller, shared out with NFS. We’ve been seeing speeds slower than software raid out of that PERC controller, which meant that once we had ~20 or so VM’s sitting on NFS, the guest OS’s would start seeing SCSI timeouts as the 2900 struggled to get I/O processed fast enough.

    So we’re slapping 750gb drives in each server and going back to local disk, for now. We’re also buying 10 more PE1950’s, bringing us up to a 25 server VMWare Server farm.

  2. Brianon 04 Feb 2008 at 7:22 am

    Ah, that’s where I see the differences. We put our VM’s on a dedicated SAN, with a 1GB cached controller behind an NFS cluster. Probably why were were seeing increased performance.

    There’s a also a decent amount of NFS tweaking required to get it working smoothly. Once I had our network dialed-in, I was able to push 20Mb/s write in a VM, over NFS.

    Check out some of the older articles on the site regarding the setup.

Trackback URI | Comments RSS

Leave a Reply