Thanks for the info! Great data points. We will still recommend a separated solution, but it's good to know that some have tried to unify compute and storage and have had some success. On Sat, Aug 9, 2014 at 5:50 PM, Mika?l Cluseau <mcluseau at isi.nc> wrote: > Hi Zach, > > > On 08/09/2014 11:33 AM, Zach Hill wrote: > >> Generally, we recommend strongly against such a deployment in order to >> ensure performance and failure isolation between the compute and storage >> sides of the system. But, I'm curious if anyone is doing this in practice >> and if they've found reasonable ways to make it work in production. >> > > we are doing this in production since more than 2 years now because we > couldn't afford more than a 2 nodes solution. There's of course a > performance impact due to the pressure on the OS, but it didn't caused any > lock or performance collapse. > > Of course, now that we had some production time behind us, we can afford > scaling up and separating storage and compute concerns and will do, but > only to be clean and allow each concern of the cluster to scale > independently. For instance, I won't hesitate to add drives to compute > hosts, should a massive increase in storage capacity demand come. > -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.ceph.com/pipermail/ceph-users-ceph.com/attachments/20140810/8be3fa92/attachment.htm>
 |