[PLUG] 1u server advice..

Kirk Harr daehlie at gmail.com
Thu Aug 28 01:49:19 UTC 2008


If you are going to run multiple VMs off a single disk system you want to
watch out for I/O contention issues.  I would suggest a RAID 10 setup with
either 4 or 8 disks that will give you redundancy and max concurrent I/O.
RAID 5 is fine, but 10 on a 4 disk set would be ideal. I also suggest
hardware RAID in lieu of software to squeeze every clock cycle out of the
CPUs for the VMs.

Kirk Harr

On Wed, Aug 27, 2008 at 2:11 PM, Rogan Creswick <creswick at gmail.com> wrote:

> On Wed, Aug 27, 2008 at 1:32 PM, Michael Ewan <mhewan1 at comcast.net> wrote:
> > then you are going to have trouble getting 1TB of RAID 5 disk in the
> > box, and please don't install it as RAID 0, you will really regret that
> > decision when the array crashes and you need to restore 1TB from tape
> > (your are planning on backing this up, right).  Hot swap is also good,
> > but doesn't sound like a necessity in your case.
>
> I'm actually not planning on using raid, and for backups, I intend to
> rsync nightly to an off-site server (the amount of disk I'm planning
> to put in there is largely so I can use multiple VMs w/out being
> worried about space).  The vast majority of the content will be
> duplicated, since the VMs will be largely identical, and I'll have
> off-site backups of them in a pristine state, along with incremental
> backups of the configuration details, data, and code running on them.
>
> I also could probably manage perfectly well with about 100 gigs, but
> disk is cheap :)  My current server is a shared penguin computing 1u,
> and we were recently bitten by the raid/lvm configuration -- some
> content was corrupted and mirrored across the raid before we noticed.
> When we *did* notice, the system went down and would no longer boot.
> We would have lost the (software) raid conf, if it weren't for an
> off-site backup (or good memory... I wasn't actually involved in that
> restoration).  In any case, I am not in a situation where that degree
> of high-availability is worth the premium it costs--periodic backups
> are sufficient to manage the times when something goes down.
>
> Re: penguin computing -- I like the company, and the issues we've had
> with our current penguin 1u are probably due to environmental problems
> (poor cooling at the colo).  My only hesitation is that it looks like
> the same hardware can be had from Dell for significantly less (there
> is about a $400 difference between the Altus 650 and a similarly
> configured PowerEdge SC1435 right now).
>
> Since neither company offers an OS I am comfortable with, I'll be
> reinstalling on my own anyway. (And probably buying the disk(s) from
> newegg, to save another couple hundred.)  I would sleep better buying
> from penguin computing though :)
>
> > You will want dual power supplies, both for capacity and for
> > redundancy.  Another reason not to do this yourself, the vendor has
> > already done all the power consumption calculations for ALL the possible
> > configurations.
>
> Very good point...
>
> --Rogan
> _______________________________________________
> PLUG mailing list
> PLUG at lists.pdxlinux.org
> http://lists.pdxlinux.org/mailman/listinfo/plug
>



More information about the PLUG mailing list