Steve,
On Thu, Mar 19, 2009 at 6:50 AM, Steve Friedman <sfriedman at woti.com>
wrote:>
> I'm in the process of selecting a storage cluster and, after reviewing
the
> gluster sites, have the following questions and greatly appreciate this
> list's thoughts:
>
> 1) The technical FAQ lists 12 x 500 GB SATA II hard drives (I presume that
> the answer hasn't been updated to 1TB drives). ?My question is: what
are the
> recommendations on the server RAM and CPU? ?Is the gluster s/w sufficiently
> multi-threaded and/or CPU-heavy that a quad core (or dual - quad core) is
> required? ?Likewise, I presume that the RAM required is based merely on the
> underlying OS file caching. ?I will be ordering new hardware for this
> cluster -- any experience not only with the current load on the servers
(and
> clients) but what can I expect in the future (e.g., the roadmap calls out
> compression).
Answering question on hardware is tough as it completely depends on
the load and usage. Having more RAM will be very useful if the server
is heavily used or if you are a heavy user of performance translators
- io-cache write-behind read-ahead. GlusterFS has io-threads
translator - reads/writes (i/o) are done in parallel on different
threads. Having dual/quad core will increase the performance.
>
> 2) I've ruled out lustre because that solution does not play well with
> selinux. ?Does gluster?
GlusterFS should run fine. We'll fix any issues that you'll face.
>
> 3) How close to the normal file semantics is a glusterFS-mounted file? Can
I
> do an fseek on a file that I am in the middle of reading?
GlusterFS, semantics wise, is as good as any of the disk based
filesystem - ext3 xfs resierfs.
Krishna
>
>
> Thanks,
>
> Steve Friedman
>
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://zresearch.com/cgi-bin/mailman/listinfo/gluster-users
>