[Beowulf] Network Filesystems performance
Buccaneer for Hire.
buccaneer at rocketmail.com
Sat Nov 10 10:19:10 PST 2007
--- Chris Samuel <csamuel at vpac.org> wrote:
> On Fri, 24 Aug 2007, Michael Will wrote:
> > I tested several NFS server configurations with a
> 19 node cluster.
> Same, but with 150+ nodes and about 500+ CPUs.
> > The first advise is to stay away from redhat for
> file servers since
> > they have some bursty I/O bugs and don't support
> We've run through Fedora (OK, but release cycle too
> quick), Ubuntu
> (Dapper has longer support, but a bit flakey) and
> now run Debian Etch
> (very happy).
> Fedora & Ubuntu were on a pair of IBM e325 with
> 2x2.0GHz Opterons and
> gigabit ethernet connected through to each side of
> an Apple XServe
> RAID array.
> Debian is running on an IBM x3550 with a pair of
> 3GHz woodcrests and a
> couple of bonded gigabit ethernets connected through
> via SAS to some
> external SATA EXP3000 arrays. We decided to use the
> hardware RAID to
> make smaller individual RAID5 arrays and then stripe
> across them
> using the Linux MD driver. Seems to work rather
A stock kernel will from Redhat will not give you the
performance you need. We use Fedora and
patch/configure a standard kernel from kernel.org.
On the back end, we are currently using Clariions
(3-80) and build smaller LUNs per EMC recommendations
and metaLUN them. That really helps performance. Our
industry is know for humongous data files.
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
More information about the Beowulf