[rhelv6-list] Distributed parallel fault-tolerant file systems

[Prentice Bisbal]

I noticed two other people already responded to this with the RH party
line of "Use Gluster", while I have nothing bad (or good) to say about
Gluster since I've never used it,  it's not the only option out there,
and you should look at others, too.

I specialize in HPC, and Lustre is definitely the most popular of the
parallel filesystems used in that space. The next most popular is
probably IBM's GPFS. Another popular option is PVFS and it's descendant
OrangeFS. There are plenty of university's and national labs are using 
Lustre, GPFS, or OrangeFS,

Lustre and PVFS/OrangeFS are open source. Since Sun was leading the
development of Lustre, WhamCloud, Inc. took stewardship of the Lustre
project after the Oracle buyout of Sun. GPFS is a commercial product,
and an expensive one at that, but I know plenty of people who've decided
it's worth the cost (or really, really want someone else to blame when
things go south. OrangeFS/PVFS are open source projects from Clemson
University.

There plenty of academic papers and whitepapers comparing the
performance and architecture of parallel filesystems. I would google the
names of each of the parallel filesystems, and spend some time learning
about all of them and parallel filesystems in general before making a
decision. They all have their strengths and weaknesses.

Prentice


On 03/05/2012 11:34 AM, Bohmer, Andre ten wrote:
> Hello,
>
> Any advise or experience from production systems regarding distributed
> parallel fault-tolerant file systems like Lustre ?
> We would like to offer high performance, redundant storage via nfs and cifs
> from Redhat servers.
>
> At this we've HP's XP9000 Ibrix in use, but performance is not all that
> great.
>
> Cheers,
> Andre