[Pvfs2-users] Does this Architecture defeat the purpose of PVFS2 in Anyways?

Kyle Schochenmaier kschoche at gmail.com
Fri May 16 18:05:52 EDT 2008


As long as you are not doing heavy application communication for your
jobs as well as being an IO node over the same gigE
link on each server I dont forsee a significant problem.

Probably best to wait to see what others have to say.  I use a similar
setup, I think.. I might be confused by the description.

~Kyle

On Fri, May 16, 2008 at 4:14 PM, sonatabar bar <sonatabar at yahoo.com> wrote:
> Hello PVFS2:
>
> Compute Cluster Architecture:
>
> Top most Layer: Frontend+Compute Nodes AS PVFS2_Client
>
> Next Layer: GigE/Infiniband Interconnect
>
> Next Layer: A Metadata Server & Bunch of PVFS I/O
> Nodes
>
> Bottom Layer: A Separate Storage array for each I/O
> nodes.
>
> With the above architecture in mind, I plan to create
> a huge pvfs2-storage pool from all of the storage
> mounted on the I/O nodes. This large pvfs2-storage
> pool will be  mounted on all of the compute nodes and
> the front end of the clusters from the PVFS2 I/O
> nodes.
>
> I just wanted to get a heads up on this architecture
> if any.
>
> I have read FAQ's and more, look forward for any
> caveats anyone can think of.
>
> Thank you - í
>
>
>
>
>
> _______________________________________________
> Pvfs2-users mailing list
> Pvfs2-users at beowulf-underground.org
> http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users
>



-- 
Kyle Schochenmaier



More information about the Pvfs2-users mailing list