[Beowulf] adding a cluster to a traditional supercomputer

[Joachim Worringen]

Florent Calvayrac wrote:
> the question is open for the filesystem ;  should they rely on NFS over
> Gigabit ethernet and attack their "old" fileserver (which is excellent),
> with some fears for some heavy I/O codes, add fiberchannel ports to
> some (or all) nodes on which the I/O bound codes would be dispatched with
> a good PBS setup (best solution but expensive), or add another, 
> dedicated NFS server with
> several gigabit ethernet cards, each for serving only parts of the new 
> cluster,
> this "new" fileserver itself getting its files via CFS, fiberchannel, or 
> even NFS from the "old" fileserver ?

You already listed many of the possible options. It is very difficult to make an 
informed decision as long as you don't have good quantitative information on the 
actual I/O load, and also specify the work-flow and reliability requirements for 
the files generated or accessed. For some applications, a solution like PVFS 
might be sufficient, or even local scratch space might fit.

Adding NFS servers to serve out the CFS to the cluster nodes is a common 
solution, although NFS is not well suited for "real" parallel I/O. More 
complex/reliable/powerful I/O solutions can become quite expensive, either in 
terms of money to effort spent (think of Lustre).

  Joachim

-- 
Joachim Worringen - NEC C&C research lab St.Augustin
fon +49-2241-9252.20 - fax .99 - http://www.ccrl-nece.de