[Beowulf] Network Filesystems performance

[Glen Dosey]

On Fri, 2007-08-24 at 08:45 -0400, Lawrence Sorrillo wrote:
> I am uncertain what this step does....
> 
> now we unmount the NFS share, recreate the file on the server, and remount
> it to clear the client cache but leave it cached on the server
> 
> [root at wopr1 ~]# dd if=/mnt/array3/file.dd of=/dev/null bs=4k
> 524287+0 records in
> 524287+0 records out
> 2147479552 bytes (2.1 GB) copied, 18.5161 seconds, 116 MB/s
> 
> 
> What does it accomplish?
> 
> Thanks.


I was using this step to illustrate that NFS itself is capable of
transferring data across the GigE network at near wires speeds. 

By unmounting the NFS share we clear the file from being cached in RAM
on the client, ensuring that it must be gotten from the server again via
NFS as opposed to being grabbed from RAM on the client. 

By recreating the file on the server we guarantee that it is cached in
the servers RAM. We could have done any activity that caused the file to
be read on the server and accomplished the same thing. Basically the
goal was to get the file into memory on the server so we eliminate any
disk activity from the NFS test.

At this point we mount the NFS filesystem on the client and run the dd
test. The file is cached on the server and not on the client. In this
situation we see the NFS read performance is basically wire speed for
the GigE.

That might seem to indicate the problem is the read speed of the
underlying disk array, but as I've said, the NFS server can read from
the underlying disk array at 160MB/s.