NFS benchmarking

[Adam McDougall]

I dont know about nfs specific benchmarks, it probably matters largely on
what the clients are doing, but you may try asking the net@freebsd.org
list, they would probably have some suggestions.  If you just want
something to play with for right now,  I would try postmark, bonnie,
iozone, and whip up a script with maybe some file copies or multiple
instances of dd if=/dev/zero of=somefile bs=1024k count=256.  I am pretty
certain a pIII 866 should be plenty of raw horsepower for simple data over
gigabit but it may not be as great with NFS on top.  Remember, NFS uses
network (big endian) byte order even if your server or clients are small
endian, so a big endian server may perform better at equal cpu speeds.  I
think most NetApp servers are Alpha (big endian)..

On Fri, 9 Feb 2001, Edward Glowacki wrote:

> OK, 2 systems here, both doing NFS filesharing:
> Network Appliance w/ 10k RPM fibrechannel disks
> PIII 866 w/ hardware RAID and 10k RPM SCSI 160 disks (running FreeBSD 4.2)
>
> How the heck do I benchmark the damn things?  Does it even matter?
>
> Basically the question is, is the PIII 866 setup sufficient to more
> or less max out a gigabit link sharing files via NFS (assuming it's
> connected directly to a sufficient number of 100mbit clients), or
> do we require a Network Appliance for the I/O performance?  My own
> thought is that the PIII will hold its own if we pop a gigabit card
> in there.
>
> So anyone know of a good NFS performance benchmark?
>
>
> --
> Edward Glowacki			glowack2@msu.edu
> Computer Laboratory, Technical Support Services
> Michigan State University
>
> _______________________________________________
> linux-user mailing list
> linux-user@egr.msu.edu
> http://www.egr.msu.edu/mailman/listinfo/linux-user
>