Hello all... I have 3 HUGE NFS servers (Each one with 2 -3 terabytes each), and 10 clients that pull a consistent 20-25 MBit total off of
the 3 NFS servers. Everything is running fine, but performance isn't the greatest... I can never pull more than 1 Megabyte per second off of any of the servers (Even though local performance of drives on the servers is GREAT)... My question is this? How would switching everything to gigabit copper instead of 100 MBit lines help. Obviously Gig is faster, but since I am not even using the 100 MBit I am wondering? The latency should go down because it transfers so much faster? Any ideas for me? Anyone do this themselves that can answer?
Thanks,
Matt
>>>>> " " =3D=3D Matt Heaton <[email protected]> writes:
> Hello all... I have 3 HUGE NFS servers (Each one with 2 -3
> terabytes each), and 10 clients that pull a consistent 20-25
> MBit total off of
> the 3 NFS servers.=A0 Everything is running fine, but performance
> isn't the greatest... I can never pull more than 1 Megabyte per
> second off of any of the servers (Even though local performance
> of drives on the servers is GREAT)...=A0 My question is this?=A0
> How would switching everything to gigabit copper instead of 100
> MBit lines help.=A0=A0 Obviously Gig is faster, but since I am not
> even using the 100 MBit I am wondering?=A0 The latency should go
> down because it transfers so much faster?=A0 Any ideas for me?=A0
> Anyone do this themselves that can answer?
You should be getting 10Mbyte/sec on a switched 100Mbit line. I
suggest you try NFS over TCP rather than UDP (note that this requires
a fairly recent kernel -2.4.20 or above- on the server side).
Cheers,
Trond
-------------------------------------------------------
This SF.NET email is sponsored by:
SourceForge Enterprise Edition + IBM + LinuxWorld = Something 2 See!
http://www.vasoftware.com
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs