Archives


- Beowulf
- Beowulf Announce
- Scyld-users
- Beowulf on Debian

[Beowulf] What are people seeing performance-wise for NFS over 10GbE

Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.

Search

Jan Heichler jan.heichler at gmx.net
Thu Feb 28 12:42:17 PST 2008


Hallo Joe,

Donnerstag, 28. Februar 2008, meintest Du:

JL>    Have a few simple setups that I am trying to figure out if I have a
JL> problem, or if what I am seeing is normal.

JL>    Two 10 GbE cards, connected with a CX4 cable.  Server is one of our
JL> JackRabbits, with 750+ MB/s direct IO and 500-650 MB/s buffered IO 
JL> (read-write), for IO about 10x system ram (~100x RAID cache).

JL>    Getting ok iperf numbers, about 7 Gb/s single thread.  Running NFS,
JL> and seeing ~200-300 MB/s best case between the systems.  Is this what 
JL> others have seen?  Latest drivers from vendor, sent a note to them to 
JL> see if we can figure this out.  Figured I would tap into the 'wulf 
JL> collective memory to learn what other people see.

The best i saw for NFS over 10 GE was about 350-400 MB/s write and about 450 MB/s read.
Single server to 8 simultaneous accessing clients (aggregated performance). 

On the blockdevice i got 550 MB/s write and 1.1 GB/s read performance. 


JL>    2.6.23.14 kernel on both sides, jumbo frames enabled. No switch, just
JL> a CX4 cable.

rsize/wsize are set to? 

NFS3 or NFS4? 

Cheers,
Jan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.scyld.com/pipermail/beowulf/attachments/20080228/29412ece/attachment.html


More information about the Beowulf mailing list