Problem with nfs speed - WAAY to slow, any ideas?
 

Has anyone ever heard of nfs writing *really* slowly because of some sort
of poor configuration?

My nfs writes run at around 400 kB/s, but the nfs reads run at around
10 MB/s.

For a 75 MB file, I tested:

• nfs write: 3:07
• nfs read: 0:07
• scp write: 0:02
• scp read: 0:02

Any ideas where I can begin to track this down? I assume that nfs should
be at least as fast as scp, correct? FYI: I did these tests on our new
computers - they're all 2.6/2.8GHz with 1GB RAM, & GigE NICs; they all
run through a 24-port GigE switch (all ports run at 10/100/1000) & all use
cat 5e cables (no longer than 7 ft, not that that should matter).

yeah u need to add some extra commands to your mount command.
nfs can be mounted for either throughput or responce time.(responce time is default) this is because it is usually used for small files.

anyway i did a test about 1 year ago and got throughput ranging from 200k-998k(10mbit) where 200k was default and 998 was after changing transmission size.


to do this u need to add the line wsize and rsize to your mount command as an option. eg -o wsize=...........

w = write r=read

have fun

Good idea, but I've already tried that... I tried ranges from 4096 to 32K - no change.

Any other ideas?

bump.. I've seen this question come up a lot online (google), but I've never seen a thread that has an answer. Anyone?

well this is how i fixed the problem, also if your benchmark is inbound increase rsize and decrease wsize as a high wsize will wast network space.