Timing file read/write over NFS

[Michael Pelletier]

bruce.labitt at autoliv.com wrote:
-----
What I did find was an asymmetry in read vs write speeds of 2:1.  Roughly I
could get 100 MB/sec read over the network, but only 45-60 MB/sec write over
the network.  Anyone have an idea why this would be true?
-----

Hi Bruce,

You'll want to have a look at the configuration options on the NFS server's
export.  Your filesystem may be exported in "sync" mode (which is the
default) which prevents the server from acknowledging write requests until
the data has been committed to disk - "async" is the opposite option.

You might also try "no_wdelay" - normally NFS will pool write requests if it
detects a sequential write pattern, but if the writes are unrelated, this
will introduced unnecessary delays.  Naturally, setting async makes
no_wdelay irrelevant.

Going to "async" might be just the ticket to improving your write
performance.  Of course, it introduces some risk of inconsistency since a
server write failure might already have been acknowledged as complete to the
NFS client.

	-Michael Pelletier.