On 4/2/07, Adam Findley <[EMAIL PROTECTED]> wrote:
So I have setup a fresh new Archlinux workstation here at the BYU CS
department, and it's NFS read speeds are tremendous, but the write
speeds are horrendous.  Anyone know what causes such problems?

Question 1: Is your server exporting filesystems with the async option?

Question 2: Why NFSv2?  NFSv3 allows clients to specifiy
non-synchronous behavior for certain classes of writes.  NFSv2 has no
such capability.

Combining the server's default "sync" (synchronous write) export
option with the client's usage of NFSv2 virtually guarantees the worst
possible (aka completely synchronous) write performance.

Some consider async writes unsafe because clients can never really
know when the server commits a write to permanent storage.  My
experience has been that server failures are rare enough, and the
increased performance of async writes high enough to override such
concerns.  YMMV depending on the nature of your workload.

Chris

/*
PLUG: http://plug.org, #utah on irc.freenode.net
Unsubscribe: http://plug.org/mailman/options/plug
Don't fear the penguin.
*/

Reply via email to