[1728] in linux-net channel archive
Re: Performance Enhancements to NFS client.
daemon@ATHENA.MIT.EDU (Alan Cox)
Mon Jan 29 08:56:51 1996
From: Alan Cox <alan@cymru.net>
To: lim@csua.berkeley.edu (Lincoln Myers)
Date: Mon, 29 Jan 1996 11:08:57 +0000 (GMT)
Cc: linux-net@vger.rutgers.edu
In-Reply-To: <199601272229.OAA07656@soda.CSUA.Berkeley.EDU> from "Lincoln Myers" at Jan 27, 96 02:29:30 pm
> Including, say, multiple reads or writes at once (e.g. for a 64k read
> sending up to 8 read requests out in a row rather than send one, wait
> for reply, send next, ...)? That's what I was going to focus on.
You generally have to be fairly conservative about doing extensive NFS
readahead. If you read too far ahead you get two problems
1. You load the network a lot (network is a valuable resource too). It
also causes server load to rise and you to drop packets on a slower
network card when 64K turns up in back to back replies. You do need
some readahead though..
2. If you readhead too far you keep having to discard the readahead
data as out of date if the program doesn't get to it in time.
Alan
