[1741] in linux-net channel archive
Re: Performance Enhancements to NFS client.
daemon@ATHENA.MIT.EDU (Alan Cox)
Tue Jan 30 17:53:45 1996
From: Alan Cox <alan@cymru.net>
To: leisner@sdsp.mc.xerox.com (Marty Leisner)
Date: Tue, 30 Jan 1996 09:34:01 +0000 (GMT)
Cc: lim@csua.berkeley.edu, linux-net@vger.rutgers.edu
In-Reply-To: <9601292205.AA25304@gnu.mc.xerox.com> from "Marty Leisner" at Jan 29, 96 02:05:43 pm
> I'm going to look at performance (probably within libc...)
> I'm not sure what's wrong,
> but I feel nfs performance is terrible...
It is badly cached, not using the buffer caches at all and has no option for
doing asynchronous writeback (though this is dangerous in NFS because it
breaks posix semantics badly).
> Also, whats the purpose of having biod on bsd systems and not on Linux?
As a way to readhead/writebehind operations. BSD needs a process context
for these to occur in so it created biod. Linux would have to do similar
although it now has proper kernel threads so it can do it more cleanly.
Alan
