Subject: Re: NetBSD - ready for prime-time fileserving?
To: Neil A. Carson <firstname.lastname@example.org>
From: Frank van der Linden <email@example.com>
Date: 10/08/1999 01:47:27
On Thu, Oct 07, 1999 at 03:39:17PM -0700, Neil A. Carson wrote:
> Solaris clients have problems with FreeBSD servers on v3 stuff. Linux
> clients have occasional attribute problems with v2 and v3 BSD servers.
> TCP NFS in BSD has problems.
> The Linux client talking to *BSD servers has occasional problems.
> To be honest, the whole NFS stuff is justa mess: If you have a solution
> that works, I'd say stick with it. Just my tuppence worth.
>From what I have seen myself, Linux has problems with NFS no matter
what it talks to / is talked to by. I wouldn't use it for this
kind of thing.
I have not seen problems between Solaris and NFS (either direction)
since NetBSD 1.4.1. FreeBSD may have had problems there because
they only recently loosened the directory cookie verifier check,
which I disabled a long time ago. Since I went through some pains
to get the directory cookie stuff right, I am pretty sure NetBSD
will interoperate correctly with other systems there.
There are some performance issues on the server side the *BSD code,
which aren't fixed in the latest releases of all of them
(commit operations do too much work).
I think FreeBSD is working on those, as are we (to be committed
before the end of the year, anyway). But that doesn't do you
any good, you want to run a release.
For a fileserver, that isn't doing much else, increasing the size
of the static buffer cache that we still have (but looking at
Chuq's UBC stuff not for long anymore) will work OK to boost
I'd currently pick Solaris as a fileserver, because its NFS performance
is pretty damn good. Oops, I shouldn't be saying that, I'm supposed
to be a rabid NetBSD advocate :-)