Subject: Re: BSDSRCDIR, was Re: mount*(8) linked as one binary ?
To: Luke Mewburn <email@example.com>
From: Ignatios Souvatzis <firstname.lastname@example.org>
Date: 11/05/2000 15:12:19
Content-Type: text/plain; charset=us-ascii
On Sun, Nov 05, 2000 at 12:52:23PM +1100, Luke Mewburn wrote:
> Frederick Bruckman writes:
> > So the objects all go on NFS? That can't be good.
> Depends on what your NFS server is.
> Even using a 2 year old NetApp filer over 100Mb full duplex ethernet
> is still quite fast, since the box has about 1/10th the latency of
> most local disks, and building is more about accessing lots of
> files vs raw seqential file throughput (which is one thing filers
> aren't really `faster' than local disk at).
> And as Chris said, even a decent solaris box with a RAID array is quite
> good too.
Bernd Sieker has built a 1.5_BETAish snapshot on a 40 MHz 68040 Amiga, with
everything over NFS, with the exception of /tmp on MFS.
Server was a fast Pentium roman($N) NetBSD BOX, with heavy usage of softdep
for most, and LFS for the DESTDIR and RELEASEDIR.
Now to the benchmarks:
My machine, 68060/50 MHz/112 MB, local disk: 31 hours
Bernds machine, 68040/40 MHz/64 MB, NFS: 60 hours
1.4.3 build, if I recall right:
he's machine, 68040/33 MHz/64 MB, local disk, 1.4 build: 28 hours
my machne, 68060/50 MHz/112 MB, local disk, 1.4 build: 15 hours
68060/50 with local disk is about twice as fast as 68040/33 with local disk.
68060/50 with local disk is about twice as fast as 68040/40 with decent NFS=
68040/40 with decent NFS is about as fast as 68040/33 with local disk.
Thats not too bad, IMHO. Your mileage will vary when other people use the
network or the NFS server.
* Progress (n.): The process through which Usenet has evolved from
smart people in front of dumb terminals to dumb people in front of
smart terminals. -- email@example.com (obscurity)
-----BEGIN PGP SIGNATURE-----
-----END PGP SIGNATURE-----