Re: [squid-users] reiserfs, 2,4 kernel, and squid

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Joe,

I read about your article about "High Performance Web Caching With Squid".
Thanks very much for that, highly appreciated.

With regards to 2.2.[18|19] being faster than any current 2.4.x kernels as
far as reiserfs + squid + single CPU machines is concerned ... do you have
any figures / benchmarks / sites for this claim, for us to appreciate?

Thanks also for saying that 2.4 is not as stable as we want it to be ... at
least I am not alone in thinking about it. 2.4.10 seems to have lots of
merges from ac and aa, plus the VM is being reworked on.

Thanks,

John

On Sat, 22 Sep 2001 17:30, Joe Cooper wrote:
> Jesus M. Salvo Jr. wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> >
> >
> > Where I work, there is an opportunity to use Linux as a server where all
> > are nt/w2k servers, mostly due to licensing issues. The administrator, an
> > msce, has approached me for some help in getting squid up and running to
> > replace ms isa server.
> >
> > We have been running squid-2.3STABLE on a redhat 7.1 distro and an ext2
> > filesystem on a very old hardware, pentium mmx and even an ISA ne2000
> > card, as a proof-of-concept. The permanent setup will use a much better
> > hardware, possibly a pentium II, PCI network card, and 128Mb of memory or
> > more.
> >
> > To cut the story short, I would like to maximise the performance of the
> > box, using a 2.4 kernel and using reiserfs instead of ext2, and using
> > squid-2.4STABLE to make use of diskd.
> >
> > The question now is:
> >
> > Which 2.4 kernel is best reiserfs + squid?
> >
> > 2.4.0 to 2.4.3 seems to be out of the question, based on this:
> >
> > * http://www.google.com/search?hl=en&q=reiserfs+2.4.3+squid
> > * http://www.namesys.com/download.html ( under DANGER )
> >
> > It seems that the best option is either:
> >
> > * 2.4.5 + reisefs umount patch
> > * 2.4.7
> > * 2.4.4
> > * 2.4.10-pre4
> >
> > Can anyone out there using squid-2.4STABLE give feedback on what 2.4
> > kernel ( plus any patches ) are you using ...and for how long??
>
> I'll say it again (check the list archives for several posts with more
> detail on the subject. 2.2 is slightly faster (and much more stable in
> my experience) for Squid on single CPU machines.
>
> If you /insist/ on using 2.4, I've been having pretty good luck with
> 2.4.5+ac5(includes the umount fix)+tons of other patches.
> 2.4.8+ac12+tons of other patches is great on IDE boxen and ReiserFS is
> fine, but it has a bug on SCSI machines leading to hideously destructive
> crashes (filesystem corruption is rampant on ext2 filesystems during
> these crashes).
>
> > Worst case, I can go back to a 2.2 kernel ... but we wont have the
> > zero-copy feature of 2.4.
>
> Squid does not make use of zero-copy features of kernel 2.4.
>
> I recommend using 2.2.18 or 2.2.19 if you have no plans to use anything
> but Squid on the machine. 2.4 is cool, but it is quite hard to create a
> stable version of it, and it's hardly worth the effort, given that 2.2
> is faster for Squid.
>
> As Chris has just said, 2.4.10 is probably a 'winner' for the 2.4 kernel
> and ReiserFS. I'm looking forward to it, and it appears from my vantage
> that all of the people who need to be coordinating things /are/ actually
> doing so now. In 2.4.0-2.4.9 there was always some piece or multiple
> pieces of the code that were enough out of sync with other pieces that
> patches were needed to even achieve some amount of stability. In other
> words, no kernel from 2.4.0-2.4.9 could be sucessfully used with the
> combination of components that I needed to work together without heavy
> patching.
> --
> Joe Cooper <joe@swelltech.com>
> Affordable Web Caching Proxy Appliances
> http://www.swelltech.com
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.4 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iEYEARECAAYFAjusS88ACgkQAvd5SY4qWYxcOACfcLDSdNenanNrCXCYEyIk2Oc9
pW8An1X1kzE7udU7KiVbxZk5RxIoX2id
=dQNq
-----END PGP SIGNATURE-----
Received on Sat Sep 22 2001 - 02:24:57 MDT