Re: [squid-users] Best OS for Squid?

Has anyone tested Squid on FBSD 4.4S since the new dirpref code went in? In
theory this could be of direct benefit to Squid.

This is NOT intended to start a who's best, fastest, better war.

Just curious if anyone had done any testing with Squid and the dirpref code
yet.

Jerry

----- Original Message -----
From: "Joe Cooper" <joe@swelltech.com>
To: <scanner@jurai.net>
Cc: "Joel Jaeggli" <joelja@darkwing.uoregon.edu>; "Eric Galarneau"
<ericga@cae.com>; "'Vosburgh, Brian P, CTR, WHS/BB'" <bvosburgh@whs.mil>;
<squid-users@squid-cache.org>
Sent: Tuesday, October 30, 2001 5:06 PM
Subject: Re: [squid-users] Best OS for Squid?

> That may be so. But there is no need to argue which OS is faster based
> on the filesystem. In fact, it's silly to argue about it. We know
> which OS is faster for Squid. There are tons of benchmarks that tell
> us. Not to mention that this same debate happened here on this very
> same list less than a year ago.
>
> Soft Updates are really cool, and everyone here thinks very highly of
> FreeBSD. But Squid+aufs+ReiserFS on Linux is faster. Sorry. That's
> just the way it is. Even with the Linux machine using DiskD, ReiserFS
> is faster for Squid workloads. It just is. Perhaps there are ten
> filesystem experts who will support your case that Soft Updates are
> "superior technology", and maybe they are--but as it is implemented
> under FreeBSD, it is not faster for Squid's millions of little tiny files.
>
> There are even better theoretical filesystem solutions than ReiserFS for
> Squid workloads, for sure, but none that work today. I'll be happy when
> COSS is finished and we can all use an overlaid 'filesystem' that
> performs better than anything currently available and performs similarly
> on most Squid supported platforms.
>
> All that said, if you prefer FreeBSD that's great, you /can/ get a
> reasonably fast Squid by using DiskD and Soft Updates. It will handle
> most peoples needs just fine.
>
> Use what you're comfortable with...it will save you time and trouble, in
> most cases. The only time it matters is if you're supporting a pretty
> large bunch of bandwidth (like 10Mbits+).
>
> But again...This sort of debate doesn't need to keep coming up. The
> Squid list isn't the place for OS wars. It accomplishes little when the
> debate is already over for the existing filesystems and Squid.
>
> scanner@jurai.net wrote:
>
> > On Tue, 30 Oct 2001, Joel Jaeggli wrote:
> >
> >
> >>In my experience softupdates brings the performance of ufs up from abysmal
> >>up to close or equal to ext2. ext2 is more brittle than ufs with
> >>soft updates. but you weren't planning on crashing it were you?
> >>
> >>I think that real comparisons of filesystem performance/stability should
> >>be made between lfs on freebsd and reiserfs/ext3/jfs on linux, given the
> >>amount of disk and the number of objects that large caches tend to
> >>handle...
> >>
> >
> > Softupdates will outperform JFS's by a small margin 90% of the
> > time. Google for greg ganger. Go to his home page and read his paper's
> > section for a paper on Softupdates VS Journaling. SU has an edge over
> > JFS. There both trying to accomplish the same thing but I believe SU to be
> > a better technology.
>
> --
> Joe Cooper <joe@swelltech.com>
> http://www.swelltech.com
> Web Caching Appliances and Support
>
Received on Tue Oct 30 2001 - 15:51:18 MST