Re: squid2.3stable2 on freebsd3.2

From: "Chris Dillon" <cdillon@wolves.k12.mo.us>
> On Tue, 18 Apr 2000, Alejandro Ramirez wrote:
> > Yes, you have to disable memory pools in your squid.conf file
"memory_pools
> > off", enable truncate instead of unlinkd "--enable-truncate", do not
enable
> > the time hack "--enable-time-hack".
>
> Is there any particular reason you're turning off memory pools and
> using truncate() instead of unlink()? I've not actually done any
> benchmarks with using these options under FreeBSD, but I'd be
> interested if anyone has.

Quote from http://www.squid-cache.org/Doc/FAQ/FAQ-9.html#paging

The meaning of ``swapping'' varies. On FreeBSD for example, swapping out is
implemented as unlocking upages, kernel stack, PTD etc for aggressive
pageout with the process. The only thing left of the process in memory is
the 'struct proc'. The FreeBSD paging system is highly adaptive and can
resort to paging in a way that is equivalent to the traditional swapping
style operation (ie: entire process). FreeBSD also tries stealing pages from
active processes in order to make space for disk cache. I suspect this is
why setting 'memory_pools off' on the non-NOVM squids on FreeBSD is reported
to work better - the VM/buffer system could be competing with squid to cache
the same pages. It's a pity that squid cannot use mmap() to do file IO on
the 4K chunks in it's memory pool (I can see that this is not a simple thing
to do though, but that won't stop me wishing. :-).
by John Line (webadm@info.cam.ac.uk)

Quote from /squid-2.3.STABLE2/configure file:

  --enable-truncate This uses truncate() instead of unlink() when
                          removing cache files. Truncate gives a little
                          performance improvement, but may cause problems
                          when used with async I/O. Truncate uses more
                          filesystem inodes than unlink.."

And also a test made by Duane Wessels (wessels@ircache.net)

http://www.squid-cache.org/Benchmarking/std1/2.2.stable3-unlink/

I dont have a benchmark, but they have certanly increased the speed respose
of my system, right now it handles 9,377 req/min (peak reported), in a 10Mb
BW & 7,000 Cable modem surfers enviroment, it outperforms a lot better than
4 Cisco Cache Engine 550 together (Proved).

> > For FreeBSD, enable SoftUpdates, mount the filesystem with the noatime
> > option, rebuild your kernel without needed drivers, and use the
following
> > options in your kernel config file too:
>
> Softupdates are definately an advantage, as well as noatime.
>
> > makeoptions COPTFLAGS="-O2 -pipe" #Optimizing the kernel
for
> > the best performance
>
> Using -02 on a kernel is not a good idea. There are documented
> problems with gcc generating bad code at these optimization levels.
> Stick with no optimization, or just -O if you absolutely feel you need
> some.

FreeBSD 3.X doesnt come with an standard gcc in the base system, it has
become the standard for FreeBSD 4.X, so the optimization bug its not present
in the 3.X branch :o)

Have Fun...
Ales
Received on Tue Apr 18 2000 - 16:37:52 MDT