Topic: FreeBSD 7.0

So I have this ongoing feud at work. Everyday people bash on BSD when a ticket comes through and they cannot use it. I've pointed out numerous times that if it weren't for BSD they wouldn't have OpenSSH. This doesn't coax them. So I'm obviously evangelizing PF, ZFS support, etc ... but always seem to be shown up with a Linux counterpart...

.... until now.

I've stirred up quite a bit of controversy here at work because of the benchmark results that Kris posted. I decided to write to him regarding this issue:


I was wondering if you could publish a little more details about how
the Linux boxes were set up, i.e. Distribution used, default setup,
any other changes made to the linux kernel. Some people here at my
place are greatly interested in these details and as they seem to
believe these benchmarks are flawed.

Also, the my.cnf link has a 404 error. Thank you. smile

David Ontiveros

And Kris' reply:

I used a stock fedora 8 release that was updated to the latest official
packages of everything.  I didn't recompile anything by hand except for
*sql and sysbench, which were built with the same options as on FreeBSD.

The my.cnf is here:

and there is a writeup here:

Hope this helps,

My reply:


One more question to put the pundits to rest. I do apologize for the
questions. I'm trying to get my company to recognize FreeBSD 7.0 as a
viable OS choice for provisioning of new servers. Was this done in an
environment similar to a liveCD? My coworkers and superiors are
curious about what services were running at the time of deployment,
was it in a live cd environment, types of disk, hardware information,
etc. I do know it's a lot to ask, especially for a stranger, but I've
been touting FreeBSD 7.0 as a valid contender against the latest
2.6.22 kernels. I seriously think this is a good product. You guys
made tremendous improvements. I'm just trying to evangelize. I've been
using FreeBSD since 4.7. I am appreciative of your work. Thank you.

Kris' reply:

It's a 2*E5320 xeon system (2 * 4-core) with 4GB of RAM, in i386
(32-bit) mode.  Performance is basically the same on other 8 core
configurations like 4*2-core opterons, and also in 64-bit mode.  Disks
are not relevant since it's memory resident.  I also didnt bother
shutting off other standard services like cron, syslogd, etc since they
were completely idle and do not interfere with the test.

This was not a live CD environment, I recompiled the kernel to switch to
the ULE scheduler and added the default packages of sysbench and mysql,
then configured mysql as in the URL I gave.  You could do the latter 2
in a live CD but not recompiling the kernel.


Re: FreeBSD 7.0 … /sysbench/

There is something new, but read the text too before starting a war big_smile


Freedom is the sure possession of those alone who have the courage to defend it. --Pericles

Re: FreeBSD 7.0

This is going to be interesting, since Kris tested with the same options compiled in on both OS's. I did read a mailing list about how the scheduler might throw off some results.