[rt-users] FYI: Some very basic benchmark results

[Tim Wilson]

Hey everyone,

I've been working on moving our RT system from Apache1 with mod_perl to
Apache2 with FastCGI. I got both configs working and thought I'd share some
very simple benchmark results with the list.

Server details: Dell 1750 with dual 2.4-GHz Xeon, 1-GB RAM, 36-GB SCSI HD in
RAID config, Debian Linux w/ 2.4.28 SMP kernel

Software: Debian's RT package (3.2.3-1), PostgreSQL (7.4.7-2), Apache1
(1.3.26-0woody6), Apache2 (2.0.53-5), mod_perl (1.26-3.0woody1), FastCGI
(2.4.2-5)

Our benchmarking procedure was very simple. We looked through our
3000-ticket RT system for the longest ticket we could find (call it Ticket
A) and identified three other big tickets to use too. Four of us logged into
RT and used the search box to pull up all four tickets simultaneously. We
simply timed how long it took for Ticket A to load completely. We did three
trials in each configuration and found very consistent results between
trials. (The server was not in use by other users at the time of our test.)

Our three configs were:

A. Apache1 with mod_perl
B. Apache2 with FastCGI (1 instance)
C. Apache2 with FastCGI (4 instances)

Results with the average time to load Ticket A for all three configs:

Config A: 14.89 s
Config B: 15.93 s
Config C: 12.31 s

The Apache2/FastCGI config certainly "feels" quicker in our everyday use. We
plan to upgrade to RT 3.4 as soon as a Debian package is available. I'm also
planning a kernel upgrade in the near future.

-Tim

-- 
Timothy Wilson
Technology Integration Specialist
Hopkins ISD #270, Hopkins, MN, USA  (44š56.013'N  93š24.736'W)
ph: 952.988.4103  fax: 952.988.4311  AIM: tis270