Hardware platform, software stack and methods
All tests were run on a Dell Latitude D620 machine, with the following hardware configuration:
-
Core2 T7200 CPU @ 2.0 GHz (dual core, 4 MB L2 cache)
-
2 GB of DDR2-667 RAM
-
Quadro NVS110 videocard (used in text-only mode)
-
a Seagate ST980825AS 7200 RPM 80 GB SATA hard disk drive (in IDE compatibility mode, as the D620's BIOS does not support AHCI operation)
-
O.S. 1: Red Hat Enterprise Linux 6.1 amd64
-
O.S. 2: Debian 6.0.2 amd64
Both Red Hat Linux 6.1 and Debian 6.0.2 have a somewhat similar software stacks. Below you can find the specific version number of some key packages:
|
Package |
Red Hat version |
Debian version |
|
kernel |
2.6.32-131.6.1 |
2.6.32-5 |
|
libc |
2.12-1.25 (standard version) |
2.11.2-10 (embedded version) |
|
apache |
2.2.15-9 |
2.2.16-6 |
|
php |
5.3.3-3 |
5.3.3-7 |
|
mysql |
5.1.52-1 |
5.1.49-3 |
|
postgresql |
8.4.7-2 |
8.4.8-0 |
The following benchmarks were run against the two OS installations:
-
CoreMark 1.0 (http://www.coremark.org)
-
STREAM (http://www.cs.virginia.edu/stream/)
-
Membench (http://www.assyoma.it/index.php/repository?func=fileinfo&id=5)
-
Apache benchmark 2.3 - from the Apache HTTPD v2.2.19 (http://www.apache.org/)
-
PHP Benchmark 1.0 (http://www.php-benchmark-script.com)
-
Sysbench 0.4.12 (http://sysbench.sourceforge.net/)
-
MySQL Benchmarks suite – from MySQL v5.1.58 (http://www.mysql.com/)
-
PgBench 1.87.2.1 – from PostgreSQL v8.4.8
Where possible, I did not use the precompiled binary benchmarks available in RH/Debian repository, but I compiled the benchmark programs from sources. Why? First, because the benchmark source code had to be from the very same version for produce comparable results. Second, in this manner I give the two different libc stacks the possibility to highlight their relative strengths / weaknesses. Third, the distribution with the better / stronger compiler can leverage it.