MyDeDub 0.4 performance data

For these tests consumer hardware is used. MySQL on fibre disks will probably give much better performance.

Test systems

database server

  • disks: 2xMaxtor 6L080P0
  • 1GB ram
  • AMD Sempron 2300+ (1.6GHz, 256KB cache)

database server disk performance

2x 16GB partition in RAID0
hdparm -t:90MB/s
IOPS (read):93
IOPS (write):390 (probably due to write caches)
dd if=/dev/zero of=/dev/md0 bs=4096 count=262144:50.4MB/s
dd if=/dev/md0 of=/dev/null bs=4096 count=262144:92.7MB/s

MySQL info

Databases stored on the RAID0 volume. Filesystem: ext2(!), tables: MyISAM, default Debian MySQL configuration (5.1.45-1)

MyDeDub system

  • 4GB ram
  • Intel Core2Duo Q9450 (2.66GHz, 6144KB cache)

network

1Gb, connected via consume 3COM switch and Via VT6102 and Realtak RTL-8169 ethernet controllers

Tests

MyDeDub datastore is emptied after each test with writes ('TRUNCATE table' in MySQL), MyDeDub is restarted after each test, runs with 256MB cache

empty datastore

hdparm -t:7.67MB/s
IOPS (read):24.9
IOPS (write):24.6
dd if=/dev/zero of=/dev/nbd0 bs=262144 count=2 oflag=dsync:1.6MB/s
mke2fs -j /dev/nbd0 (*1)2m56.062s
tar xjf /linux-source-2.6.30.tar.bz2 + sync (*2):7m56.735s
copying /home-filesystem using dd, bs=262144 (6GB of sources and photo's) (*3):1.4MB/s
dd if=/dev/nbd0 of=/dev/null (*4):3.6MB/s
Note: step *2 uses the data from *1 with a restart of MyDeDub.
Note: step *4 uses the data from *3 with a restart of MyDeDub.
Performance graph of *3:

full datastore

hdparm -t:3.16MB/s
IOPS (read):24.83
IOPS (write):17.7
dd if=/dev/zero of=/dev/nbd0 bs=262144 count=2 oflag=dsync:505KB/s
dd if=/dev/nbd0 of=/dev/null bs=262144 count=2:3.4GB/s
mke2fs -j /dev/nbd05m29.491s
tar xjf /linux-source-2.6.30.tar.bz2 + sync:10m12.498s
copying /home-filesystem using measurecopy, blocksize=1048576
(6GB of sources and photo's initial
data is / of a Debian system):
see below *5
Performance graph of *5:


feedback