Benchmarking OCFS2

Posted on March 18, 2010 by Chintana

Many bad things can be said about benchmarks but they are useful to have some understanding about the thing involved. This benchmark is no different. Take it for what it’s worth.

The setup is as follows. A 1TB iSCSI volume is formatted with OCFS2. This iSCSI volume is mounted on 3 machines running Debian. Numbers are taken after running bonnie++ on the mounted iSCSI volume at the same time on the 3 machines simultaneously. 3 machines are connected to a Dell PowerConnect 3450 switch with CAT6 cables (1Gbps) and iSCSI device is connected to the same switch with a CAT6 cable (1Gbps). Speed in the graph K/sec is Kilo-bytes per second.

Machine A (Dell PowerEdge R410) – bonnie++ version 1.03d. 2 Quad Core Intel Xeon @ 2.27GHz (with HT OS sees 16 CPUs) with 32GB RAM.

  Sequential output Sequential input Random
  Per chr Block Rewrite Per chr Block Seeks
Size K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
63G 21049 39 46081 14 20134 6 58457 91 107392 13 537.5 1
  Sequential create Random create
  Create Read Delete Create Read Delete
files /sec %CPU /sec %CPU /sec %CPU /sec %CPU /sec %CPU /sec %CPU
16 468 28 +++++ +++ 10862 44 2377 83 +++++ +++ 6713 63

Machine B (Dell PowerEdge 1950)- bonnie++ version 1.96. 2 Core 2 Duo Intel Xeon 3.20GHz (with HT OS sees 8 CPUs) with 10GB RAM.

  Sequential output Sequential input Random
  Per chr Block Rewrite Per chr Block Seeks
Size K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
19904M 102 99 22214 10 4159 1 272 93 45714 4 213.4 0
  Sequential create Random create
  Create Read Delete Create Read Delete
files /sec %CPU /sec %CPU /sec %CPU /sec %CPU /sec %CPU /sec %CPU
16 388 12 +++++ +++ 3964 13 731 25 +++++ +++ 3139 58

Machine C (Dell PowerEdge 1950) – bonnie++ version 1.03d. 2 Core 2 Duo Intel Xeon 3.20GHz (with HT OS sees 8 CPUs) with 10GB RAM.

  Sequential output Sequential input Random
  Per chr Block Rewrite Per chr Block Seeks
Size K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
20G 13949 28 34214 13 5771 2 26713 47 53736 7 153.0 0
  Sequential create Random create
  Create Read Delete Create Read Delete
files /sec %CPU /sec %CPU /sec %CPU /sec %CPU /sec %CPU /sec %CPU
16 461 30 +++++ +++ 8116 60 2085 85 +++++ +++ 4714 71

Have to perform another benchmark with 3 identical servers and the same version of the program. On a later date…

Another interesting file system is GFS by RedHat. In Debian, configuring GFS turned out to be a royal pain in the arse. Documentation on GFS says, “current supported maximum size of a GFS/GFS2 file system is 25 TB” and “maximum number of nodes supported in a Red Hat Cluster deployment of GFS/GFS2 is 16″. Feh.

This entry was posted in benchmark, distributed file systems, iscsi, ocfs2. Bookmark the permalink.

One Response to Benchmarking OCFS2

  1. Pingback: cpu q9550 | Buy CPU