Performance benchmarks
This section defines the process for running the performance benchmarks used in this documentation.
Disk I/O with bonnie++
Bonnie++ provides a simple read/write disk I/O performance benchmark.
yum install -y bonnie++
bonnie++ -d /tmp -u root
This test runs for about 10 minutes.
Version 1.96 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
timperf01.cer 7656M 294 99 68834 15 38694 9 1458 96 114325 10 398.0 18
Latency 30575us 883ms 2524ms 29026us 95506us 90168us
Version 1.96 ------Sequential Create------ --------Random Create--------
timperf01.cern.ch -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 10082 74 +++++ +++ 28534 61 11125 81 +++++ +++ 26378 57
Latency 662us 645us 626us 628us 190us 177us
Write I/O is thus around 68MB/s sequential (from 68834K/sec sequential output), read I/O at around 110MB/s. IOPS corresponds to the Seeks/s value (398).
See the performance tuning chapter for the options to improve I/O
Memory bandwidth
The reference benchmark used for memory performance is the stream benchmark.
yum -y install gcc wget
wget http://www.cs.virginia.edu/stream/FTP/Code/Makefile
wget http://www.cs.virginia.edu/stream/FTP/Code/stream.c
wget http://www.cs.virginia.edu/stream/FTP/Code/mysecond.c
make stream_c.exe
./stream_c.exe
Output such as below is produced
-------------------------------------------------------------
Function Best Rate MB/s Avg time Min time Max time
Copy: 8503.4 0.020086 0.018816 0.020925
Scale: 8342.8 0.020838 0.019178 0.021833
Add: 8644.6 0.028610 0.027763 0.029654
Triad: 8739.0 0.029437 0.027463 0.031352
-------------------------------------------------------------
The performance figure for the service level is based on the Triad figure (i.e. 10927.5 MB/s)