Difference between revisions of "Benchmarking: CPUBurn-in"

From Define Wiki
Jump to navigation Jump to search
Line 3: Line 3:
 
CPUBurn-in files can be obtained from:
 
CPUBurn-in files can be obtained from:
 
<code>//storage1/software/hpc/benchmarks/cpuburn-in/cpuburn-in.tar.gz</code>
 
<code>//storage1/software/hpc/benchmarks/cpuburn-in/cpuburn-in.tar.gz</code>
 +
 +
== Install Libraries ==
 +
<syntaxhighlight>
 +
[root@node1 ~]#yum -y install glibc.i686
 +
</syntaxhighlight>
  
 
== Extract CPUBurn-in ==
 
== Extract CPUBurn-in ==
 
 
<syntaxhighlight>
 
<syntaxhighlight>
 
[root@raldisk ~]# tar zxvf cpuburn-in.tar.gz  
 
[root@raldisk ~]# tar zxvf cpuburn-in.tar.gz  

Revision as of 16:07, 9 January 2013

Download CPUBurn-in

CPUBurn-in files can be obtained from: //storage1/software/hpc/benchmarks/cpuburn-in/cpuburn-in.tar.gz

Install Libraries

[root@node1 ~]#yum -y install glibc.i686

Extract CPUBurn-in

[root@raldisk ~]# tar zxvf cpuburn-in.tar.gz 
cpuburn-in
README

Run CPUBurn-in

It may be necessary to chmod the extracted binary to run (or execute the binary with sh):

[root@raldisk ~]# chmod +x cpuburn-in

Note: cpuburn-in should have one running instance per core/virtual core. To achieve this, create a script to calculate the number of available core and run cpuburn-in a sufficient number of times to fully stress the cpus:

START=1
END=`cat /proc/cpuinfo | grep processor | wc -l`

for (( c=$START; c<=$END; c++ ))
do
        nohup ./cpuburn-in 3600 &
done
top

NB - the "3600" above is the number of seconds for the stress test to run.

Ending CPUBurn-in

To kill cpuburn-in manually before the stated completion time, use killall:

[root@raldisk ~]# killall cpuburn-in