Testing Cluster

[Mr.C]

I'm new to this, and I am trying to find a program that I can run to test the performance and speed of the cluster. (i need numbers).
Can you please advise me.

Thanks, Mr. C

[Shporang]

You can check out the tutorial at http://pareto.uab.es/mcreel/PelicanHPC/Tutorial/PelicanTutorial.html

Section 3 is titled example software, and under that the first example is how to use the HPL benchmark to test your cluster.  If you need further help, I'm sure someone with a bit more experience would be willing to help you out.  If you need help tuning your hpldat file check Michael's post here: http://www.nabble.com/How-to-get-big-numbers-with-the-HPL-benchmark-to19685268.html

[Mr.C]

Ok, so I ran the simple pi calculating fortran program. The interesting thing is when I increase the number of nodes used to calculate pi to 10^-16 it takes longer then if i used a 1 or 2 nodes.
Im useing 13 dual cores.
It could be a slow connection between the nodes and the server. Please does anyone have any answers...thanks


Im trying to make a graph showing the amount of nodes used and the amount of time it took to finish the task.

[Michael Creel]

OK, I'm glad that you were able to get started. Problems that run very fast on a single core will not show a speedup running on multiple cores or on a cluster, since the communication overhead is large relative to the CPU time. For GNU Octave, there are some examples that use more CPU time and which will show a speedup. After having set up the cluster, open a terminal, type "octave" and then type "kernel_example". You can then edit kernel_example.m to specify the size of the data (larger uses more time) and the number of compute nodes to use in computations. I have a few journal articles that discuss this and other examples in more detail, see
http://econpapers.repec.org/article/kapcompec/v_3A32_3Ay_3A2008_3Ai_3A4_3Ap_3A343-352.htm
http://econpapers.repec.org/article/jaejapmet/v_3A22_3Ay_3A2007_3Ai_3A1_3Ap_3A215-223.htm
http://econpapers.repec.org/article/kapcompec/v_3A26_3Ay_3A2005_3Ai_3A2_3Ap_3A107-128.htm
Cheers, Michael

[Vall]

Try runing a few benchmark programs such as LINPACK...