Quantcast

How can i find out which are processes are being sent to which node?

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate
star

How can i find out which are processes are being sent to which node?

parikh27
Hi Team, I am planning to do a demonstration on clustering in my college. I am very new to Clustering, and also linux as for that matter. I have been able to setup a cluster of 2 nodes (pxe boot ) and 1 acting as the primary machine ( Total 3 ), using PelicanHPC on my Oracle VirtualBox. I have been able to run the kernel_Setup, and other examples from GNU octave and also using MPIRUN command in the terminal. The question is: is there anyway i can show a list of processes which are being transferred to different nodes? or is there any software that can display some sort of 'task manager' like we have in windows on all the nodes? Basically, my primary objective is demonstrate and prove to my audience that the processes are being transferred to different nodes and are running parallel. all help would be appreciated. Thanks
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate
star

Re: How can i find out which are processes are being sent to which node?

Michael Creel
Administrator
My favorite tool for this is ksysguard. To use it
1. open a terminal, and type "ksysguard"
2. in ksysguard, create a worksheet with the same number of panels as hosts in your cluster. For example, I have 4 hosts, so I create a worksheet with 2 rows and 2 columns.
3. On the left, click on localhost, and drag "process controller" into one of the blank cells of the worksheet
4. In the "File" menu, click on "Connect host...", and type in the IP address of one of the compute nodes (for example 10.11.12.2)
5. on the L, the IP of the compute node will appear, click on it, and then drag "process controller" into another blank cell
6. Repeat 4 and 5 for all compute nodes.

You will end up with a worksheet that looks like this:
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate
star

Re: How can i find out which are processes are being sent to which node?

parikh27
Hi Michael,

Thanks a bunch, you have been great help. This looks perfect... Cant wait to demonstrate it.

Keep UP the good work !

Cheers Mate !


Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate
star

Re: How can i find out which are processes are being sent to which node?

A. Jorge Garcia
In reply to this post by Michael Creel
Wow, this looks a lot like mtop in openMOSIX.

Thanx,
A. Jorge Garcia
Teacher and Professor
Applied Math and Computer Sci
http://shadowfaxrant.blogspot.com
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate
star

Please tell me some real time applications were I can show the use of the cluster

anilkumarpanda
In reply to this post by Michael Creel
Hi Michael,
I want to demonstrate cluster building in my college, can u please suggest me some applications/softwares tht I can use to demomstrate the power generated by the cluster.
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate
star

Re: Please tell me some real time applications were I can show the use of the cluster

Michael Creel
Administrator
kernel_example.m for octave will eat up a lot of cpu cycles, and scales well. From the help info:


octave:1> help kernel_example
`kernel_example' is a function from the file /home/user/Econometrics/MyOctaveFiles/Econometrics/Kernel/kernel_example.m

 kernel_example: examples of how to use kernel density and regression functions

 usage: kernel_example(n, use_mpi, show_plot);
 INPUTS:
 n: sample size (default n = 500)
 use_mpi: (default false) Use MPI or not. To use MPI, you should execute
        mpirun -np X octave --eval "kernel_example(1000, true);"
        where X>1 is an integer number of ranks to use.
        requires open_mpi extensions from Octave Forge
 show_plot (default true) display plot?


So, if you have a large cluster, set n to 20000 or so, and you'll see that it scales more or less linearly in X as long as X <= total cores +1.
Loading...