Command Line Executions

Lion0928 · 01-21-2005, 10:43 PM

Hello, im pretty new to linux and am doing a project on building a cluster. Ive already built the cluster and have some software to test it with. However, I want to run a program X number of times at the same time so they will all begin executing and split up the number of tasks among the computers. So, my question is Does anyone know how to execute one binary that I have compiled X number of times?

What I was thinking was that it could be done like:
./testapp -someargument X

where someargument will make that progran execute X number of times begining them all at the same time.

Thanks,
Taylor

homey · 01-21-2005, 11:50 PM

Maybe a little loop? Put the command you want to run in place of where I have the ls

Code:

#!/bin/bash

# set this to the number of times you want to run
num=5

for (( i=1; i<=$num; ++i )); do
   ls &
done

Dark_Helmet · 01-22-2005, 12:37 AM

I think homey has the right idea (in fact, I think his script is only missing one character):

Code:

#!/bin/bash

# set this to the number of times you want to run
num=5

for (( i=1; i<$num; i++ )); do
   ls &
done

Without the '&' the script will launch all of the ls commands one-at-a-time. That won't tax a cluster much. Running them in the background will allow the script to shoot off each ls without waiting for it to finish.

Quote:

Originally posted by Lion0928
What I was thinking was that it could be done like:
./testapp -someargument X

Short answer: no, there's no global command-line argument to cause a process to spawn X times.

Long answer: It depends on the software. You say you compiled it. I don't know if that means you wrote it yourself, or just downloaded a tarball from a site. I'm assuming you downloaded it, because you'd probably know whether the program supports that kind of command-line argument if you wrote it

So, you'll need to check the software's documentation. That's not a fequently needed option, so I would guess you'll have to use a script like homey suggested. You'll also need to know how a job gets placed into the cluster's job pool. I don't have any experience with clusters, but I would imagine there is some sort of special command to submit a job into the cluster (as opposed to just running a command on the local machine/node you're using). Again, that should be in documentation somewhere, and you'd have to modify the script accordingly.

homey · 01-22-2005, 06:43 AM

Thanks Dark_Helmet,

I also changed the loop abit.

for (( i=1; i<=$num; ++i )); do