ObjectWeb Consortium
Search ObjectWeb Mail Archive: 

Advanced Search - Powered by Google


Mail Archive Home | proactive List | April 2006 Index

<--  Date Index     <--  Thread Index    

Proactive and task farming (newbie)


Hello all,

Let me first warn you that I am very new to ProActive (however I have worked
a lot with Jini) but after some research I think it is exactly what I need.
This mail is to check I can really do with ProActive what I want and to
clarify some questions.

Context:
* We have developed a simulation/modelling toolbox that requires distributed
execution of a native executable.  The executable takes a few numbers as
command line input and produces the results (numbers) on standard out. As of
yet the executable has no dependencies but these may become a requirement
eventually.
So you see, nothing fancy, just simple task farming.

* Available resources:
- a band new 256 node cluster administerd by Sun N1 Grid Engine 6, the front
node is only accessible through SSH
- Globus administered resources (with PBS queue), also with a SSH reachable
front node
- small, private clusters with a shared drive, accessible through ssh

I basically want to write a distributed backend to our modelling toolbox
that supports as many distributed resources as possible in the most
transparent way possible.  I think ProActive can help me here.

So as far as I understand the steps I have to take to get up and running are
(for SGE case):

1) write a ComputeActiveObject class with a method "double[]
compute(double[], pathToExe)"
2) make sure the Proactive scrips are installed on the cluster frontend 
3) Deploy my active object using a modified SSH_SGE_Example.xml. If I would
want to switch resources I would just have to change the XML descriptor?

Are there any other steps, problems,... I have to be aware of?

I already tried the Hello example with a modifed SSH_SGE_Example.xml but
that did not work.  I will try again to be sure and post the problem if Im
stuck.

* Questions
- Can proactive take care of staging the executable?  Or do I have to copy
it manually to the submit node or add a method to my service that can read a
binary stream.
- In the SGE case, dont I need to specify the name of the submission queue
somewhere?  Or does Proactive completely bypass qsub and the likes?
- what firewall requirements does ProActive have between the clustor nodes
themselves and between the cluster nodes and the frontend node?

* Suggestions
- How about adding a generic compute service to the list of examples?  I
think many people would benefit from having a service readily available that
can execute a given binary and return the results (ie. master-worker
pattern).

Many thanks in advance,

Dirk Gorissen
PhD Student
Antwerp University
Belgium
www.coms.ua.ac.be


-- 
Echte DSL-Flatrate dauerhaft für 0,- Euro*!
"Feel free" mit GMX DSL! http://www.gmx.net/de/go/dsl



<--  Date Index     <--  Thread Index    

Reply via email to:

Powered by MHonArc.

Copyright © 1999-2005, ObjectWeb Consortium | contact | webmaster.