Jeremy Enos wrote:
Jose-The cluster has 8 client nodes, each one has a single PIII 800 MHz Processor, with 128 MB of main memory and a 40 GB HDD. By now it's a little outdated. We intend to upgrade it early next year.
It depends on the nature of the jobs you're running. Typically, if you're running the same type of jobs on a shared node, then they just split the bottleneck and run less efficiently. In most cases, I would also recommend setting the number of virtual procs = physical procs for the PBS config. How many physical procs/box does your cluster have?
Well, we have a little of everything. Most of them are heavy computing tasks (i.e. Neural Nets training, Neuro-Fuzzy Systems modelling, Genetic Algorithms executions); some others have large memory requirements (feature selection processes). Finally, I/O isn't being a very big issue by now.Also, are the jobs you are running CPU, memory, or I/O bottlenecked?
So it's hard to find a perfect, common solution for everyone!!
--
Jose Manuel Benitez Sanchez e-mail: [EMAIL PROTECTED]
http://decsai.ugr.es/~jmbs
Dpto. de Ciencias de la Computacion e Inteligencia Artificial
E.T.S. Ingenieria Informatica
Universidad de Granada Tel. +34 - 958 - 24 61 43
18071 - GRANADA (Spain) Fax: +34 - 958 - 24 33 17
-------------------------------------------------------
This sf.net email is sponsored by:
With Great Power, Comes Great Responsibility Learn to use your power at OSDN's High Performance Computing Channel
http://hpc.devchannel.org/
_______________________________________________
Oscar-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/oscar-users
