Clustering

[Austin Godber]

Eric Shubert wrote:
> Has anyone here implemented any clusters?
> Is any particular distro better or worse at clustering?
> Any pointers regarding clustering you'd like to share?
>   
In my experience, the base distro doesn't matter much unless the apps 
you plan on running have specific requirements.

I have run a couple hundred node batch cluster (not much if any IPC on 
the jobs).  During the time I managed that system we used PBS Pro 
(Commercial IIRC, there is OpenPBS as an alternative), torque (free, 
open source), maui, and Moab(commercial).

The torque was pretty straight forward.  The general concept is you have 
a cluster manager that tracks resources (resource manager, like torque) 
and schedules jobs (like pbs_sched that comes with torque or maui/moab) 
and then each node has an agent that gets managed by that cluster 
manager.  Typically there is shared storage between nodes but that is 
not required.

http://www.clusterresources.com/products/torque-resource-manager.php
http://www.clusterresources.com/torquedocs21/
http://www.clusterresources.com/torquedocs21/a.ltorquequickstart.shtml


There were some other options we looked at and briefly tested.  But I 
was most fond of torque/maui combination IIRC.  Moab seemed to be 
overkill for our needs at the time.

Austin