[StarCluster] Cluster size

[Butson, Christopher]

I'm looking at StarCluster for use in simulations using PBS or SGE. This seems like a great tool that could save a lot of development time! The goal is to setup a cluster that can allocate hundreds of nodes (I think these are elastic compute units (ECU) in the Amazon nomenclature), up to a maximum of around 1000. I've read the documentation and have successfully run a small test cluster but I'm still having trouble understanding the relationship between the following:

  1.  NODE_INSTANCE_TYPE: I understand that this is the Amazon instance type, but I don't see how this interacts with load balancing or the default template.
  2.  DEFAULT_TEMPLATE=smallcluster: Where is this template defined? I note that another is largecluster, but I'm not sure the difference between the two.

  3.  Elastic load balancing: This feature makes sense, but is the maximum number of nodes a function of selections I make for the default template or instance type?

Thanks for any insights on how to configure the cluster,
Chris

Christopher R. Butson, Ph.D.
Associate Professor
Biotechnology & Bioengineering Center
Departments of Neurology & Neurosurgery
Medical College of Wisconsin
(414) 955-2678
cbutson at mcw.edu