How do I use the super queue?

How do I access the super queue?
Access to the quad-processor and large memory nodes

13-Mar-2008

Introduction

The RTC has a special queue called the super queue which is assigned to nodes within the cluster that have special resources available that the remainder of the RTC nodes do not have. Specifically, these nodes might have more RAM, faster processors, or be quad-processor nodes instead of dual-processor. There are a total of 38 processors in the super queue. The remainder of this document will describe the super queue in more detail.

Hardware Inventory

The super queue has the following hardware and associated PBS properties:

Number of Nodes	Processors per node (ppn)	RAM per node	Clock speed	PBS features
4	4	16GB	900MHz	quad
3	4	16GB	1.3GHz	quad, fast
2	4	32GB	1.3GHz	quad, largemem, fast
1	2	16GB	1.3GHz	dual, fast

Accessing Specific Nodes

If you need to run jobs on specific super queue nodes, you will do so by specifying one of the PBS features listed in the table above in your job script. Here are several examples.

Accessing a 32GB RAM node

#PBS -q super
#PBS -l nodes=1:ppn=4:largemem,walltime=04:00:00

Note: you can request less than 4 ppn and still be able to run on this node, but you will not have access to all four processors.

Accessing all nodes with 1.3GHz clock speed regardless of RAM

#PBS -q super
#PBS -l nodes=6:ppn=2:fast,walltime=04:00:00

Note: if you want to access all 1.3GHz nodes, then the largest ppn you can ask for is 2 since one of the nodes is a dual-processor node.
Accessing all quad processor nodes, regardless of RAM or clock speed

#PBS -q super
#PBS -l nodes=9:ppn=4:quad,walltime=04:00:00

Note: you can request less than 4 ppn and still be able to run on the quad-processor nodes, but you will not have access to all four processors.