Computational Cluster Programs

Computing on the Hydro Cluster

The preferred way to access the Hydro Cluster is through the UCLA Grid Portal at http://grid.ucla.edu. The grid portal makes the Hydro Cluster accessible from any web browser. You can submit batch jobs, run interactive GUI applications and get an xterm from the UCLA Grid Portal. If you are going to use the Hydro Cluster from the UCLA Grid Portal, to use batch or interactive services (except for xterm) you will not need to know the information presented here.

Role of the Head Node

Use the Hydro head node for logging into the Hydro Cluster.

You may do the following when logged onto a Hydro Cluster head node:

File management
Editing
Compile programs (from source code)
Test parallel jobs. Run interactive applications. However note that, any program that requires more than 1 hour to complete and which does not require considerable human interaction must be submitted to a batch job queue.

Batch Jobs

Most of the nodes on the Hydro Cluster are dedicated to running batch jobs. All non-interactive programs that require more than 1 hour to complete must be submitted as batch jobs.

The Hydro Cluster has 28 compute nodes with 2 GB memory per processor, 20 of which have InfiniBand interconnects. You can run at most a 40 processor job.