Overview of the PPPL Cluster

The PPPL Cluster consists of several sub-clusters. The kestrel subcluster has a large number of nodes with dualcore Opteron CPUs and a 1Gb ethernet network. The kite subcluster has many dual CPUs systems connected via gigabit or infiniband. The sque cluster has many systems available for single CPU jobs. The ganesh cluster has several systems with 4 cpus and large memory size (32GB of RAM per system).

Users do not login directly to the cluster. Instead, there is a pool of systems called portal.pppl.gov, into which users login, stage their jobs, and submit their jobs.

Users log into portal.pppl.gov using their PPPL Unix accounts. To get an account, send email to lscimeca@pppl.gov.

On portal, users create, in their home directories, scripts that will run their program. (A sample batch script is available on this web site). These scripts are submitted to a batch job queue and run on one or more batch nodes. Results of the run are written back to the user's home directory.

The PPPL cluster also offers interactive processing services. Users can use the 'use' command on portal to reserve a set of systems for their exclusive use.

Home directories are served by UNIX fileservers which are responsible for storage and backup of files.