Introduction to Research Computing on Palmetto Cluster: Glossary

Key Points

Introduction
  • Palmetto is a very powerful high-performance computing cluster

Accessing the Palmetto Cluster
  • Palmetto can be accessed in a web browser through Open On Demand interface

  • This interface can be used to transfer the files and to start a terminal

The structure of the Palmetto Cluster
  • Palmetto contains more than 2000 interconnected compute nodes

  • a phase is a group of compute nodes that have the same architecture (CPUs, RAM, GPUs)

  • a specialized login node runs the SSH server

Storage on Palmetto
  • users get 100 GB of backed-up storage in their home directories

  • they also have access to more than 2 PB of scratch storage

  • scratch storage is not backed up, and files left unused for 4 months are deleted

Running an interactive job on Palmetto
  • whatsfree shows the current Palmetto usage

  • qsub sends a request for a compute node to the scheduler

  • software available on Palmetto is organized into modules according to version

  • modules need to be loaded before use

Running a batch job
  • batch jobs don’t require interaction with the user and run on the compute nodes on the background

  • to submit a batch job, users need to provide a PBS script which is passed to the scheduler

  • jobs are assigned to queues, according to the amount of requested resources

  • different queues have different limits on the walltime and the number of parallel jobs

Web-based access to the Palmetto Cluster
Transferring files to and from Palmetto
  • Windows and Mac users can use CyberDuck for file transfer

Accessing the Palmetto Cluster via Terminal
  • Palmetto can be accessed by an SSh (secure shell) client

  • Windows user can use MobaXTerm application

  • Mac users can use the Terminal application

Glossary

FIXME