Signed Off

ARC PROJECT MANAGEMENT

PSU Clusters

last updated on Thu, 10/23/2008 - 10:15

Contents

General Information

Description

A high performance computing solution to be used for parallel processing using latest standard in this technology such as MPICH, openMPI, etc. Currently we operate two separate clusters Rocks and Gravel.

Purpose and goal of the project

Provide additional computing ressources to the scientific community at PSU by providing HPC services and support.

Client contact

Personnel

Martin Lukac
lukacm@pdx.edu

Scope and Scale

Customer

Researchers in General and Funded Projects

Users/ audience

Researchers in General and Funded Projects

Scope

University Wide

Partners

Architecture/Design

Project Specification

Rocks is a 48 queues (dedicated CPU) AMD Opteron 64bit Linux build (Rocks Cluster) with each compute queues has 2GB dedicated SDRAM.

Rocks cluster consists from:

- Headnode with 2 AMD Opteron 264 and 2GB RAM

- NFS shared partition with user accounts (storage on demande)

- 20 computes each with 2 AMD Opteron 264 with 4GB RAM

- Gigabit Network Connection on the cluster intranet as well as on the access internet

Gravel, representing our latest addition, is a Dell cluster having a 2950 headnode and so far 23 compute nodes Dell 1955. Gravel has 184 parallel queues (dedicated CPU's) each with 1GB dedicated memory. Gravel is built using Platform's Open Custer Stack (OCS)

Gravel cluster consists from:

- Headnode with 4 Cores and 4GB RAM

- Local 3TB of storage space

- 23 computes each with 8 Cores with 8GB RAM

- Gigabit Network Connection on the cluster intranet as well as on the access internet

Requirements

Consult Manuals and Help page

Project home and resource storage

The main control access is ssh using PDX credentials to host:gravel.research.pdx.edu.

Gravel cluster is also accessible through the web at http://gravel.research.pdx.edu or alternatively https://gravel.research.pdx.edu.

The main control access is ssh using PDX credentials to host:rocks.research.pdx.edu.

Rocks cluster is accessible through the web at https://rocks.research.pdx.edu 

Related Projects

Project status

Timeline

Cluster has been upgraded to latest software and is 100% functional. The cluster will be soon upgraded by an Intel based cluster (out of the box) from Dell.