Information Services

High Performance Computing

High Performance Computing

 

The Research Computing team at Queen's University provide a number of centrally managed High Performance Computing (HPC) systems (detailed below).  Researchers can avail of these systems in order to solve compute and data intensive problems. 

Kelvin Cluster

Kelvin is a scalable High Performance Computing (HPC) and Research Data Storage environment (this cluster is the Dell Cluster replacement).

Installation date: Nov 2015

  • Compute nodes run CENTOS operating system
  • HP hardware
  • 51 HP Apollo compute nodes with Intel Haswell generation processors
  • High memory nodes ranging from 128GB to 1TB of RAM
  • 500TB of lustre parallel file system for scratch
  • Each node has 20 cores
  • All compute nodes and storage are connected by an Infiniband fabric
  • Visualisation node (Nvidia K4200) for openGL applications. This must be requested.
  • Private cloud provision for applications such as Galaxy and CLC Bio.

Windows Cluster

The Windows cluster provides support for Windows applications and users who want to work in Microsoft/GUI environment.

Installation year: 2013

Dell hardware

  • Windows server 2012 operating system
  • 16 x Dell c6220 servers

=> 256 cores

  • Dual Intel Quad-core E5-2660 processors (64 GB memory )
  • 10GB interconnect
  • 30 terabytes of storage

HPC Service

HPC Service

Software

A wide range of scientific software is available for all the systems. New software can be installed on request providing a valid software license is available.

Support for researchers

The Research Support team provide help in a number of ways:

  • Training - a 90min introduction to Kelvin can be organized and delivered on request.
  • Help with job scripts
  • Software installation
  • Consultancy
  • Programming help

Accessing the service

Email hpc@qub.ac.uk  and please provide the following information:

  • Staff or student number
  • Name of supervisor
  • What software you intend to run
  • The number and size of jobs (in terms of cores) you plan to run
  • Any specific data storage requirements
  • In general, support requests that can be dealt with in less than an hour will be completed within 24hrs. For requests requiring more time it will be logged and dealt with within 1-2 weeks, dependent on how many support requests are current.

Notes:

  • New accounts are usually given out within 2 days of request.
  • The cluster runs the sge job scheduler software which ensures fair use of the resources.
  • Users are restricted to using no more than 64 cores at the same time.
  • Home folders have a quota of 50GB.
  • There is 500GB of scratch space available.
  • There is 1TB of local disk on each compute node.
  • If you have never used Linux before please allow time to get up to speed working with the command line.

Last Updated: June 2017