QuickStart Guides

QuickStart Guides are documents that give a short overview of a topic, mostly comprised of links to more detailed documentation elsewhere on the web site. The available QuickStart Guides are listed below and in the menus to the left.

General QuickStart Guides

  • QuickStart Guide for New Users

    The QuickStart Guide for New Users is a collection of links to pages covering topics of interest to researchers who have just obtained a WestGrid account, such as choosing the most appropriate WestGrid system, logging in, working interactively, programming, using installed software and submitting batch jobs.

QuickStart Guides to WestGrid Facilities

The WestGrid computing facilities are comprised of a diverse set of resources. The QuickStart pages referenced below each give a description of one particular machine (or set of related machines), highlighting some of the features that distinguish it from other WestGrid resources and giving links to further information about how to work with that particular system. They are intended to be read by new WestGrid account holders and by current users considering whether to start using a part of WestGrid that is new to them. Although a few comments are given here about the roles of the various machines, please see the Computing Facilities page for an overview.

Storage sites

Computational resources for large, shared-memory jobs

Breezy and Hungabee are intended for serial programs that require more memory ( >45 GB, say) than available on other WestGrid machines or for shared-memory parallel programs (such as those based on OpenMP) that can effectively make use of more than 12 cores.  As noted in the respective QuickStart Guides, Breezy and Hungabee accounts are not set up automatically, but, require a special request.

Computational resources for distributed-memory parallel jobs

There are a number of clusters with Infiniband networking between nodes that support distributed-memory parallel processing.  Memory per core can be used as a guide to determine which cluster to try.  Grex has about 4 GB/core, Hermes and Nestor about 3 GB/core, Bugaboo, Checkers, Jasper, Parallel and Orcinus have 2 GB/core and Lattice has 1.5 GB/core.  Bugaboo is recommended for those who require large and/or high performance storage systems.  The Parallel cluster has 60 nodes with general purpose graphics processing units that can accelerate codes that are programmed to use them. As noted in the respective QuickStart Guides, Lattice and Parallel accounts are not set up automatically, but, require a special request.

Computational resources for serial processing

Serial jobs are accepted on a number of clusters, with Checkers and Hermes being the systems to which users with serial code are most often directed.  Bugaboo is also often used for serial processing due to the large variety of software installed there.  Breezy and Hungabee can be used for large-memory serial jobs. Glacier can be used for serial jobs, but, is an older cluster that is not generally recommended and requires a special request for access.

Pre-2009 machines

Glacier is available by special request for serial or low-demand parallel computing.  Although this older cluster still has significant computational capabilities, you might like to consider this only if you encounter congestion on the newer machines mentioned above.

Updated 2012-10-03.