Campus Computing News

UNT Installing a new High Performance Computing Cluster

By Dr. Philip Baczewski, Director of  Academic Computing and User Services

The CITC's Academic Computing and User Services will soon take delivery of the largest computing system ever installed on the UNT campus. The computing system may be among the most powerful in the world and will bring to UNT the capability of performing large-scale computing tasks that are required by many areas of scientific and engineering research being done today.

What is a high performance computing cluster?

A high performance computing (HPC) cluster is a set of individual servers which can work together, combining their individual resources, in order to run large-scale computation programs.  Academic Computing is not new to HPC cluster management. In 2001, ACUS staff assembled a modest 8-node cluster and over the intervening years has upgraded resources to take advantage of new developments in processor and networking technology. This new HPC acquisition, however, represents a quantum leap in the cluster resources that are managed centrally and potentially available to all researchers on campus. The new cluster will provide 10 times more processing power than what ACUS can currently offer.

The HPC cluster purchase was made possible through funding support from the offices of the Vice President for Research and Economic Development and the Vice President for Finance and Administration.  A committee of faculty researchers helped determine the specifications for the cluster acquisition and Dell was chosen as the vendor to supply the cluster equipment.   

Cluster configuration and installation

As configured, the HPC cluster will consist of 224*  individual Dell R410 servers (called nodes), each with two quad-core Intel X5550 (Nehalem) processors running at 2.66GHz (with performance levels up to about 3 GHz in “turbo” mode).  This number of CPUs  provides a total of 1792 processor cores, each of which can work independently to run all or part of a computer program. All nodes are interconnected by a high-speed Infiniband network which allows a set of nodes running a single program in parallel execution to share data and items held in memory at a very high level of efficiency.

Three kinds of node configurations are included within the cluster:

·         32 large-memory nodes with 64GB of RAM and 2TB of disk space;

·         64 medium-memory nodes with 32GB of RAM and 2TB of disk space;

·         128 small-memory nodes with 24GB of RAM and 1TB of disk space.

Large and medium-memory sized nodes will be useful for running individual programs which require resources to support manipulation of large data sets in memory.  The “small-memory”-sized nodes will be available for individual (serial) execution of programs as well as the parallel execution of a single program where some of the computation is  done on  each of the nodes.  The entire cluster will have the ability to act as a single very-large parallel processing system if a job is large enough to require it.

Another aspect to high-performance computing is the need to generate or analyze large amounts of data. In order to accommodate researchers' use of large data sets, the HPC cluster will be attached to 200TB of high-performance disk storage.  This storage system, manufactured by Data Direct Networks (DDN), will allow programs running on the cluster to read or write extremely large amounts of data at access speeds that won't slow down the rate at which the program can execute.

In order to accommodate the new HPC system, the CITC machine room on the UNT campus has undergone upgrades to its power and cooling capacity. The HPC cluster will have a dedicated cooling system and it and other computers in the machine room will be protected by a newly-expanded uninterruptible power supply (UPS) system and a diesel-powered generator to provide electricity in the event of a loss of City of Denton power.

Installation of the cluster will begin at the start of August and the cluster will be made available to researchers during the Fall semester of 2009.  Acquisition of this new computing resource represents one part of UNT's effort to expand research activity in the areas of math, sciences, and engineering.  Academic Computing and User Services is excited to be able to support such a significant new service for the research community at UNT.

 

Originally published, July 2009 -- Please note that information published in Benchmarks Online is likely to degrade over time, especially links to various Websites. To make sure you have the most current information on a specific topic, it may be best to search the UNT Website - http://www.unt.edu . You can also search Benchmarks Online - http://www.unt.edu/benchmarks/archives/back.htm as well as consult the UNT Helpdesk - http://www.unt.edu/helpdesk/ Questions and comments should be directed to benchmarks@unt.edu

Return to top