Computing & Grid computing in COSM
Information technology is critical to conducting frontier research and it is an indispensable tool for many areas, including teaching. Particle physics, particularly accelerator-based research, has often motivated and enabled advances in information handling. An example is the World-wide web, developed at CERN where COSM researchers are now working on the ATLAS project.
For us, information technology includes a cluster of computers to simulate and analyze data from the ATLAS experiment, a cluster on the Open Science Grid, video conferencing tools, and an office applications network.
Grid computing is one of the tools that COSM brings to bear to help understand the world around us. In grid computing, many collaborating groups put computing resources (processors, storage, software) on the Internet in a sharing mode. Computing tasks are then submitted to the whole collection of resources and run where resources are free. No one group has preferential access to these resources, but there is an efficiency gain. A group may have 50 computers, but need none one day and 100 the next. The grid enables others to use the 50 computers the first day and return the favor the next. In the case of ATLAS (and the other OSG users), about 10,000 computers are in the OSG.
- Grid Computing Effort:
In 2002 COSM built the first grid-enabled cluster at Hampton University. The site is currently being upgraded, so it is off the air. - Cluster Computing Effort:
The Center also maintains a standalone cluster that is not grid-enabled. This resource is for Hampton University ATLAS researchers only. This cluster has been used to produce ATLAS simulation data that is made available to the larger ATLAS collaboration. - Microsoft Windows Environment:
COSM operates a Microsoft Windows 'organizational unit' in support of local group members. - Development of web/grid tools:
Through student projects, the Center has been involved in building a web portal to use grid resources.
All clusters run Scientific Linux with the Virtual Data Toolkit (VDT). The VDT provides the grid middleware tools including the condor job queues employed at the Hampton site.
The Hampton clusters produced an appreciable fraction of the simulated data for the ATLAS 2005 Data Challenge (DC2/Rome), and Hampton staff took 15% of all grid production monitoring shifts.
A real-time check of the state of the COSM grid site at Hampton University and the OSG catalog of sites:
- COSM's OSG grid site
- The OSG catalog of all sites, OSG Site Catalog.
| COSM cluster evolution | |||
|---|---|---|---|
 |
 |
 |
|
| The first HU cluster on the grid (HU-Huatlas): October 2002 - November 2003. | The previous and current HU clusters. Not shown: Webserver, Windows file server, GbE switch. | ||
Supported
by the