skip to main content search contact us
CREW logo CREW logo University of Michigan Seal
home about CREW research people publications seminars
research - current projects
make font size smaller font size make font size larger
Learning from TeraGrid
Learning from TeraGrid
Tom Finholt, Ann Zimmerman,
Magia G. Krause
National Science Foundation

A key goal of cyberinfrastructure (CI) is to bring together distributed resources such as computational tools and services, instruments, data, and people to create virtual organizations that will accelerate the pace of scientific and engineering discoveries by reducing constraints of distance and time.  Even though CI efforts are in their infancy, there are already examples of exciting findings, life-saving systems, and unique learning environments enabled by high performance computing, networking, and information tools and resources.

In contrast to these inspiring successes, past experiences with large-scale information infrastructures show that achieving such goals is not easy. The final report from a National Science Foundation (NSF) workshop on CI for the social and behavioral sciences alluded to the causes for some of these failures:

"…implementing CI is not easy. The high failure rate of large information technology projects and their tendency to overrun budgets indicate that we must pay attention to implementation. Strong empirical evidence demonstrates that the sources of failures and cost overruns typically are organizational and managerial rather than technological. The higher level of complexity of Cyberinfrastructure demands a commensurably higher level of organizational and managerial knowledge and expertise" (Berman & Brady, 2005, p. 28 ).

TeraGrid is a national, comprehensive, and distributed infrastructure that integrates high-performance computing resources at nine resource provider facilities. Following a 5-year construction phase, TeraGrid became operational in late 2004. TeraGrid's resources include more than 102 teraflops of computing capability and greater than 15 petabytes of online and archival data storage. High-performance networks provide rapid access and retrieval to data. TeraGrid supports a variety of use cases ranging from exploiting a single TeraGrid resource to combining resources across sites. The latter capability opens up new possibilities for conducting scientific work.

TeraGrid is a key part of the nation's long-term cyberinfrastructure plan. Like other CI efforts, TeraGrid is an administratively complex partnership among multiple, geographically-distributed institutions. There is some risk that CI investments will not be optimized unless these projects have access to techniques that will help them identify, monitor and address potential problems. In late spring 2006, the NSF awarded a one-year grant to the University of Michigan ’s School of Information (UM-SI) to conduct an external evaluation of TeraGrid. The primary goals of the evaluation are a) to provide specific information to TeraGrid managers that will increase the likelihood of TeraGrid success; and b) to give NSF leaders and policy makers general data that will assist them in making strategic decisions about future directions for cyberinfrastructure. In order to accomplish these objectives, the UM-SI evaluation study is assessing four aspects of the TeraGrid project:

1) progress in meeting user requirements;
2) impact on research practice and outcomes;
3) quality and content of TeraGrid education, outreach, and training efforts; and
4) satisfaction among TeraGrid partners.

Project researchers are employing multiple research methods in the TeraGrid study including interviews, surveys, and analysis of documents and publications.  

Related Materials / Links
- TeraGrid User Workshop Final Report ( PDF 129KB)
copyright disclaimer