The Computational Grid, as described in The Grid: Blueprint for a New Computing Infrastructure and demonstrated by many proof-of-concept applications, promises to connect computers, databases, and people in a network to solve problems in scientific research and other diverse fields. However, the complexity, unreliability, and overhead of low-level operations in today's systems obscure the Grid's potential. The Virtual Grid Application Development Software (VGrADS) project attacks a fundamental part of this problem - how to more effectively program these highly complex and dynamic systems. It develops software systems that (attempt to) simplify and accelerate the development of Grid applications and services while delivering high levels of performance and resource efficiency. This improved usability helps expand the community of Grid users and developers. In the process, VGrADS contributes to both the theory and practice of distributed computation.

VGrADS is based on the earlier GrADS project.  That project explored the scientific and technical problems of grid application development and performance tuning for real applications. In particular, it developed a framework to incorporate rescheduling based on performance monitoring of dynamically-changing resources.

VGrADS is a five-year project, started in October 2003, that extends GrADS in several respects:

• Virtual Grids (VGs) provide an additional level of abstraction, allowing separation of concerns between levels of the system.
  
• VGs enable the system to present additional resource capabilities to applications. In particular, VGs will be key to controlling fault tolerance and allowing scalable scheduling in VGrADS applications.
• VGs also enable support for a "slot" abstraction, representing the times when resources are available. They provide the leverage for both scheduling availability of those slots, and mapping computations into the provided slots.
  
• Advanced prediction strategies (including predictions of queue delays and computational performance) and scheduling strategies (which may soon include market-based scheduling) allow more-advanced applications to run with VGrADS support.
  

VGrADS is pursuing this agenda by collaborating with leading scientific applications to elicit key challenges, validate results, and disseminate technology. VGrADS researchers at UCSB developed one of the world-leading boolean satisfiability programs (GridSAT) some time ago.  We are now working with the LEAD project on atmospheric modeling. 

VGrADS distributes some software that it creates in open-source form for the research community. This is particularly true of sub-technologies, such as the prediction strategies, can be packaged as modules for inclusion in general Grid computing packages. VGrADS will also build on its PIs' past successes in human resource development by using existing programs to attract and retain women and minorities in computational science.

For more information about VGrADS, please browse the folders listed on the left:

• Research presents many of our technical thrusts, particularly
• Execution Systems,
• Programming Tools, and
• Applications
• Education, Outreach, & Training presents our efforts in reaching beyond today's research community, including
• Support for diversity-oriented conferences (Richard Tapia conference and Grace Hopper conference), and
  
• Research Experiences for Undergraduates
• Participants lists individuals involved with the project
• Publications lists our papers and technical reports (many available on-line)
• Formal Documents provides various proposals and annual reports
• Presentations provides some significant presentations made at conferences
  

The Virtual Grid Application Development Software (VGrADS) project is supported by the National Science Foundation (NSF) under award #0331645. It is a collaboration of Rice University, University of California at San Diego, University of California at Santa Barbara,  University of Houston, University of North Carolina at Chapel Hill, University of Southern California Information Sciences Institute, and University of Tennessee at Knoxville.