Research


I am currently researching as a Ph.D. candidate in Computer Science at SUNY Binghamton with my advisor Michael Lewis, specializing in grid computing. My current research direction involves researching grid resource availability prediction as well as scheduling in a heterogeneous grid environment. More specifically, I'm examining how prediction of the availabilities of resources that comprise the grid can be utilized to increase the performance of the grid in executing applications. The key novelty in my contribution is in studying the different causes of unavailability that grid resources experience and in attempting to predict those unavailability types. The initial work highlighted below concerns analyzing traces gathered from actual grid resources for their availability data as well as making an initial assessment on the predictability of those availability states and the scheduling benefit that is possible based on this. My current work includes developing more accurate and sophisticated algorithms for predicting resource availability in addition to studying how to utilize those predictions to make more effective scheduling decisions.



Publications


Brent Rood and Michael J. Lewis, "Grid Resource Availability Prediction-Based Scheduling and Task Replication," Journal of Grid Computing, 2009. [PDF]

Brent Rood and Michael J. Lewis, "Resource Availability Prediction for Improved Grid Scheduling," to appear in Workshop on Advances in High-Performance E-Science Middleware and Applications (in conjunction with e-Science 2008), Indianapolis, Indiana, December 7 - 12, 2008. [PDF] [Presentation Video]

Brent Rood and Michael J. Lewis, "Scheduling on the Grid via Multi-State Resource Availability Prediction," Grid 2008: The 9th IEEE/ACM International Conference on Grid Computing, Tsukuba, Japan, September 29 - October 1, 2008. [PDF]

Brent Rood and Michael J. Lewis, "Multi-State Grid Resource Availability Characterization," Grid 2007: The 8th IEEE/ACM International Conference on Grid Computing, Austin, Texas, September 17 - 19, 2007. [PDF]

Brent Rood, John Paul Walters, Vipin Chaudhary, and Michael J. Lewis, "Failure Prediction and Scalable Checkpointing for Reliable Large-Scale Grid Computing," HPDC-16: The 16th IEEE International Symposium on High Performance Distributed Computing (Hot Topics Session), Monterey, California, June 27 - 29, 2007. [PDF]


Why Distributed Computing?


Grid computing and it's cousin, volunteer computing have many direct positive consequences for the community at large. One of the key benefits is in providing scientists and researchers in virtually any field with access to a large amount computing power for relatively cheap or at no cost at all. This computing power can be utilized to help solve large problems as evidenced in several current projects. These projects aim at utilizing computational power that otherwise would have been wasted. Instead it is put to use for the benefit of mankind. The problems solved in this way may not have been feasible without access to such a large amount of computational power and as such, these grids can be an invaluable tool. Research that enables grids to become more efficient, easier to setup or in the most practical terms, more easily utilized by the end users allows the technology to be more powerful and accessible thus enabling widespread use and benefit to the scientific community and by extension, the community at large.

Examples of such projects include:


Related Work


Below I've highlighted some further related material whose availability is dependent on the specific paper in question. Good resources for exploring these and other papers include Google Scholar and CiteSeer. Adobe Acrobat Reader is required for viewing.

Availability Characterization

Prediction