CAREER: Affinity and Scheduling for the Computational Grid

点击翻译按钮获取中文摘要

The design and implementation of a national computing systems grid has become areachable goal for both computer and computational scientists. In order to better use thissuite of resources, consisting of geographically disperse machines, instruments, and datastorage, better scheduling and resource management techniques are needed.This project concentrates on the issue of matching the requirements of an applicationto the capabilities of a resource or set of resources on the grid. This is often referred to as determining the affinity of an application for a resource. The research will formalizethe folklore currently used in order to increase the accessibility of the grid to thoseapplication scientists that have less experience than current users. Work willcontinue in the development of a language specification to describe both applicationsand resources as needed today, but with the extensibility needed for future applications aswell. This will then be used as an indexing structure into a case-based reasoning expertsystem.This project also has an extended educational portion to it. As a part of the effort toexpand Northwestern's Computer Science education in the systems area and a cross-schoolinitiative, we plan to offer a course in experimental systems using ideas in paralleldistributed scheduling as the basis for a larger view into practical systems,experimentation, and teamwork. This project-oriented class will be focused aroundteaching students about current parallel distributed scheduling efforts and applications.In addition, a new seminar series will be started featuring women researchers incomputer science to help encourage female students to continue with careers in computerscience.