An Approach for Fair Scheduling with Efficient Methods for Managing Advance Reservations in Grids
Scheduling with advance reservations is an important class of problems related to resource management in distributed systems. Advance reservation is usually used to run parallel jobs that require allocation of multiple distributed resources at once or to execute interactive applications. Advance reservation becomes particularly important if a user must pay for the utilization of resources. In that case advance reservation capability is crucial to guarantee requested quality of service in terms of time and cost. Nevertheless, this type of scheduling requires efficient methods to manage reservations. To this end, we propose efficient data structures and algorithms to search for resource availability time slots. Furthermore, we define conditions under which particular data structures and algorithms should be applied. Furthermore, we adapt these proposals for a multicriteria resource management problem in which we find fair schedules for all users that express their preferences using criteria related to time and cost. Efficiency of proposed methods is demonstrated using the GSSIM simulator, a tool for Grid scheduling experiments described in the presentation.