In many real world systems, servers are assigned to work in parallel to increase the system throughput and resource utilization. A class of multiple server queues, called pseudo parallel queues, was investigated. Exponential service times and interarrival times are assumed throughout. We are especially interested in establishing strategies for special customers to lower their sojourn time. Regular customers are assumed to join the shortest queue (JSQ) among the parallel queues. Decisions of the regular customers are made at their arrival time. The only information available to a special customer is the current queue lengths of the parallel queues at and after the time it arrives at the system. Two job scheduling strategies for the special customers are proposed based on a u22gathering informationu22 philosophy. They are studied for pseudo parallel queues. The thesis statement is that under certain circumstances, the average sojourn time of an individual job can be shortened by gathering information about the pseudo parallel queues before joining the queue. The purpose of this thesis is: (1) to provide a comprehensive literature review of multiple server queues, (2) to explore some mathematical properties of the parallel queues (by assuming JSQ), (3) to use the properties of the parallel queues that we found in analyzing new scheduling strategies. Paper copy at Leddy Library: Theses u26 Major Papers - Basement, West Bldg. / Call Number: Thesis1992 .P665. Source: Masters Abstracts International, Volume: 31-03, page: 1273. Co-Supervisors: Myron Hlynka; Henry Toews. Thesis (M.C.Sc.)--University of Windsor (Canada), 1991.
展开▼
