In this thesis, we study different scheduling policies in service networks. In Chapter 2, we consider two service level (SL) measures in a two-server tandem queue system: the average sojourn time and the probability of long waits. We demonstrate that a family of Threshold Based Policies (TBP) can reduce the probability of long waits while maintaining sojourn times that are only slightly higher than those of a non-idling policy. In Chapter 3, we present a case study for improving the operations of a healthcare provider that has an open-shop queueing network. We propose an effective implementation of Dynamic Scheduling Policies (DSPs) and a generalized TBP to improve the SL in an open-shop queueing networks. Using a simulation model we demonstrate that an open-shop queueing network can be managed in a systematic fashion to deliver improved SL. In Chapter 4, we study the waiting time distribution of two different priority classes in an M/M/c queue with different service times. For the c=2 case, we provide closed form expression of the Generating Function (GF) of the number of low-priority jobs in the system, which can lead to the waiting time distribution. For c>2 case, we present an efficient numerical algorithm for deriving this GF. We discuss several insights gained from numerical results.
Both Chapter 2 and 3 were supervised by Professors Opher Baron, Oded Berman, and Dmitry Krass. Chapter 4 was supervised by Professors Opher Baron and Alan Scheller-Wolf.
| Identifer | oai:union.ndltd.org:TORONTO/oai:tspace.library.utoronto.ca:1807/65765 |
| Date | 01 September 2014 |
| Creators | Wang, Jianfu |
| Contributors | Baron, Opher, Berman, Oded, Krass, Dmitry |
| Source Sets | University of Toronto |
| Language | en_ca |
| Detected Language | English |
| Type | Thesis |
Page generated in 0.002 seconds