Choosing Optimal Queue Weights for Switched Processing Systems

• Main Authors: Wen-Chih Cheng, 鄭文智
• Other Authors: 洪英超
• Format: Others
• Language: en_US
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/23791116824635158534

碩士 === 國立中央大學 === 統計研究所 === 95 === Switched Processing Systems (SPS) represent crucial models for many applications in communication, computer, and manufacturing networks. They are characterized by flexible, independent service capabilities and multiple classes of job flows. Over the years, a fairly rich literature has been developed for maximizing the system’s throughput and at the same time constructing scheduling policies that maintain a certain level of system stability. Recently, research has been shifted to improving the quality of service (QoS) performance with respect to the performance metrics such as delay and backlog. In this study, we investigate a class of throughput maximizing scheduling policies called MaxProduct policies. The goal is to place the optimal weight on each queue so that the delay performance under the MaxProduct policies can be significantly improved. The proposed approach is motivated by the well known Shortest Processing Time First (SPTF) rule. For systems where the input traffic does not change, a one-dimensional search procedure for finding the optimal queue weights is introduced. For systems where the input traffic statistics might change, a more flexible and practical search procedure is suggested. The simulation results reveal that our proposed methods can substantially improve the average system delay and the 95th percentile of delays for various types of input traffic.