版权所有:内蒙古大学图书馆 技术提供:维普资讯• 智图
内蒙古自治区呼和浩特市赛罕区大学西街235号 邮编: 010021
作者机构:Wuhan Univ Technol Dept Comp Sci Wuhan 430063 Peoples R China
出 版 物:《INFORMATION SCIENCES》 (信息科学)
年 卷 期:2007年第177卷第15期
页 面:3038-3059页
核心收录:
学科分类:12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 08[工学] 0812[工学-计算机科学与技术(可授工学、理学学位)]
基 金:Wuhan Younger Dawning Foundation, (20045006071-15) National Natural Science Foundation of China, NSFC, (60402028, 60672137) Specialized Research Fund for the Doctoral Program of Higher Education of China, SRFDP, (20060497015)
主 题:joint QoS optimization computational grid layered architecture decomposition
摘 要:Many existing grid resource allocation and scheduling algorithms mainly focus on isolated layers of the grid architecture. The inflexibility of the strict layering structure results in an inefficient utilization of the grid resources. This paper takes a system view of the computational grid and aims to jointly optimize global QoS by adopting cross-layer design. Cross-layer design is based on information exchange and joint optimization among multiple grid layers. Parameters from different layers are provided to a cross-layer optimizer, which selects the values of the layer specific parameters maximizing joint global QoS. The objective of the paper is to jointly optimize the parameters of all layers in a decentralized optimization problem and decompose joint QoS optimization into three sub problems at fabric layer, collective layer and application layer. In simulation part, we compare the performance of the global joint QoS optimization approach with application layer local optimization and resource layer local optimization approach, respectively. (c) 2007 Elsevier Inc. All rights reserved.