Multiple QoS modeling and algorithm in grid system is considered. Grid QoS requirements can be formulated as a utility function for each task as a weighted sum of its each dimensional QoS utility functions. Multiple QoS constraint resource scheduling optimization in computational grid is distributed to two subproblems: optimization of grid user and grid resource provider. Grid QoS scheduling can be achieved by solving sub problems via an iterative algorithm.
研究运用投标模型对网格资源进行合理分配.建议了网格任务A gen t的投标方式,并提出了任务A gen t的爬坡式投标算法,任务A gen t可以通过该算法搜索效用函数获得其期望效用最大化.网格任务A gen t的效用函数是基于期望的效用函数.如果任务A gen t在限定的时间内完成所有的任务并且将剩余的费用归还,用户将表示满意,否则,用户表示不满意.在仿真实验中,着重研究了网格任务A gen t效用函数与作业响应时间之间的关系,并对基于效用函数的经济调度方法与传统的先来先服务(FCFS)方法进行了比较分析.
A wireless MANET is a self-configurable network in which mobile,battery-powered devices dynamically create and change the network topology without relying on any intrastructure or administrative support.It offers unique benefits for certain environments, but there are still several open issues to be resolved before realizing these benefits,including QoS routing,multicasting,and mobility management. Specifically,an efficient QoS multicast routing protocol is needed.This paper pre- sents a hierarchical QoS multicast routing protocol(QMRP)for mobile ad hoc networks.It can provide QoS-sensitive routes in a scalable and flexible way,in the network environment with mobility.In the proposed QMRP scheme,each local node only needs to maintain local multicast routing information and/or summary infor- mation of other clusters(or domains)but does not require any global ad hoc net- work states to be maintained.The QMRP also allows any ad hoc group member to join/leave the multicast group dynamically and supports multiple QoS constraints. This paper presents formal description and main procedures for realizing routing decision process of the QMRP and the proof of correctness and complexity analy- sis of the protocol.The performance measures of QMRP are evaluated using simulation.The simulation results verify the efficiency and availability of QMRP.
设计了网格资源发现过程和网格资源经济调度过程以实现网格的资源管理.网格资源发现过程提供了一种基本的机制,它可使网格请求A gen t能够发现所需要的网格资源.网格资源经济调度过程主要用于管理网格任务A gen t通过竞标向网格资源A gen t购买资源,以完成计算任务.文中对网格资源发现过程和网格资源经济调度过程进行了形式描述,并描述了所使用的主要数据结构.
提出了一种基于效用函数优化的网格资源分配策略.试图利用效用函数和竞标函数对网格资源进行合理分配,采用拉格朗日方法对网格任务A gen t效用函数进行优化,从而可使网格任务A gen t在能估计资源节点拥塞度,并能完成其所有任务的情况下,产生一个合理的费用预算.对网格任务A gen t的竞标函数的特征进行了研究分析,研究结果表明如果网络状态不变,网格任务A gen t不能通过改变投标来获取效益.这种恒定性在任务A gen t不知晓竞标结果的情况下,可使它无需做徒劳的重投标.
