Abstract: In the past years, most existing grid resource management techniques focus on performance improvement for one kind of workload, while ignoring the scalability of resource provision for multiple types of workload, which restricts the grid system application to more fields besides e-science. This paper proposed a new resource management model and its implementation framework for mixed types of workload that can support diverse workload simultaneously, such as batch workload, service workload etc. The proposed policy that constrained resource allocation method can make the resource achievement satisfy the performance specification of each workload by abstracting "workloads-resources" sharing relationship into consumer sharing policy tree, so that the user utility of whole system is greatly improved.

Key words: grid, resource management, Mixed types of Workloads, Consumer Sharing Policy Tree, Consumer Utility

摘要： 多年来，网格资源管理技术的研究只是关注单一应用负载的性能提高，忽略了资源的可扩展服务能力，制约了网格系统从科学计算领域向更广应用领域的推广。针对以上问题提出了一种支持混合负载的网格资源管理框架，所提的策略约束资源分配方法通过将多种“负载—资源”的共享关系抽象成消费者共享策略树，使资源的获取能够同时满足批处理、SOA服务等多种类型混合负载的性能要求，从而大大提高了系统整体的用户效益。

关键词: 网格, 资源管理, 混合负载, 消费者共享策略树, 用户效益