Research On QoS-Based Grid Resource Allocation

As interest in Grids continued to grow, and in particular as industrial interest emerged, more and more distributed, heterogeneous resources participated in Grid. In order to achieve the goal of highly resources sharing and cooperating, these resource need to interact in a uniform, well-known way. Thus the importance and demands for the open standards are increased. Under the push of the demands, the architecture of Grid developed from early hourglass structure, which is computing-oriented, to Open Grid Service Architecture (OGSA), which is service-oriented. Because in OGSA-based Grid environment, service is the kernel and service can span multiple, different abstract levels, a service application may include multiple resource requests, which have different abstract levels. And the required resources are presented unified as logical resource in Virtual Organization (VO) and transparent to the user in physics, although they have different native managerial strategy and are distributed in different locations actually. Hence, how to guarantee the user's Quality of Service (QoS) demands and how to modulate the resource allocation and task scheduling between logical resource and physical resource, as well as how to improve the resource utilization, these questions become more urgent with the continual developing of Grid.Because the assurance of the user's QoS demands depends on the performance of the resource allocated by the system, it is especially important to carry on QoS-based resource allocation management in service-oriented Grid.Our research focuses on Grid QoS and resource advance reservation. In theory, we do a deep research in the area of the characteristics of Grid QoS, the guarantee mechanism of Grid QoS and the model of QoS-based resource allocation management. In practice, aimed at the demands of Grid simulation platform and the deficiency of the existing Grid simulation toolkit, we design and implement a Grid simulation platform, which supports advance reservation and with multi QoS metric constrain. Our research works mainly include:1. Analyze the characteristics of Grid QoS, classify the QoS parameters according their properties, put forward the hierarchical structure of gird QoS2. Analyze the definition and requirements of advance reservation established by GRAAP-WG of GGF, set up the state transition figure of advance reservation. Introduce the concept of virtual resource container to extend SNAP, present a SLA-based Grid resource advance reservation strategy.3. Analyze the content of QoS-based resource allocation management and referring to the OGSA specification, put forward a QoS-based Model for Grid Resource Allocation Management (GRAM-QoS).4. Design the Grid simulation platform, which supports advance reservation and with multi-QoS metric constraint. Implement the grid environment simulation module, and the resource allocation simulation module based on GridSim.