S.B. Musunoori and F. Eliassen (Norway)
Quality of Service; Service Composition; Service Config uration; Scheduling.
By sharing computational resources of individual nodes, computational grids promise to provide easy-to-use infrastructures for parallel and distributed systems. For real-time distributed applications, it is the quality of ser vice (QoS) which decides their performance. The current existing grid solutions do not support QoS issues such as QoS specification and management, and are limited in per formance optimization. In this paper, we propose to use a platform managed QoS-aware service configuration ap proach, which enables application developers to separate functional specification and QoS requirements from im plementation decisions that depend on the deployment en vironment. The middleware platform is responsible for achieving QoS goals of the grid application service con figuration at deployment time. We refer to this task of middleware platform as service planning. We formulate the service planning problem, and present a basic algo rithm to search for different configurations. We will also present an analytical model of service planning for find ing QoS-tradeoff points in making configuration decisions. This model itself is an improvement of a common solution technique used for QoS management. We verify the model in the context of a real-time multimedia application service configuration.
Important Links:
Go Back
