一种面向应用的光子网格资源联合分配策略

A policy of combined resource allocation for application-oriented optical grid systems

基于对业务驱动的光子网格系统的研究,提出了一种面向应用的光子网格服务提供模型,它将应用资源和光网络资源看作对等的实体,并将其抽象和封装为网格可以控制和共享的服务。基于对光子网格资源分配问题的研究,提出了一种用于资源联合分配的最小代价策略,它从最小化作业执行代价的角度解决了应用资源和光网络资源的联合分配问题。构建光子网格实验平台,并以网格计算业务在分布式岩心图像处理中的应用为例,对资源分配策略进行实验分析。结果表明,最小代价策略能够降低光子网格系统的业务阻塞率和平均处理时间,优化网格应用资源和光网络资源的使用。

A service provisioning model for application-oriented optical grid systems was proposed, which regards application resources and network resources as peering entities, and abstracts and encapsulates them into the controllable and sharable services for grid systems. A minimum cost policy for combined resource allocation was presented, which can resolve the problem of combined allocation of application resources and optical network resources to minimize the execution cost of service requests. A test bed of optical grid was designed to validate the feasibility of the service provisioning model, and the performance of resource allocation policies were analyzed experimentally through an example of grid computing in distributed core image processing. The experimental results show that the minimum cost policy can improve the performance in terms of blocking probability and average handling time upon optical grid systems and co-optimize the usages of application resources and network resources.