The Research And Implementation Of Policy Based Self-optimization System

With the rapid development of Internet technology, large-scale distributed systems can be widely used in various fields. However, the complexity of the system increased with the expansion of the system size, the system administrator must face greater challenges in management of these large-scale distributed systems. There is an urgent need to reform the traditional approach to system management. The concept of autonomic computing has been proposed by IBM. Its purpose is to build a distributed system capable of self-management with the theories and methodology of autonomic computing. The management of autonomic system is based on system-related knowledge base and policy, to achieve properties of self-configuration, self-optimization, self-healing and self-protection. And the ultimate goal of autonomic system is to complete the cumbersome and complicated maintenance work autonomously which originally operated by the system administrators manually.This paper focuses on the aspect of self-optimization in autonomic computing, the architecture of autonomic computing and the process of autonomic managenment has been studied. A model of dynamic resource allocation has proposed. And a unified model of policy and knowledge base has been built according to the characteristic of self-optimization. Implement the property of self-optimization in the process of web service request control, to achive policy based management of request.First, the characteristic of resources allocation and demand for self-management in distributed systems has been analysed. And the significance and feasibility of achieving self-optimization features in systems to evaluate. To predict the resource requirement of system in dynamic environment, a model of resource allocation based on Support Vector Machine regression algorith has proposed. Use data samples of recent time window by monitoring the resource configuration and performance parameter of autonomic element, to build regression model of dynamic resource requirement. Perform online learning for regression model, and use it to prodict the resource allocation adjustment of automonic element. Eventual optimize the performence of resource allocation.In autonomic system, behavior of self-management is guilded by the high-level manamgement policies which work out by administrator of system. Autonomic element is on the basis of knowledge base, and transfer policy to instruction that system can execute directly. On the basis of analysis of the mean to implement the self-optimization feature which based on policy, an approach of self-optimization based on hybird of action policy and utility function has proposed. And a model of knowledge base and system model that realized the feature of self-optimization have designed.Finally, according to system model of self-optimization proposed before, and experiment to evaluate the feature of the self-optimization. The result of experiment indicated that, with implementing self-optimization in web services system, more effective control to management of service request is easy to accomplish. And the quality of service also improved.