Resarch of a Machine-Learning Based Load Prediction Approach for the Service-oriented Computing Environment
服务计算环境下一种基于机器学习的负载预测方法研究

With the rapid development of computer technology,the distributed applications scale up increasingly,and more software systems begin to make use of service-oriented architecture SOA.To improve the dependability and scal- ability of SOA,one effective way is to provide service replicas and balance loads among the replicas via adaptive load balancing service based on the middleware.By using middleware,we can satisfy the urgent demands of performance, scalability and availability in current distributed service-oriented applications.However,we must pay attention to the fact that the computing of the load should be predicative to avoid the affection of the peak load.To the complex service- oriented applications,the peak means the system may suffer extremely high load for a short period which will cause the system to be overload and unstable.The response time will be increased so that the overall throughput will be affected too.Therefore,in order to decrease response time and to utilize the services effectively especially when the workloads fluctuate frequently,we have proposed a new technique based on machine learning for adaptive and flexible load balan- cing mechanism within the distributed middleware.