Tuning adaptive computations for the performance improvement of applications in JEE server

With the increasing use of autonomic computing technologies, a Java Enterprise Edition (JEE) application server is implemented with more and more adaptive computations for self-managing the Middleware as well as its hosted applications. However, these adaptive computations consume resources such as CPU and memory, and can interfere with the normal business processing of applications at runtime due to resource competition, especially when the whole system is under heavy load. Tuning these adaptive computations from the perspective of resource management becomes necessary. In this article, we propose a tuning model for adaptive computations. Based on the model, tuning is carried out dynamically by upgrading or degrading the autonomic level of an adaptive computation so as to control its resource consumption. We implement the RSpring tuner and use it to optimize autonomic JEE servers such as PkuAS and JOnAS. RSpring is evaluated on ECperf and RUBiS benchmark applications. The results show that it can effectively improve the application performance by 13.6 % in PkuAS and 19.2 % in JOnAS with the same amount of resources.