Seda-based Adaptive Performance Optimization Techniques

This thesis presents a self-adaptive performence control mechanism based on staged event-driven architecture(SEDA).The explosive growth of the Internet in the last few years has given rise to a vast range of new services being deployed on a global scale. Internet services have become a vital resource for many people. With the fast-growing number of users and the increasing services, Web servers are expected to provide the information with an ever-faster speed to meeting the need of a variety of demands. How to improve the performance of servers is one of the hot fields in the computer science. At present, there are mainly two kinds of methods to improve the web server performance. One is to make use of many servers to compose of a cluster of servers. The other one is to make use of a high-performance web server software to improve the performance of the single sever. The main focus of this thesis is how to design and implement a high-performance server architecture.SEDA, staged event-driven architecture, which is a new design for highly concurrent Internet services. SEDA decomposes an application into a network of stages separated by event queues, and makes use of dynamic resource control to allow applications to adapt to changing load. SEDA combines the advantages of threads and event-based programming models to achieve high concurrency and robustness to load. Though SEDA exhibits high performance, it has some disadvantages. For example, there are some conflicts between its resource controllers, and the performance of the controller is strongly determined by the controlled parameter settings. These disadvantages make SEDA degrade the quality of service to requests. Thus a self-adaptive performance control mechanism is proposed to improve SEDA. The works described in this thesis are listed below:1. This thesis sums up and analyzes several methods to improve the performance of Web servers, which include QoS control, server cluster, server architecture and operating system.2. Based on summary, SEDA, a high-performance architecture is analyzed, both its components and performance control mechanism.3. After analyzing performance control mechanism of SEDA, this thesis proposes some disadvantages of controllers in SEDA, and designs a new performance control mechanism to improve performance of SEDA. This new mechanism includes optimization of stage, congestion control between stages, and overload control of whole system.4. Based on theoretical design, a new server architecture based on SEDA and self-adaptive performance control mechanism is proposed.5. A Http server based on new architecture is implemented for benchmarking and evaluation. The testing results show that new control mechanism is better than original one.6. At last, the contents of the thesis are summarized and the future works to do are proposed.