Research On Cluster Server Consolidation System

Server Consolidation is a new IT model that emerged in later 1990's. It combines the openness and scalability of distributed client/server model and the centralization and easy-to-management features of mainstream model. Cluster system is the best platform to implement scalable large-scale server consolidation system. In this dissertation, key technologies to build scalable large-scale Internet server consolidation system are studied, with the emphasis on single entry point (SEP), high availability management and scalable parallel Internet application.After analyzing all aspects of designing cluster server consolidation system, we study the performance model of network Server based on M/G/1 PS queue. The condition for minimum average-wait-time under random dispatch algorithm is presented and proved.Then this dissertation concentrate on the implementation of SEP. Five technologies based on packet forwarding are discussed in detail. Then the design, implementation and performance analysis of the SEP software, PowerRouter on the Dawning 2000 Cluster, are given. PowerRouter adopted the direct forwarding technology. It has the advantages of direct reply without source package modification and OS modification. Benchmark shows that PowerRouter can well support large Internet application such as large-scale web server. The technology based on application layer forwarding is also discussed. Then a new technology, delayed-decision network address translation, is presented, which combined the efficiency of transport-layer forwarding and the flexibility of application-layer forwarding.A central, flexible system management software is necessary for each server consolidation system. In this dissertation, the design and implementation of application-management software for Dawning Server Consolidation (DSC) are presented. DSC is independent of OS, application software and special nodes. The virtual service and load balancing design combined the parallel-application management and high availability control. It also has mechanisms for global resource allocation and reclamation, flexible event processing, dependable communication, dynamic configuration etc. The system is a clear and flexible...