Clustering, resource management, and replication support for scalable network services.

摘要

With the increasing demand of providing highly scalable, available and easy-to-manage services, the deployment of large-scale complex server clusters has been rapidly emerging in which service components are usually partitioned, replicated, and aggregated. This dissertation investigates techniques in building a middleware system, called Neptune, that provides clustering support for scalable network services. In particular, Neptune addresses three specific aspects in support of network service clustering: (1) the development of a flexible and scalable clustering architecture with efficient load balancing support for fine-grain services; (2) the design and implementation of an integrated resource management framework that combines the “response time”-based service quality, overall resource utilization efficiency, and service differentiation support; and (3) the design and implementation of a service replication framework focusing on providing flexible replica consistency, performance scalability, and failure recovery support. Neptune has been implemented on Linux and Solaris clusters and a number of applications have been successfully deployed on Neptune platforms, including a large-scale document search engine.