分布式失效检测算法的研究计算机软件与理论专业论文.docx

重庆大学硕士学位论文英文 重庆大学硕士学位论文 英文摘要 PAGE PAGE IV ABSTRACT Nowadays, the development and application of computer technologies, telecommunication and computer networks make the style of information processing shift from the traditional centralized way into decentralized pattern. In other words, people can now integrate and coordinate computational entities and resources (which are spatially distributed, loosely connected and based on different platforms) to carry out a task-driven transaction. However, the above-mentioned shift also brings challenges and difficulties along with brilliant prospect. Different from traditional centralized computing, distributed computing relies itself on interconnection which is not always reliable. Connection failure, message loss, unexpected long-delay can make interconnection in distributed systems fail and prevent components and computational entities from effective interaction and coordination. An important research topic of distributed computing is to ensure correctness and efficiency (quality of service) with the possibilities of unreliable interconnections and other unavoidable network failures. In recent years, there is a growing trend of using failure-detection mechanisms to implement fault-tolerant algorithm in unreliable distributed environment. Corresponding solutions has been used and implemented in many real systems and applications, such as PAXOS、 HARP、Chain Replication、BOXWOOD, etc. For correctness, two properties should be satisfied, namely the completeness property and the accuracy property. The completeness property requires that any process failure must be detected after the failure happens. While the accuracy property requires that if a failure-detection is made, there exists a failure before the detection time. This paper proposes failure-detection algorithm. The algorithm is based on a system model which consists of one main process, one customer process and various observering processes. The algorithm employs a “surv