(An) implementation of file server in a distributed computing system

Abstract

In the implementation of file system, there are two approaches; file server implementation and kernel implementation. The kernel implementation is used in centralized time-sharing computer system. But most distributed file systems are implemented on several computers, on which file servers are running to manage file system and provide file services to clients. We chose file server approach to implement a file system. The design of file server was focused on file system reliability, consistency and expansibility. During file operations, machine failure or network failure can cause partial results to remain in system. System reliability is enhanced by disk shadowing. To keep system consistent under concurrent file access, transaction concept and optimistic concurrency control are used. And file server has layered structure to support system expansibility.