Distributed File Systems

Abstract

AbstractA distributed file system, also known as a network file system, is a means for to access data transparently across a network. A user should not have to know whether a file actually resides locally or on a remote server. In the typical distributed file system architecture in Fig. 1(a), a client uses the same standard file system calls that are translated into network requests at runtime. If the access to remote data is not transparent, a client might have to retrieve explicitly a file locally, make modifications, and then move explicitly the file back to the server as in Fig. 1(b).Distributed file systems are a broad topic in computer science. We have organized our discussion into four distinct areas. We begin by focusing on high availability techniques that improve uptime and provide support for mobile or disconnected operation. Then, we closely examine several protocol standards for compatibility and implementation‐independent optimizations. The next section describes several distributed file systems that are tuned for high‐performance computing (HPC) applications. Finally, we discuss several distributed file systems that have application‐specific features.