Treffer: Supplying High Availability with a Standard Network File System

This paper describes the design of a network file service that is tolerant to fail-stop failures and can be run on top of a standard network file service. The fault-tolerance is completely transparent, so the resulting file system supports the same set of heterogeneous workstations and applications as the chosen standard. To demonstrate that our design can provide the benefit of highly available files at a reasonable cost to the user, we implemented a prototype based on the Sun NFS protocol. Our approach is not limited to being used with NFS, however. And, the methodology used should apply to any network file service built along the client-server model. 1 Introduction There are two approaches to building fault-tolerant distributed programs. The first is to choose an available programming abstraction that reasonably fits the problem at hand (e.g. transactions (e.g. [7]), replicated procedure calls [4] or reliable objects [2]) and implement the program using the abstraction. The second.