No abstract

This research proposes and tests an approach to engineering distributed file systems that are aimed at wide-scale, Internet-based use. The premise is that replication is essential to deliver performance and availability, yet the traditional conservative replica consistency algorithms do not scale to this environment. Our Ficus replicated file system uses a single-copy availability, optimistic update policy with reconciliation algorithms that reliably detect concurrent updates and automatically restore the consistency of directory replicas. The system uses the peer-to-peer model in which all machines are architectural equals but still permits configuration in a client-server arrangement where appropriate. Ficus has been used for six years at several geographically scattered installations. This paper details and evaluates the use of optimistic replica consistency, automatic update conflict detection and repair, the peer-to-peer (as opposed to client-server) interaction model, and the stackable file system architecture in the design and construction of Ficus. The paper concludes with a number of lessons learned from the experience of designing, building, measuring, and living with an optimistically replicated file system. * For example, Ficus uses an 'update notification' daemon (a push) to tell other replicas asynchronously of a new file version. This typically results in a much faster propagation than relying on periodic volume-wide file reconciliation (pulls).

Ficus is a replicated general ling environment for Unix intended to scale to very large nationwide networks. The system employs an optimistic one copy a vailability" model in which con icting updates to the le system's directory information are automatically reconciled, while con icting le updates are reliably detected and reported. The system architecture is based on a stackable layers methodology which permits a high degree of modularity and extensibility of le system services. This paper presents the motivations for replication and summarizes the case for optimistic concurrency control for large scale distributed le systems. It presents a brief description of the Ficus le system and concludes with a number of outstanding issues which m ust be addressed.

Previous distributed file systems have relied on either convention or obtaining dynamic global agreement to provide network transparent file naming. This paper argues that neither approach can succeed as systems scale to the kind of size that is anticipated in the current decade. We propose instead a novel name mapping scheme which relies on a fragmented] selectively replicated name transl* tion database. Updates to the naming database are coordinated by an optimistic concurrency control strategy with automatic propagation and reconciliation. A prototype of the name mapping mechanism has been implemented and is in use in the Ficus replicated file system. services.This paper argues that name transparency is essential to the success of a very large scale wide area distributed computing environment. It describes the difficulties of scaling current small scale name transparency techniques] and proposes a new mechanism suitable for use in a very large scale DCE.Section 2 describes salient characteristics of a large scale DCE. File system aspects of network transparency are outlined in Section 3, followed by a discussion of name transparency in Section4. The naming mechanisms of several distributed file systems are studied in Section 5. Section 6 then presents a new mechanism which relies on optimistic replication to support name transparency. We present our conclusions in Section 7.