Performance analysis of coda file systems

Abstract

Distributed file systems provide access to data stored at servers using file system interfaces. These systems are useful from the point of view of the users due to many reasons. Data sharing of multiple users, user mobility, location transparency, location independence and backups and centralized management are some of them. Coda is one such distributed file system and it makes file available to a collection of client computer as part of their directory tree, but ultimately maintains the authoritative copy of the file data on servers. In addition to common facilities provided by distributed file system Coda has some more desirable features that makes it stand out among other distributed file systems. In order to support disconnected operation, Coda provides full access to a caches section of the file space during voluntary or involuntary network or server outages, and Coda will automatically reintegrate the changes made on disconnected clients when reconnecting is one such feature. Another important feature is the read write failover server replication where data is stored and fetch from any of a group of servers and Coda will continue to operate even only a subset of ail servers is available. The primary objective of this study was to implement Coda file system one a server running Windows NT 4 server operating system and a client machine running windows NT4 Workstation operating system. Such an implementation provides Unix like features on a windows platform. The second objective was to analyze the performance of the Coda file system whe the client cache manager was implemented as a user level process instead of havir it implemented in kernel level. The benchmark tests performed in the presence of the user level cache manage clearly showed improvement in the file system performance compared to when the client cache manager was not present.