Abstract

This paper describes the design and implementation of a distributed file system (dfs) for embedded applications. The services provided in dfs include transaction management, file management and file access. Client processes use message passing to access these services. The file system provides client processes with different types of files. These include replicated files, with identical copies stored at multiple nodes, and consistent files, stored at a single node. Both are accessed only within transactions. Multiple readers or a single writer are allowed access to the files, and updates are performed atomically using a co-ordinated commit protocol over multiple sites. A novel concurrency control algorithm which provides deterministic response and avoids deadlock has been designed to resolve contention. Other types of file are also provided for low-overhead storage of less critical data. This paper motivates the need for different types of file, and describes the design and implementation of dfs using the CONIC programming environment for the construction of distributed systems.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call