|
Coda is a distributed filesystem, developed at Carnegie Mellon University since 1987, under the direction of M. Satyanarayanan. It is based on the older AFS, a similar but less advanced distributed file system.
Major features include:
- Disconnected operation - Disconnect from the file server and carry on working
- Client-side caching - Improves performance by reducing the need to fetch data from the server
- Replication - Store and manipulate the same data on multiple servers, to improve reliability and scalability
- Security - Uses a security system derived from Kerberos
- Well-defined semantics of sharing, even with network failures
- Native support in the Linux 2.6 kernel (as opposed to AFS or NFS3)
Major misfeatures of the standard CMU implementation:
- The server software maps its RVM metadata to memory, meaning it occupies about 4% of the maximum served disk space in memory (!). Also, RVM metadata is limited to about 300MB, resulting in a maximum volume size of about 8GB.
The emphasis of the Coda development appears to be on moving from a very good research project to a filesystem ready for production use.
External links
|
