NFS Protocol Overview

The Network File System (NFS), developed by Sun Microsystems, is the de facto standard for file sharing among UN*X hosts. NFS Version 3 is documented in RFC 1813. The Mount Protocol is closely related.

NFS is implemented using the RPC Protocol, designed to support remote procedure calls. All NFS operations are implemented as RPC procedures. A summary of NFS procedures is show below:

Procedure 0:  NULL - Do nothing 
Procedure 1:  GETATTR - Get file attributes 
Procedure 2:  SETATTR - Set file attributes 
Procedure 3:  LOOKUP - Lookup filename 
Procedure 4:  ACCESS - Check Access Permission 
Procedure 5:  READLINK - Read from symbolic link 
Procedure 6:  READ - Read From file 
Procedure 7:  WRITE - Write to file 
Procedure 8:  CREATE - Create a file 
Procedure 9:  MKDIR - Create a directory 
Procedure 10: SYMLINK - Create a symbolic link 
Procedure 11: MKNOD - Create a special device 
Procedure 12: REMOVE - Remove a File 
Procedure 13: RMDIR - Remove a Directory 
Procedure 14: RENAME - Rename a File or Directory 
Procedure 15: LINK - Create Link to an object 
Procedure 16: READDIR - Read From Directory 
Procedure 17: READDIRPLUS - Extended read from directory 
Procedure 18: FSSTAT - Get dynamic file system information 
Procedure 19: FSINFO - Get static file system Information 
Procedure 20: PATHCONF - Retrieve POSIX information 
Procedure 21: COMMIT - Commit cached data on a server to stable storage 

NFS is a stateless protocol. This means that the file server stores no per-client information, and there are no NFS "connections". For example, NFS has no operation to open a file, since this would require the server to store state information (that a file is open; what its file descriptor is; the next byte to read; etc). Instead, NFS supports a Lookup procedure, which converts a filename into a file handle. This file handle is an unique, immutable identifier, usually an i-node number, or disk block address. NFS does have a Read procedure, but the client must specify a file handle and starting offset for every call to Read. Two identical calls to Read will yield the exact same results. If the client wants to read further in the file, it must call Read with a larger offset.

Of course, this is seldom seen by the end user. Most operating systems provide system calls to open files, and read from them sequentially. The client's operating system must maintain the required state information, and translate system calls into stateless NFS operations.

NFS Performance

NFS performance is closely related to RPC performance. Since RPC is a request-reply protocol, it exhibits very poor performance over wide area networks. NFS performs best on fast LANs.