Network File System (NFS)

Network File System ( NFS ) is a distributed file system ( DFS ) developed by Sun Microsystems. This allows directory structures to be spread over the net- worked computing systems.

A DFS is a file system whose clients, servers and storage devices are dis- persed among the machines of distributed system. A file system provides a set of file operations like read, write, open, close, delete etc. which forms the file services. The clients are provided with these file services. The basic features of DFS are multiplicity and autonomy of clients and servers.

NFS follows the directory structure almost same as that in non-NFS system but there are some differences between them with respect to:

• Naming
• Path Names
• Semantics

Naming

Naming is a mapping between logical and physical objects. For example, users refers to a file by a textual name, but it is mapped to disk blocks. There are two notions regarding name mapping used in DFS.

• Location Transparency: The name of a file does not give any hint of file's physical storage location.
• Location Independence: The name of a file does not need to be changed when file's physical storage location changes.

A location independent naming scheme is basically a dynamic mapping. NFS does not support location independency.

There are three major naming schemes used in DFS. In the simplest approach, files are named by some combination of machine or host name and the path name. This naming scheme is neither location independent nor location transparent. This may be used in server side. Second approach is to attach or mount the remote directories to the local directories. This gives an appearance of a coherent directory. This scheme is used by NFS. Early NFS allowed only previously mounted remote directories. But with the advent of automount , remote directories are mounted on demand based on the table of mount points and file structure names. This has other advantages like the file-mount table size is much smaller and for each mount point, we can specify many servers. The third approach of naming is to use name space which is identical to all machines. In practice, there are many special files that make this approach difficult to implement.

Mounting

The mount protocol is used to establish the initial logical connection between a server and a client. A mount operation includes the name of the remote directory to be mounted and the name of the server machine storing it. The server maintains an export list which specifies local file system that it exports for mounting along with the permitted machine names. Unix uses /etc/exports for this purpose. Since, the list has a maximum length, NFS is limited in scalabilty. Any directory within an exported file system can be mounted remotely on a machine. When the server receives a mount request, it returns a file handle to the client. File handle is basically a data-structure of length 32 bytes. It serves as the key for further access to files within the mounted system. In Unix term, the file handle consists of a file system identifier that is stored in super block and an inode number to identify the exact mounted directory within the exported file system. In NFS, one new field is added in inode that is called the generic number.

Mount can be is of three types -

1. Soft mount: A time bound is there.
2. Hard mount: No time bound.
3. Automount: Mount operation done on demand.

NFS Protocol and Remote Operations

The NFS protocol provides a set of RPCs for remote operations like lookup, create, rename, getattr, setattr, read, write, remove, mkdir etc. The procedures can be invoked only after a file handle for the remotely mounted directory has been esta- blished. NFS servers are stateless servers. A stateless file server avoids to keep state informations by making each request self-contained. That is, each request iden- tifies the file and the position of the file in full. So, the server needs not to store file pointer. Moreover, it needs not to establish or terminate a connection by opening a file or closing a file, repectively. For reading a directory, NFS does not use any file pointer, it uses a magic cookie.

Except the opening and closing a file, there is almost one-to-one mapping between Unix system calls for file operations and the NFS protocol RPCs. A remote file operation can be translated directly to the corresponding RPC. Though conceptu- ally, NFS adheres to the remote service paradigm, in practice, it uses buffering and caching. File blocks and attributes are fetched by RPCs and cached locally. Future remote operations use the cached data, subject to consistency constraints.

Since, NFS runs on RPC and RPC runs on UDP/IP which is unreliable, operations should be idempotent.

Cache Update Policy

The policy used to write modified data blocks to the server's master copy has critical effect on the system performance and reliability. The simplest policy is to write through the disk as soon as they are placed on any cache. It's advantageous because it ensures the reliability but it gives poor performance. In server site this policy is often followed. Another policy is delayed write. It does not ensure reliability. Client sites can use this policy. Another policy is write-on-close. It is a variation of delayed write. This is used by Andrews File System (AFS).

In NFS, clients use delayed write. But they don't free delayed written block until the server confirms that the data have been written on disk. So, here, Unix semantics are not preserved. NFS does not handle client crash recovery like Unix. Since, servers in NFS are stateless, there is no need to handle server crash recovery also.

Time Skew

Because of differences of time at server and client, this problem occures. This may lead to problems in performing some operations like " make ".

Performance Issues

To increase the reliability and system performance, the following things are generally done.

1. Cache, file blocks and directory informations are maintained.
2. All attributes of file / directory are cached. These stay 3 sec. for files and 30 sec. for directory.
3. For large caches, bigger block size ( 8K ) is benificial.

This is a brief description of NFS version 2. NFS version 3 has already been come out and this new version is an enhancement of the previous version. It removes many of the difficulties and drawbacks of NFS 2.