Looking from the user.

Server NFS exporting one or more file systems. Each exported file system can be either a disk or its section of a subtree. (Various versions of UNIX have their own rules fragmentation exported systems. Some of them might, for example, allow only export the file system entirely, others only one of a subtree in each file system). The server can determine, usually through the lines in the file / etc / exports, which customers can access each exported file system, and allowed access to it : "read-only" or "read and write". The client machines can mount a file system or of a subtree to any location in its existing hierarchy of files just as they would be able to mount any local file system. Customers can install directory with a "read-only", even if the server is exported as a "read-write". NFS mount are two types : hard and myakgoe. The type of mount dependent behavior of the customer if the server fails to respond to the request. If the filesystem mounted rigidly, the client continues to repeated requests for a response. In the case of soft mount client take some time refused to repeat requests and receives an error. When Mount fired, the client may request that the files in a remote file system using the same operations that apply to local files. Some systems also support this type of mount, which corresponds to conduct rigorous mount for the organization repeated attempts to mount a file system, but it turns soft for the next I / O. Operation mount NFS less limited compared to mount local file systems. Protocol does not require the caller to mount an operation was privileged user, but most users are imposed on them. (For example, Digital ULTRIX company allows anyone to mount NFS file system until the user has access to records in the directory mount point. Users can mount the same file system to multiple locations tree directories even your subdirectory. It can only export their local file systems and can pass their own mount points during the passage of the track name. Thus, to see the files server, the client must mount all file systems.

The original design.

design of NFS had the following objectives : NFS should not be limited to the operating system UNIX. Any operating system should be able to implement NFS server and client. Protocol should not depend on any of the specific hardware. Should be implemented simple repair mechanisms in the event of server failure or client. Applications must be transparent access to files without the use of special travel names or libraries, without recompiling. To UNIX-klientov be maintained UNIX semantics. Productivity NFS must be compared with a capacity of local disks. Implementation should be independent of the vehicles.

Components of NFS

The NFS consists of several components. Some of them are located either on the server or the client, and some use both. Some components are not required to provide basic functional vozmozhnosteey but are part of an expanded interface NFS : NFS Protocol defines a set of requests (operations), which can be sent to the client to the server, as well as a set of arguments and return values for each of these requests. Version 1 of the protocol existed only in the bowels of Sun Microsystems, and has never been released. All of NFS (including NFSv3) supports NFS version 2 (NFSv2), which was first published in 1985 under SunOS 2.0. Version 3 of the protocol was published in 1993 and implemented by some firms. Table 3.1 provides a complete set of NFS requests. The remote procedure call (RPC) to determine the format of the interaction between client and server. Every NFS sent as a package RPC. Advanced reporting (XDR-Extended Data Representation) provides mashinno-nezavisimy encoding method for sending data through the network. All requests RPC XDR encoding used for data transmission. The XDR and RPC are used for many other services, in addition to NFS. The code NFS server is responsible for processing all client requests and provides access to exported file systems. The code NFS client implements all treatment client files to the server by sending one or more of RPC requests. Mount Protocol defines semantics mount and unmount filesystems NFS. NFS uses several protsessov-demonov background. The server set demons waiting nfsd NFS client and answer them. Demon mountd process requests mount. The client set biod daemons handle asynchronous I / O blocks NFS files. Blocking Network Manager (NLM-Network Lock Manager), and monitor network status (NSM-Network Status Monitor) together provide a means to lock files on the network. These funds, although not formally associated with NFS can be found in most NFS implementations. They provide services not possible in the core protocol. NLM and NSM implement server operations by demons lockd and statd respectively