Solaris NFS

Selecting a distributed file service with reliable transparent file access is one of the most fundamental network decisions both small businesses and large corporations must make. SunSoft simplifies the choice with NFS, an open standard that implements a client/server architecture for file sharing across heterogeneous systems, from PCs to mainframes.

The NFS distributed file service increases productivity and efficiency by giving users fast, transparent access to files located on global, enterprise wide networks. It lets users on multi-vendor client systems share information residing on a wide variety of server platforms. At the same time, NFS lowers the cost of computing by centralizing the management of heterogeneous, distributed disk resources.

NFS is part of SunSoft's ONC+ distributed computing solution in Solaris. ONC+ is one of the most successful and widely implemented distributed computing solutions in the industry today. In addition to NFS, ONC+ includes the NIS+ enterprise naming service, a powerful distributed application developer platform based on transport independent remote procedure call (TI-RPC) technology, and local disk caching technology. Together these components help Solaris users and developers reap the benefits of information sharing on today's distributed networks.

The overwhelming success and pervasiveness of NFS cannot be disputed. According to a study conducted in 1995 by Dataquest, NFS reached an installed base of 8.5 million systems in 1994. "We [Dataquest] also expect NFS to continue to grow rapidly, with commercial Unix and PC markets reaching an installed base of 12 million systems in 1997, and to remain the dominant distributed file service for heterogeneous environments. It is no surprise that NFS protocols are used extensively for file sharing over the Internet.

Figure 1: The estimated installed base and projected growth of Unix and PC based NFS systems from 1991 through 1997 according to Dataquest (1995).

Some of the major features of NFS include:

Scalability: Easily scales to support small, mid-range, and large enterprise networks.

Transparency: Lets users access remote files as if they were local.

Reliability: Ensures quick recovery from system and network failures.

Efficiency: Uses TCP for efficient file data exchange on LANs and WANs.

Global File Naming Consistency: File names are the same from every workstation.

Security: NFS authentication, authorization, and Access Control Lists (ACLs) together protect against illegal file access.

Centralized Administration: Reduces administration time and effort.

Preservation of File Integrity: File locking prevents file integrity problems that can occur when files are accessed by multiple processes at one time.

Industry leading performance: Proven by accepted industry benchmarks.

Continuous Service: Solstice Highly Available NFS is tolerant of system and network failures.

Global Enterprise File Sharing

Global enterprise wide access to information is a requirement for companies wishing to compete successfully in today's global economy. To reach this goal, the corporate network must be able to span multiple geographies while ensuring a high level of productivity. NFS supports these environments by giving users transparent, fast access to server file systems connected over LANs and WANs. The transparency and high performance file access offered by NFS gives global workgroups continuous easy access to file information.

Location Transparent File Access with the Automounter

The NFS automounter adds remote file systems to the client's local file system. This happens through a process called file mounting. With the automounter, file mounting occurs automatically and on demand. A user only needs to attempt to access a file system to cause it to be mounted. The user requires no prior knowledge of where the file system actually resides. This makes the global network appear to users as if it is a single disk resource.

Global Network Support with TCP Protocols

NFS in Solaris now utilizes TCP (Transport Control Protocol) for efficient, data exchange over global networks. TCP is a "reliable" transport protocol that can handle high latency and low bandwidth network connections. It independently performs functions such as dynamic retransmissions, packet sequencing, congestion control and error recovery, resulting in guaranteed data delivery. The result is that TCP increases both the efficiency and performance of NFS.

Figure 2: TCP enhances NFS global file sharing capability.

TCP is now the default protocol for NFS in Solaris. However, in order to support backward compatibility with the installed base of NFS UDP systems, NFS clients and servers on Solaris can utilize either UDP or TCP protocols. Therefore, NFS clients and servers can negotiate which protocol to use based on what protocol the system they are communicating with supports.

Improved Scalability and Performance with Local Disk Caching

NFS gives users faster access to files by storing large chunks of file data in a local, quick access data facility called a cache. Caching large amounts of file information on the client means fewer demands are made on the server. The load on the server is decreased while scalability and performance are increased.

CacheFS provides NFS with extensive local disk caching capability. NFS clients still cache as much data in fast access RAM as they can. But when additional space is needed, CacheFS lets clients use the local disk. Because the local disk is available for storage, NFS clients can cache data in large (64K) chunks, as well as entire directories. By caching large amounts of information, the client NFS system can more easily ensure that the data an application wants to access will already be in the cache in advance of demand.

Introducing NFS Version 3

NFS has recently undergone a major revision resulting in significant improvements in performance, scalability, security and efficiency. This new revision, called NFS Version 3, was jointly defined in 1992 by a group of vendors including IBM, Digital and SunSoft. NFS Version 3 offers substantial new improvements over NFS Version 2. These include:

Improved client write throughput.

Reduced server load resulting in increased scalability and performance.

Support for multi-gigabyte files on NFS servers.

The industry is rapidly converting from Version 2 to Version 3 due to its many benefits. However, in order to support backward compatibility with the installed base of NFS Version 2 systems, NFS has been implemented in Solaris to support both protocols. In other words, both NFS clients and servers can utilize either Version 2 or Version 3 protocols based on what protocol the system they are communicating with can support. Therefore, network administrators can easily transition to NFS Version 3 without impacting interoperability with installed Version 2 systems on the network.

Figure 3: Solaris NFS can interoperate with both Version 2 and Version 3 NFS systems

Solid Performance

NFS performance has increased an order of magnitude in the last five years, from hundreds of operations per second to thousands. Tests run with the accepted industry benchmark SPECsfs (formerly LADDIS) show that NFS Version 2 is now an industry leader in performance.

NFS Version 3 Performance Advantages

Preliminary measurements show NFS Version 3 is exhibits even greater performance improvements over Version 2. One of the most important reasons for this is a new feature called "safe asynchronous writes." With safe asynchronous writes, clients no longer have to wait for individual write requests to complete as they did with synchronous writes.

NFS Version 3 also increases performance by reducing requests for file attribute information. Version 3 is designed to return attribute information along with the results of every operation. This increases the likelihood that cached attribute information is up to date and decreases the need for separate attribute requests to the server.

Finally, NFS Version 3 has no maximum transfer size limitation. This means NFS can construct and send larger chunks of data over the network at one time. The increased buffer size also lets NFS better utilize higher bandwidth network technologies such as FDDI and 100baseT Ethernet, which substantially increases NFS performance.

Preservation of File Integrity

There are times when a file might be opened by multiple processes concurrently. This poses a problem if one or more processes are writing to the file. When this happens, the file readers must somehow detect that a change has been made to the file and then get a copy of the updated information. The challenge is to coordinate changes made by the writers, while at the same time give accurate file snapshots to the readers. More complicated still is to synchronize changes among readers and writers when portions of the file are cached on each user's local system.

Using so called "cache consistency" mechanisms to solve this problem cannot always guarantee appropriate results. Therefore, the best solution is to make sure that problems associated with concurrent file access can't occur in the first place. NFS handles this with a newly improved file locking feature called the lock manager. The lock manager and another process, the status monitor, work together to guarantee that multiple readers and writers do not collide with each other.

The NFS lock manager provides Unix style locking (also known as advisory locking) between operating client and server processes enabling synchronized access to files. The lock manager lets a process lock a file or part of a file for shared or exclusive access. Exclusive access means no other process can share the lock. Shared access allows other processes to share the lock, but no one process will be able to obtain an exclusive lock.

The status monitor gives the lock manager information on host status. It monitors the system to ensure that locks will be handled properly if a system crashes while a file is locked. Non-monitored locking is available for client systems that do not support multi-tasking and cannot run both lock manager and status manager at the same time.

Simplified Administration

NFS eases the burden of network installation, configuration, and maintenance. It eliminates or simplifies routine administration tasks. NFS administration is also centralized. This means many tasks can be performed in one location instead of being spread out across multiple systems on the network.

The Global NFS Namespace

The NFS automounter uses special configuration tables called "maps" in order to make server file systems available to users on client systems. Maps associate client pathnames to server file systems. The client pathnames collectively represent the NFS namespace. Using the maps, NFS administrators are free to tailor the namespace to their particular environments so that users can locate their files quickly and easily. Storing the maps in the naming service enables the NFS namespace to be administered centrally.

Figure 4: The shared NFS namespace can be tailored to suit the specific needs of different environments. In this example, all users home directories are accessible by using the pathnames "/home/username".

The NFS namespace is a global or "shared" namespace. This means users can move freely to different network locations and still access files using the same pathnames they used on their home systems. The same is true of applications that are invoked from different client systems on the network.

As an alternative to using automounter maps, the NFS namespace can now be defined and managed using the guidelines encompassed by the X/Open Federated Naming Specification (XFN). XFN incorporates the file system namespace into an overall enterprise namespace that also encompasses other enterprise objects such as users, hosts, organizations, services etc. This simplifies and unifies naming for all objects across the enterprise.

Automounter Improvements

As stated earlier, the automounter enables users to access global files transparently. After initial configuration, the automounter mounts file systems automatically "on demand" and does not require administrator intervention.

The latest implementation of the automounter, the autofs automounter, offers improvements in performance and the ability to manage changes on the network. The autofs automounter is implemented as a virtual file system (VFS). This enables NFS file systems to be mounted on the mount points named in the automounter maps. Therefore, the automounter no longer requires the use of symbolic links increasing performance and efficiency. In addition, changes to client mount points require only a simple map edit.

Lowered Costs on the Network

NFS enables administrators to lower costs by supporting cost effective diskless and dateless client systems and low cost PC, Mac and Sun desktops on the network. Sun's Solstice AutoClient systems are also supported by NFS. Solstice AutoClient offers many advantages over traditional diskless and dataless systems including centralizing frequently required tasks like system backups and upgrades.

Secure File Access

Distributed networks are inherently vulnerable to "illegal" intruders. This vulnerability grows as the networks grow and connections extend beyond the local organization, exposing files to a larger unknown community of strangers. System administrators must take care to select a distributed file system that gives them the freedom to choose from a variety of security mechanisms that can be tailored to their environments.

Authentication, Encryption and Authorization Services

NFS uses authentication and authorization services to maintain security on distributed networks. An authentication service checks a users network identification before allowing access to remote resources. NFS can optionally be configured to utilize any one of multiple authentication services including well known services such as Kerberos and Diffie-Hellman. Note that this list is being extended to include new authentication services in the future.

Some of the authentication services utilized by NFS employ encryption services based on the Data Encryption Standard (DES) developed by IBM and adopted by the National Bureau of Standards. Encryption helps ensure that information transferred over the network cannot be easily interpreted by an unintended recipient who has gained unauthorized access. Protecting information by encrypting it is especially important when confidential information is being exchanged.

The purpose of an authorization service is to make sure users have permission to access resources once they have been authenticated. NFS supports two authorization mechanisms: "Unix style" permission checking and Access Control Lists (ACLs).

With permission checking, a set of permission bits set "on" or "off" indicates whether read, write and/or execute permission is allowed for the file's owner, groups of users, or for everyone. With permission bits, access to a file is granted or denied on a per group basis for everyone other than the file's owner.

Access control lists are just what the name implies: a list that controls access to files. With ACLs, extensive lists of authorization information can be maintained for each file, enabling a finer granularity of control over file access. For example, with ACLs access can be controlled on a per user basis in addition to on a per group basis. NFS supports ACLs based on the POSIX 1003.6 draft standard.

NFS supports Access Control Lists (ACLs)

The Solstice Highly Available NFS Service

In some environments, any significant disruption of service can lead directly to a significant loss of business. Solstice Highly Available NFS has been developed to address the needs of these environments. SunSoft's Solstice product line includes components necessary to make NFS highly available. Clients connected to the Solstice Highly Available NFS server will not experience a significant disruption of service despite failure of a single component. This is because Solstice Highly Available NFS replicates hardware components such that a failure is masked, enabling service to the client to continue. Some of the features of Solstice Highly Available NFS include:

Primary and secondary servers and dual ported disks

Secondary server takes over if primary fails

Heartbeat and fault detection software decide when a "take over" is necessary.

File locking state is recovered and preserved after a take over.

Figure 6: Solstice Highly Available NFS service

Solstice Highly Available NFS is available on Solaris SPARC platforms from Sun Microsystems Computer Corporation (SMCC).

Heterogenous Systems Integration

NFS is available on a wide variety of hardware and software platforms, providing the network administrator with unlimited options for integrating heterogeneous systems on enterprise networks. NFS is licensed to over 300 organizations including IBM, Apple, SGI, DEC, Hewlett-Packard, and Novell to name only a few. NFS implementations are available on every major operating system platform including MS-DOS, Windows 95, Windows NT, Unix, VMS, MacOS and MVS. NFS is also supported on a wide range of hardware platforms from PCs to mainframes.

NFS: The Global Enterprise File Sharing Solution

NFS fully satisfies enterprise requirements for reliable, transparent global file sharing today and in the future. A new, highly improved version of the NFS protocol, NFS Version 3, is now available. An industry leader in performance, NFS provides fast access to file information and has the scalability to support small to large corporate environments. It enables administrators to choose security solutions that fit their ever changing needs. Centralized administration reduces the time and effort it takes to perform a variety of routine administrative tasks, including managing the shared NFS namespace. File locking guarantees coordinated file access. And, for mission critical environments, Sun Microsystems provides Solstice Highly Available NFS.

These features, together with the vast array of multi-vendor NFS products to choose from and a rapidly growing licensee and installed base, reaffirm that NFS will remain the best distributed file sharing solution for the global heterogeneous enterprise well into the future.

Contents

Global Heterogeneous File Sharing