 |
Network Operating System Architectures - Servers
by
Matthew Martin
The subject of server systems follows on from the discussion of client systems. The transport layer is used in the same way by servers as it is by clients.
|
by
Matthew Martin
The subject of server systems follows on from the discussion of client systems. The transport layer is used in the same way by servers as it is by clients.
Some common network OS:
Network OS are increasingly able to work together (inter-operate). This is important because the IT infrastructures of businesses tend to evolve organically, rather than be carefully planned from the beginning. This means that systems working in very different ways may need to communicate with each other.
One solution where different systems have been purchased (usually over a long period of time but sometimes out of necessity) that will not readily exchange information, is the use of middleware. Middleware is specialised software that translates between systems. Often this is specially written (bespoke) to meet the communication requirements of the organisation. The specialised nature of middleware ensures that it can provide the speed required for rapid data interchange between different systems, this often underlies the necessity for bespoke middleware software.
Servers provide services that are available over a network. Often these services are the main or only reason for an organisation installing a network. Services like file sharing, e-mail, printer sharing or networking, web services and database connectivity can provide efficiencies in working practices and justify the installation of a network. All of these services are provided by separate, specialised server software.
Server software uses the same kind of network software for accessing the transport layer as used by clients. The server software typically includes some form of security, which not only restricts access to the system but also determines the access rights of users.
Advanced server Operating Systems (server OS):
These network OS are multitasking, using this ability to serve many clients simultaneously.
The central function of file management is:
Approaches to efficient file access by many clients include the following:
Advanced file sharing systems have disc caching, which transfers data into RAM for faster access by clients. This can also be used for updates, the updates are first made in RAM, leaving writing the changes to the hard drive until the workload as fallen-off sufficiently. This is more efficient because access times to RAM are much faster than for hard drives.
They may also use elevator seeker, which queues the requests for data not in the order in which they arrive but in order of how they most efficiently be accessed from the hard drive.
Directory hashing lists files by directory in an index that is mathematically derived in order to produce the most efficient access to files. For example, NetWare uses directory hashing.
Servers may use a number of different techniques, often in combination, to ensure that the file access systems are functioning reliably. Typically these systems are focused on the hard drive.
A second hard disk is used to “mirror” the main hard disk. All the data sent to the main hard disk for writing is also sent to the mirror disc. Usually the mirror disc will be placed in a different physical location for safety. If there is a problem with the first hard disc, the mirror disc can immediately take over. This can happen if the main disc completely fails (or the machine it is installed in has a power outage), or if there are bad blocks on the main hard disc causing data reading or writing to fail. In the latter case, the data will be read from the mirror disc at a later time, and read to good blocks on the main hard disc once the blocks that failed have been marked as bad. The mirror disc provides a failsafe system.
An application on the server constantly monitors the state of the hard drive for errors. If an error should occur, the software will immediately attempt to recover any data written to the bad blocks on the hard drive, mark the blocks as bad so that there are no future attempts to write to these blocks, and re-direct the file address to a different location on the hard drive.
This is similar to disc mirroring but instead the requests for data are fed to two discs simultaneously, each being controlled separately (either by separate disc controller cards in the same machine, or in separate machines). This effectively doubles the disc hardware being applied to the retrieval of the data. The server software looks at both hard drives and determines which one can obtain the data the fastest. This both provides a safety factor in the same way that disc mirroring does and improves retrieval performance on the network. The disadvantage of disc duplexing is that this is processor intensive.
Clustering is a way of allowing a group of machines to function together. Every machine in the cluster is connected by a local area network (LAN). Work is shared among the machines in the cluster, depending upon their work load. Should a machine fail, the work can be automatically transferred to another machine in the cluster. The machines in the cluster all function as servers, signalling to each other as they perform tasks. A number of different companies offer clustering services and in theory the machines on in the cluster do not need to all be of the same type.
Security is a key aspect to networking and server software must ensure that security is maintained. There are two main approaches to network security.
All resources are given a name and each name has a password associated with it. The ability to access a resource depends upon having the correct name and a valid password. Different passwords can be used to provide different levels of authority, such as read and write permissions.
Users are placed in groups. Each group has specific access rights. Thus most users mat be placed in a group called something like, “basic user”, guest users in “guest” and administrators in “administrators”. Each group will have its own set of access rights.
Much more could be said about security since this is a whole subject of itself.
Novell were the first company to produce a true network OS. Today they are still a major player but Microsoft have eaten away at their market share with NT and Windows 2000 & 2003 servers. Netware is very reliable and tends not to slow down under heavy work loads. Netware is a 32-bit OS able to handle up to 32 terabytes (1015 bytes) of hard disc. The security of Netware is boosted by security auditing and encrypted backups.
Netware Ladable Modules (NLMs): Netware is able to provide a large range of services, partly through Netware Ladable Modules (NLMs). These are program packages, written by both Novell and other software houses, that provide specific network functionality. This makes Netware servers very powerful and flexible. The disadvantage is that everything can end up being run on a single machine which can be a problem if it fails.
Dynamic Resource Configuration (DRC): Netware has a powerful system for allocating memory for disc caching and routing known as Dynamic Resource Configuration (DRC). This is essentially an Artificial Intelligence (AI) system that controls the allocation of specific network resources, adjusting them automatically as system usage alters.
Interroperability with Multiple Name Space: Netware is interoperable, being able to files held on machines running different OS through its Multiple Name Space technology. This allows files on different OS to be associated with a single name within Netware.
Multi-Protocol Routing: This allows Netware to route between networks (LANs) that use different network protocols.
Microsoft started its current strategy for the networking OS market place with Windows 3.11, also known as Windows for Workgroups. The idea is to build networking functionality into as much of its software as it can. At the same time as introducing Windows for Workgroups in 1992, Microsoft introduced LAN Manager. Later followed Windows NT with workstation and server versions in NT 4.0. Then came Windows 2000, with client (Professional) and three server versions, called Server, Server Professional and Data Center. The latest is Windows 2003 Server.
Windows for Workgroups was an extended version of Windows 3.1. It included both client and server software, so every installation would contain the software for both.
Windows NT 4.0 introduced NT File System (NTFS), providing fast access to files across a LAN. Previously Microsoft has used the File Allocation System (FAT) and this was still used for Windows 95, 98 and ME – and is also available for Windows XP as an option instead of NTFS (note that these are not network OS). NTFS provides faster file access with greater fault tolerance.
Windows NT uses NT domains, with an NT server maintaining all the accounts and security information for a single domain. Network services are advertised to other computers through NetBIOS broadcasts, or on IP-based networks through Windows Internet Name Service (WINS). This provides a distributed system of control of domains.
While Windows 2000 Professional is the workstation end, the server version of Windows 2000 provide progressive increases in power. Win 2000 Server can be used with up to 2 CPUs. Win 2k Server Professional can be used with up to 4 CPUs and 64Gb RAM, Win 2k Data Center with up to 16 CPUs.
The introduction of Windows 2000 Server brought with it a new feature, Active Directory Services (ADS). This is an object database for network objects, allowing for the use of containers for network objects. This makes administration much simpler. ADS also uses Domain Name Services (DNS) names, older NT names are converted into DNS names. The core of ADS is Lightweight Directory Services (LDAP). The organisation of ADS is hierarchical. Since ADS uses the DNS global naming system for LANs, it can provide trusted connections between businesses for business-to-business (B2B) networking. This all sounds very good but it is very complex and this could cause future issues.
The Windows network OS provide a range of network facilities including file sharing, printer sharing, drive mirroring, drive duplexing, file replication and uninterruptible power supply (UPS) communications and the sharing of serial devices among other features.
UNIX is a powerful, reliable and flexible OS, widely used by large organisations, universities, government departments and the military. Originally command line only, a GUI front-end is available in the form of X-Windows. LINUX has been around for a long time and probably will be for some time to come. With the development of its “little brother” Linux, it may begin to threaten Microsoft’s dominant position.
UNIX runs on powerful workstations (like Sun Microsystem’s SPARK Stations) and on mini & mainframe systems. Originally written in 1969-1970, it was completely re-written in 1973 in C. There are several versions by different companies, e.g. IBM’s AIX. One of the key versions is the Berkeley Standard Distribution (BSD), which is used as the basis for a number of different OSs, including Macintosh’s OS-X for its latest Apple computers. An off-shoot of UNIX is Linux, originally written to run on PCs with Intel’s 8086-family of processors (486, Pentium, etc.). UNIX varies in price from the very expensive to, in the case of Linux, the virtually free.
A key feature of UNIX is its Directory Services, which provides everything on the network with a name and user with varying authority to access named services and devices. This is achieved through a database of devices, users and connections.
Novell extended the Directory Services with Novell Directory Services (NDS), allowing information about the Directory Services to be displayed as a tree, making it easier for users and administrators to understand the system. It also allows for effective network desktop management, providing the basis for groupware functionality. A number of applications have been built on the basis of NDS, by Novell and other software providers.
The Lightweight Directory Access Protocol (LDAP) was introduced in order to provide network interoperability. This is achieved through a system of alias names. The system is not perfect but it works.
by
Matthew Martin