This chapter aims to introduce Local Area Networks (LANs) to library computer users and to provide some background information to acquaint the reader with some of the concepts used in the case studies that comprise the body of this volume. It does not assume any previous technical knowledge of the reader, but only a general familiarity with some type of microcomputer. The intended audience of this material includes library staff who have some degree of responsibility for microcomputers in the library, especially those who anticipate installing a LAN in the near future or those evaluating the need for a LAN.
Computer networking can be a complicated topic, but one does not need to know the subject exhaustively to set up and manage a LAN in a library. It is quite possible for a library staff member with little technical knowledge to purchase a network package and set up a LAN by reading the manuals and documentation. This introductory chapter attempts to present enough of an overview of the concepts of computer networking so the reader can place the networks described in the following articles, or other networks that they work with, into some perspective and see how they compare to other networking options.
My general attitude about LANs in the library environment is that network implementors should do as much as possible to mitigate the degree of complication that a network imposes on computer users. Establishing a LAN will inevitably introduce some degree of increased complexity to using a microcomputer, but it should not cause dramatic complications. Microcomputers have become an important tool for library staff to accomplish their work. A LAN should simply be an extension of that tool and a means for that tool to become more powerful and efficient. If the LAN implementor does his/her job well, the computer users should be pleased with the new things that they can do with the network, once they have recovered from the initial learning curve, and not be disgruntled because they can't figure out the network.
The main purposes of a LAN involve sharing computer resources and facilitating communication among computer users. With a LAN, individual microcomputers have access to more resources than would be available otherwise. LANs allow users of the network to share common databases, spreadsheets, and documents, high-quality printers, as well as to communicate with each other throughout the network through electronic mail.
Not only does a LAN provide better functionality than isolated stand-alone microcomputers, but in many cases can provide more overall computing resources to staff at less cost. For example, once connected to a LAN, a minimally configured floppy disk system can greatly extend its functionality through access to a hard disk and laser printer. In many cases it is more economical to concentrate resources on a LAN server rather than purchase hard disks and printers for each microcomputer in an organization. Especially when purchasing fairly large numbers of microcomputers, libraries can economically provide users with faster processors and better displays if disk storage and printer access can be relegated to the network. The cost of a group of diskless workstations and a file server and laser printer is generally less than a comparable number of fully equipped stand-alone microcomputers.
Components of a LAN
A Local Area Network includes many components. These components take many forms—some software, some hardware, and some human. Each of these components must fit together well in order to cohere into a successful network. These components include workstations, servers, and the network cabling and hardware.
Workstations are microcomputers that use network resources. Computers attached to a LAN may provide resources, use resources, or both. When a computer only uses network resources, network terminology deems it as either a workstation or a client.
The minimal configuration for a network workstation includes a central processing unit, keyboard, monitor, and network interface card. It is possible for all disk access and printing capability to be provided only by the network. In most networks, however, it is typical for most workstations to have their own disk storage and printing capability, and to depend on the network for supplementary functions. It is common, for example, for a network workstation to have its own dot-matrix printer to print draft copies and to use a laser printer on the network for the finished copy. Such an arrangement leaves each individual workstation less vulnerable to network failures and relieves the network of the strain that might be imposed through the constant access of programs that could reside on local disk systems.
A microcomputer on a network can function without any disk storage of its own—either in the form of a floppy disk or hard disk—if its network interface card includes a PROM (Programmable Read-Only Memory) chip that allows the system to receive its start-up programs from the network. Such a system without any local disk drive is termed a diskless workstation. Network implementors may choose to implement diskless workstations for both economic and security concerns. The economic reasons seem clear: it is less expensive to set up diskless workstations because the cost of the boot PROM chips are much less than a floppy disk drive. Security concerns may be less obvious. Without disk drives, it is impossible for users to load software onto the network that may not be authorized or that may include undesirable functions. Software tainted with computer viruses and Trojan Horse programs that have received so much attention in recent years fall into this category. Many government and business environments may want to prevent network users from copying and distributing sensitive data, and might use diskless workstations to discourage this activity. In the library environment the opposite usually applies. Libraries generally promote the distribution of information, and would therefore want to provide disk drives so that network users can store and keep retrieved information.
The computers that provide network resources of some type are called servers. Servers are usually some type of microcomputer to which some special equipment and/or software has been added to perform special functions for the benefit of the network. The types of servers include file servers, print servers, communications servers, fax servers, and database servers.
When a computer system exists for the sole purpose of performing a network function it is called a dedicated server. The term non-dedicated server refers to systems that can function as a workstation at the same time that it performs its network role. When a user's workstation can double as a network server, then the network implementors can avoid the cost of an additional computer system. Such non-dedicated servers, however, may not perform as well as dedicated servers since their computing ability must be split between two tasks. Smaller networks tend to implement non-dedicated servers, while large networks generally require dedicated servers.
File Servers. One main function of a LAN lies in providing users access to supplementary disk storage on a large hard disk located on a central machine on the network. A network file server generally consists of a high-performance microcomputer system with a very large hard disk. Since the file server must service the disk storage needs of many users on the network, the network implementor must carefully consider the performance and capacity of this system.
Network file servers require special software to manage access to disk storage by the network users. This network software must include features to allow access to the server by multiple users at the same time but to restrict access to only authorized users. The server's software should also include capabilities for network administrators to configure, monitor, and manage the resources on the server.
The server's disk storage can store both software and data. All network users can potentially access software loaded on a network file server. Rather than each user having to load individual copies of their software from their local disk drives, each workstation on the network may execute programs from a common copy located on a network file server. Network use of software facilitates standard use of software—everybody uses the same version, and upgrading to a new version only has to be done once.
Although networks allow multiple users to share a single copy of a given piece of software, this does not remove the legal requirements of software licenses. The shared use of network software does not mean that libraries only have to purchase one copy of each application. Software must still be used within the restrictions that the vendors impose. This may mean that the library will need to purchase a copy of the software package for each user on the network, even though the network will function with only one copy of the software. Many vendors, however, will offer a network license for their products that provides the one necessary copy of the software, along with a license for each person on the network to legally use that copy, and usually additional copies of documentation and manuals. Although the cost of a network license of the software generally exceeds the cost of a single copy, such a license will generally be much less than the cost of buying individual copies for each network user.
Network file servers can also store data for all the users of a network. File servers generally provide both private and public disk storage space. Each network user usually owns a private area on the file server that no other network user may access. File servers may also offer public directories which all users can view. Data may also be stored in areas that are accessed by particular authorized groups of users, but not others. The network administrator may authorize some users to create, add to, and modify data files, while other users can only view data.
A complication to storing data on a network lies in the dangers of simultaneous access by multiple users. If two people attempt to update the same piece of data at the same time the data might become corrupted. Network file servers protect data through file and record locking. File locking means that the network software will permit only one user to have update capability to a file at a time. If someone has the file open with update privilege, then anyone else who tries to access the same file will have only viewing privilege. Some network data systems have record locking capability. With record locking, multiple users may simultaneously access a file with update capability, and the system will control updates to individual records.
One of the great advantages to using shared data storage on a central file server relates to data security. Practically all networks have some system established for the regular backup of data. Not only are network file servers generally much more reliable than a hard disk on a standalone microcomputer, but since the data of so many users are at stake, network managers are highly motivated to ensure regular backups. Individual users almost always take a lax attitude toward protecting their own data files. Network administrators must also protect data from the possible dangers imposed by computer viruses and other ill-behaved software.
LAN users access a network file server in just the same way that they access non-networked disks. Any good network system will make access to remote disks space on the network just as easily accessible as the ones attached to the user's own system. With MS-DOS machines, local drives are accessed through logical drive letters such as A: B: or C:. Once attached to a file server, users access their authorized network drives through higher drive letters. For example, one might have a drive G: defined as private disk space on the network file server and H: as a directory of departmental report documents. With the Macintosh and other machines with a graphical interface, network drives appear as icons just as do local disk drives. The network software on the user's workstation in conjunction with the software on the network server takes care of assigning logical drives on the user's system that correspond to the physical disk space on the server. As part of the security system on the network, users may have to enter a password in order to access network drives.
Print Servers or Network Printers. Local area networks also facilitate the sharing of printers. Most organizations cannot afford to purchase a laser printer for each microcomputer user. With a network printer on a LAN, all authorized users on the network can print on a common laser printer as if it were attached to their own machine. Networks can usually include high-end laser printers since their cost can be amortized across many users. Laser printers provide fast, quiet, and highly flexible printing functions.
A print server consists of a microcomputer connected to the network, which has one or more printers connected to it. Just as with the file server functions on the network, some type of software is required to control and manage access to the network printers. Many network systems combine the file server and print server functions into a single system.
The print server must manage printing for a lot of people at the same time, controlling all the printing requests that network users send to the printer. In order to manage multiple simultaneous print requests, the print server must receive each request, store it, and send documents to the printer one at a time. The term spooling refers to this process of receiving, storing, and releasing print requests directed to the server. The print server must also contend with requests for special forms. Downloadable fonts pose another complication for network print servers, but most network printing systems will accommodate them.
Just as described for file servers, access to a network printer follows whatever conventions that apply to local printers for your system. For example, MS-DOS systems use LPT1: as the logical designator for a locally attached printer. If connected to a LAN, the user might user LPT2: to access a network printer.
Communications Servers. One other type of network server relates to communications functions. Some LANs may have some common means of communication with other networks or systems. If microcomputer users in your organization need access to mainframe systems or dial-up access to outside computer services, you might want to set up a communications server so that all network users can share a high-speed modem, or a pool of modems.
Modem sharing is one communications function that may be implemented on a LAN. There are products available for most network systems that allow all authorized users to access a high-speed modem. Although the number of users that can simultaneously access network modems is limited to the number of modems and phone lines available, such a scheme may be more economical and convenient than providing individual modems for each user.
Fax Servers. The ability to send and receive material by fax has become a necessity for most organizations. Any microcomputer can send and receive a fax once equipped with a fax board and software. Recently, many products have emerged in the market that make it possible to perform fax transmission and receipt over a network. Through the implementation of such a fax server, all the users of the network can send and receive faxes through the network without having to have their own fax board and phone line. These products often require a dedicated microcomputer to manage incoming and outgoing fax messages for network users.
Database Servers. Database servers not only store data for the network, but also perform most of the work related to retrieving data. In traditional networked database applications, the data may be stored on a network file server, but the workstation must run the software that searches, retrieves, and displays the data. A database server takes over much of the work from client workstations. The software on the workstation formulates a request for data which is then passed to the database server which then performs the search on the database and passes back the results of the search to the workstation. The computations involved in performing the search are done on the server's processor, not on the workstation's. This model of database access allows a more powerful processor of a database server to relieve workstations on the network of some of the work involved in information retrieval. The most common type of database server implementation is known as Structured Query Language, or SQL.
A network requires several hardware items in order for the individual computers to function as interconnected systems. Each computer must have some type of interface board installed in it to communicate with the network, and this network board must have some type of cable that attaches it to the other machines on the network. Some networks may need other equipment such as repeaters, routers, bridges, and gateways in order for all the systems to communicate.
Network Interface Cards. The network interface card (NIC) connects the system unit of the microcomputer to the cabling system. The current computer network market offers dozens of options to choose from when selecting a NIC. There are types of network interface cards available, including Ethernet, Token Ring, Arcnet, and AppleTalk. Buyers of network equipment will find many models and brands for each of these types. Some network software packages will run on a variety of NIC types, while others have limited options. Some network products may either come packaged with NICs or specify a particular model, while others give the network implementor a wide selection of possibilities. Each category of NIC requires a particular type and configuration of network cabling.
Each class or type of NIC uses a particular network protocol. The network protocol relates to a standard set of rules and conventions used in low-level network communications. One does not need to know a great deal about how low level communications works to install or use a network, but one does need to make sure that the network software supports the network protocol and its corresponding NICs and cabling. The network protocols operate transparently to the user, or even to the network software. The software that implements these protocols usually resides in firmware on the network interface card itself.
Network Cabling. In order for a LAN to work, the microcomputers involved must connect to each other through some sort of cabling system. Network cabling comes in many different types including coaxial, shielded twisted-pair, unshielded twisted-pair, or fiber optic. The cabling system consists not only of the cable itself, but also the various connectors needed to connect the cables to the computers and other network equipment.
Many microcomputer LANs use coaxial cabling. This type of cable consists of a central wire surrounded by a layer of insulation, a layer of shielding, and then by an outer layer of insulation. Coaxial cable comes in many forms, each with varying thickness and electrical characteristics. If your network requires coaxial cabling, make sure that the cable that you buy corresponds to the specifications required by your system.
Fiber optic cable supports faster data communications than any other transmission medium, but comes at a higher price. With this type of cable, pulses of light transmit information rather than an electrical current. Fiber optic cable is a thin cable with one or more thin glass fibers in its core. To be used in computer networks, fiber optics require interfaces to convert the digital signal between its electrical and optical forms.
Twisted-pair cabling, also widely used for LANs, can come with or without shielding, and multiple pairs of wires may be bundled within a single cable. A twisted-pair cable may include a single pair of wires, or as many as 100 pairs. With shielded cable, each pair within a bundle may have a shield, and the whole bundle will have an outer shield. Unshielded twisted-pair cabling in recent years has become popular in computer networks. Because telephone systems use this same type of wiring, and therefore it already exists in most buildings, networks that use unshielded twisted-pair cabling can be established without significant cable installation expense.
Cable installation often requires considerable effort. However, many recently built libraries may have been built with computer networks in mind and will have various types of network cabling integrated into the building structure. If the network that you are implementing consists of a group of computers located in the same room or in adjacent rooms on the same floor, then you might be able to install the cabling yourself. If your network will span multiple floors or across the entire building, you will most likely need to delegate the task to a qualified electrician or cable installer.
Part of the cable installation process must include the testing of the cables and connectors before they are placed into service. For most cable types, installers can use time domain reflectometry (TDR) equipment to verify the cable integrity before and after installation. This equipment will detect the presence of breaks or flaws in the cables.
Network Communications Equipment. Some networks may require special equipment in addition to the NICs and cabling system. Such equipment might include repeaters, network hubs or concentrators, multiuser access units, gateways and routers. You are likely to need such network devices with larger networks, or when interconnecting networks.
A repeater allows networks to span distances greater than the specifications of a particular cable type. If, for example, your network requires a total cable length of 2000 feet and your cabling system allows only 1000 feet, you could divide your network into two cable segments of 1000 feet each and connect the two segments with a repeater. The repeater will amplify and retransmit all the data as it is passed between the two cable segments. Multiport repeaters allow several cable segments to connect together into a central hub. From the network software's point of view, such a network appears as a single unit, while the cables actually exist as individual segments.
Many types of networks require some type of central device to manage network communications. Token Ring networks, for example, require a multistation access unit (MAU). Each computer on the network must be connected to a port on the MAU. When the number of nodes on the network exceeds the ports on a single MAU, multiple MAUs can be chained together. Ethernet networks on unshielded twisted-pair cabling systems also require that each machine connect to a central hub.
While repeaters allow multiple cable segments of the same type to interconnect within a single network, bridges allow separate networks to communicate with each other. A repeater passes all information between the network segments. Information passes across a bridge only when its destination is not on the network in which it originated. Because of this capability, a bridge can help to eliminate unnecessary traffic on each network. Bridges can also connect some different types of networks. A bridge could, for example, connect an Ethernet LAN with a Token Ring LAN, or connect a broadband Ethernet LAN with a baseband Ethernet LAN.
Gateways allow interconnection of entirely dissimilar networks. When the protocols of two LANs differ so much that they cannot be handled by a bridge, then a gateway will handle the translation from one network format to the other. A gateway, for example, could be used to connect a Novell Ethernet LAN to an SNA network, or could connect an AppleTalk network to a TCP/IP Ethernet network.
In addition to the hardware components described above, a network also requires several layers of software components. These layers include the software drivers to the NIC, an interface between the computer's operating system and the network, and software that defines and controls access to network resources. Other optional software components of a LAN might include a menuing system, network administration and configuration software, or electronic mail or messaging programs.
When you buy a network operating environment, it will likely include most of the software layers described above. It is not necessarily the case that a network implementor will need to acquire each of these software components separately, but it is fairly important that network administrators understand the purpose and general concepts behind each software element.
Device Drivers. Most systems require a device driver that tells your computer about the particular hardware characteristics of the NIC. This driver may be loaded as part of the computers initialization in the CONFIG.SYS file, or it may be executed as a terminate-and-stay-ready program from the command line. This device driver will pass information to the computer about how the NIC uses interrupts (IRQ), Input/Output (I/O) addresses and memory blocks. Each NIC comes with a certain IRQ and I/O address pre-selected. If your computer uses that IRQ or I/O address for something else, you will need to change the options on the NIC and make corresponding changes in the way that you load the device driver for the NIC.
Suppose, for example, that you have purchased a Novell NE2000 Ethernet board to connect your microcomputer to the network. You read the documentation and learn that it comes set to use IRQ3 and I/O address 300. Your network requires that you load the device driver in your CONFIG.SYS with the entry: DEVICE=C:\NETWORK\NE2000.SYS. You find, however, that the computer system locks up because of this device driver. You then check your computer's documentation and learn that the second serial port uses IRQ3 and your CD-ROM interface board uses I/O address 300. To make the NIC work in your computer you must change jumpers on the board so that it uses alternate settings, probably something like IRQ 5 and I/O address 360. After making the changes on the board and reinstalling it in your system, you would then load the device driver specifying that the NIC now uses non-default settings. The command to load the driver might look something like: DEVICE=C:\NETWORK\NE2000.SYS /IRQ=5 /IOBASE=360. This represents one of the most common software problems that a network implementor will encounter while installing a network.
Network Operating System. Another layer of the network software environment are the various modules of the network operating system itself. What these modules do exactly and how they are loaded into the system vary greatly among all the network products on the market. Networks need software to allow the operating system of the microcomputer to interface with the resources on the network. In general terms, the network operating system extends the normal operating system of the computer so that it can address network resources as if they were part of the local system. The computer system needs to know which logical disk drives and printers are local and which it needs to request through the network. Most network operating environments employ a strategy of redirection so that the system channels requests for non-local drives and printers to the NIC, and the network services the requests.
One of the main concerns that network implementors should study relates to the amount of memory used by the network software. Memory used by network modules takes away from the memory space that other software applications can use. If the network software uses too much memory, users may no longer have enough memory to run some large software applications. This issue frequently arises in networks involving CD-ROM applications since CD-ROM drives require additional software drivers, and since the software to search the products tends to require a lot of memory.
Network Server Software. Network servers require software to enable them to provide and manage resources on the network. Providing resources is naturally more complex than using them. A server needs the same type of device drivers and network operating software required by workstations, but also needs the capability to define network resources, manage access to these resources so that only authorized users access each resource, plus it has to manage many simultaneous requests. The tasks required of network servers are so complex that almost all the high-end network products use an operating system other than MS-DOS as the native environment of the server. MS-DOS, designed as a single tasking, eight bit operating system, simply lacks sufficient sophistication to manage a large network server. Novell's 286 and 386 products use their own non-DOS operating system, Banyan/VINES uses UNIX, and Microsoft LAN Manager uses OS/2—all multitasking operating systems.
It is only the network products designed for smaller implementations that can rely on MS-DOS to manage network servers. Some of these scaled-down systems often allow the same microcomputer to be used both as a server and a workstation. But these products have limited features and performance compared to the ones that employ dedicated, non-DOS servers.
The network software that runs on a server will include management utilities that a network administrator will use to define and control all the resources provided by that server. One function of the network management utility is to define and set privilege levels for each person that will use the network. The network administrator assigns a username and password to all network users. The network operating system
will not allow a user to gain access to any resources until that user has entered the correct name and password.
The network management utility also defines the resources on the server. The network administrator will organize the disk space on the server into various directories, give network names to those directories, and will specify who has access to those directories, and whether that access includes the ability to read, write, browse, erase, create, or scan data in that directory.
Most network packages choose to give access by individual users rather than to specific microcomputers on the network. This access strategy allows a user to have the same level of access no matter what system on the network she/he happens to be using. In practice, however, each user generally accesses the network from the same system. In these cases, the network administrator will likely set up most of the workstations to log onto the network, enter the password and assign network drives and printers as part of the computer's automatic start-up procedure. With MS-DOS machines this automatic network sign-on would be done by placing network commands into the AUTOEXEC.BAT file. Once set up this way, each microcomputer user on the network never has to do anything special to get onto the network.
An example of typical AUTOEXEC.BAT file for a network might look like this:
@echo off ;don't display commands< prompt $p$g ;modify dos prompt path c:\dos;c:\network;c:\utility ;establish search path ne2000 /irq=5 /iobase=360 ;load NIC device driver netbios ;load NetBIOS extensions redir ;load network redirector net login \\server myname secret ;logon to network server net use e: \\server\mystuff ;set up logical E: drive net use f: \\server\deptstuf ;set up logical F: drive net use lpt2: \\server\@laser1 ;access laser printer menu ;run menu program
Most LANs will include some type of electronic communication facility. One of the great advantages to a LAN lies in the ability to communicate electronically with other members of your organization. Furthermore, electronic mail works much more quickly and efficiently than paper memos. Coworkers can often exchange information more easily through electronic mail than over the telephone. Electronic mail software may come bundled with the network operating system, or it might be an optional feature purchased separately.
One of the issues relevant to selecting electronic mail systems involves its capabilities for interfacing with other mail systems. If the network connects with other networks, then network planners should make sure that the mail systems of each network can accept and receive messages from the others.
Even though a computer operates on a network, the tasks required of the computer remain much the same as stand-alone systems. Applications software such as databases, word processors, and spreadsheet will need to operate on the network. However, software exhibits various levels of compatibility with networks. Some software packages do not tolerate a network as well as others. For example, many software packages that employ a copy-protection scheme will run only from a local hard disk and cannot be loaded from the network. Given the current proliferation of LANs, most software developers make great efforts to ensure that their products will run on all the popular networks. In fact, many software applications have special network versions specifically designed to take advantage of network services.
Network administrators should take special care to provide for frequent backup of data stored on network file servers. Once a network user stores data on a file server, it generally becomes the network administrator's responsibility to ensure the security of the data. Floppy disk backup may be inadequate for many large file servers. When network data exceeds 20 megabytes or so, some other media should be used for backups.
A very common method for backing up large file servers involves magnetic tape cartridges. A single tape cartridge can hold over 100 megabytes of data. Not only do tape backup subsystems store large amounts of data, but most can be programmed to backup the server automatically and unattended. Thus, a network administrator could configure the backup system to perform the backup each night after the users are off the system.
Good backup procedures will include multiple copies of the backup data. This adds another degree of data security since it takes into consideration the possibility of both a disk failure and a defective tape cartridge. Another safeguard involves storing a copy of the backup tapes in some building other than the one that houses the file server. If a fire or other catastrophic event destroys the file server, at least a copy of the data would be preserved.
Network Concepts and Models
While the preceding section described the more tangible pieces that come together to form a network, this section aims to highlight some of the concepts that apply to local area networks.
LAN Models: Server-based vs Peer-to-Peer
LANs come in many varieties. Some LANs are patterned after a centralized model while others follow a more distributed approach. For example, peer-to-peer networks work quite differently than server-based networks.
Peer-to-Peer networks operate democratically. In these networks, each workstation has the option of both contributing resources to the network as well as using resources on the network. Printers and hard disks attached to any workstation in the network can be defined as a network resource to be shared by all workstations.
One of the greatest advantages that peer-to-peer networks hold over server-based networks lies in their great flexibility. Any system on the network can contribute its drives and printers to the service of the network. This capability, for example, allows network administrators to establish network printers in many locations throughout the physical layout of the network.
Large peer-to-peer networks can become very complicated and difficult to manage. When each system on a large network potentially contributes disk and printer resources, the process of defining, maintaining, and organizing these resources can become unmanageable. Without high-performance servers, the load of a large network exceeds the capabilities of many peer-to-peer networking products.
Server-based networks, on the other hand, follow a centralized model. Computers in these networks are either workstations or servers. A server provides resources to the network while workstations use the network without contributing resources. In a typical large server-based LAN one might find a large central file server, one or more print servers, and dozens of workstations.
These distinctions are not absolute. It is possible to have a server-based network where some servers can double as a workstation. It is also common to have non-dedicated print servers in a server-based LAN where some individual workstations can have their printers defined as network resources.
The OSI Layers of a Network
No discussion of local area networks would be complete without mentioning the Open Systems Interconnect (OSI) model of network functions. The International Standards Organization created this model in order to promote the interconnectivity of networks through the standardization and isolation of network functions. This model divides the many tasks that comprise a network into distinct, clearly defined, layers.
In addition to its intended function as a tool for network designers, the OSI model has become a great pedagogical tool. With this model in mind, one can better understand the functions of the various components of a LAN. The OSI layers provide a reference in which to compare one network with another and to conceptualize how network communications operate.
The OSI model divides network functionality into the following seven layers:
Layer 1: Physical. This layer defines the characteristics of the raw electrical signals that travel across the network. It includes properties such as the voltage used for the electrical signals and the length of signal pulses. This layer does not concern itself with specifications of cable types and connectors. These issues fall outside the scope of the OSI reference model.
Layer 2: Data Link. These definitions concern the basic rules for the sending of information across the network. Basic error checking and correction occur on this layer, but the processes that occur within this layer are totally unaware of the content of the transmitted information. The data stream is broken down into blocks and each block is encapsulated with header and trailer information so that other data link layer processes can recognize the block, verify its integrity, and move it through the network.
Layer 3: Network. The main concerns here relate to establishing the routes that data blocks can take across the network, or from one network to another. In some networks the routes that the data blocks may take are static and require very little activity on the network layer. Other networks may have multiple routes possible between nodes and may change these routes dynamically according to the current load of network traffic.
Layer 4: Transport. This layer ensures that data get all the way from the source to the destination intact. This is the lowest layer that deals with end-to-end transmission of data.
Layer 5: Session. The session layer defines the rules for how two machines can communicate with each other across the network. It determines how systems can establish, maintain, and terminate communications between themselves. The session layer represents what goes on when a user logs into a file server, for example.
Layer 6: Presentation. One of the main concerns of this layer lies with conventions used in the representation of data. Issues such as the transmission of textual information in ASCII or EBCDIC, the format of integers and other abstract data types belongs to this layer. Services such as the encryption/decryption of information and data compression also belong here.
Layer 7: Application. This top layer of the OSI model is the only one that the users of the network see. This layer includes network services such as file transfer, electronic mail, remote printing, and terminal emulation.
Although future networks may eventually strictly follow the OSI model, the networks described in this book do not. The network components found in most current network products may span multiple OSI layers, or perform part of one layer and part of another. As network products emerge conforming to the OSI standards, a much greater degree of interconnectivity among networks can be accomplished than is currently possible.
The physical layout or topology of a network can take several forms. Different network products require different methods for connecting the nodes on the network together. In some networks each of the systems connect to each other, and in others each system connects only to a central hub. Network implementors need to know the various options concerning network topology and the advantages and disadvantages that correspond to each type.
One possible network topology is that of a bus. With this topology, used widely in Ethernet networks, each node on the network attaches to a single cable that spans the length of the network. Each end of the bus will have some type of terminating device and each node on the network must tap into the central cable that forms the bus. Figure 1.1 illustrates the simple bus topology.
Illustration of Simple Bus Topology
The bus topology works well for a cluster of network nodes located in close proximity to each other. But as networks become larger and more complex, a simple network bus may not be adequate. Limitations apply to the total length of bus cable segments and to the number of network nodes that can exist on any one cable segment. To alleviate these limitations, most networks allow branching of the bus or allow multiple busses to be combined into a central hub. Figure 1.2 shows multiple bus segments combined through a multi-port repeater.
One of the disadvantages of the bus topology relates to cable failures. When connected in a bus topology, all the systems on the bus segment will fail in the event of a cable break. Strategies such as network segmentation can allow network planners to mitigate the number of systems affected from any given cable problem.
Bus Segments connected through a Repeater
The star topology requires each network node to connect to a central hub. One of the significant advantages of this topology is that most cable problems will affect only a single system. Since each network node connects directly to the central hub independently of the other nodes on the network, a single broken cable will affect only one system, and will be easy to isolate. This network topology does, however, involve the installation of significantly more cable than a bus topology. Figure 1.3 shows a star network, illustrating that each system on the network connects to some central communications device.
Illustration of a Star Topology Network
A third network topology, that of a ring, consists of a bus network that has its ends connected to form a complete circle. Figure 1.4 shows a ring network.
Illustration of a Ring Topology Network
It is fairly rare to see a network cabling system follow a ring topology. Although Token Ring networks require that the nodes of the network combine into a logical ring, the ring exists within the communications equipment, usually a MAU. Most Token Ring networks follow a physical star topology even though the network protocol requires a logical ring.
Access Methods and Protocols
In order for a network to work smoothly, certain rules must govern how each workstation behaves on the network. These rules, or access methods, are transparent to users of the network, and are usually implemented in the network boards themselves. These access methods concern the rules that the network hardware must follow in order to transmit data on the network. The three most common access methods are Carrier Sense Multiple Access with Collision Detection (CSMA/CD), Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA), and token passing.
Three main types of network implementations prevail in the case studies in this volume, Ethernet, Token Ring, and LocalTalk. Each of these network implementations uses a different access method, Ethernet uses CSMA/CD, Token Ring networks use token passing, and LocalTalk networks use CSMA/CA. Besides these mainstream protocols, some network vendors may use their own proprietary protocols with nonstandard NICs and cabling.
Ethernet. Ethernet emerged early as a standard protocol for computer networks. The IEEE issued a standard, known as 802.3, that defines this type of network. Ethernet and 802.3, though very closely related, differ in some subtle aspects. Especially in cases where one network will interconnect with other networks, network implementors need to know whether their network follows the Ethernet conventions or the 802.3 standard, or whether it supports either. For the broad, nontechnical discussion of this chapter, the differences are insignificant, and the term Ethernet will be used consistently.
One aspect of the theoretical operation of a network protocol has to do with its access method. When multiple computers need to communicate simultaneously over the same piece of cable, some set of rules must govern the communication.
When sending data over Ethernet, the network takes the content of the data and encapsulates it with codes that will address it to its correct destination. The encapsulation process also incorporates codes into the packet so that the recipient will know that the network transmitted the data correctly. If the receiving machine detects an error in transmission, it automatically requests retransmission. These encapsulated blocks of data are called datagrams.
A broad characteristic of Ethernet is that it uses broadcast transmission techniques over a bus topology network. Computers communicate by broadcasting datagrams onto the entire network. The sending computer broadcasts its message throughout the network, but only the addressed recipient node pays attention to the content of the message.
The broadcast messages that Ethernet uses follow an access method called CSMA/CD, meaning Carrier Sense Multiple Access with Collision Detection. Under this access method each machine senses for the presence of an active signal on the network, and may broadcast at any time. When multiple computers, also called nodes, attempt to broadcast on the network at the same time, the datagrams become garbled, much like two people talking at once on the telephone. Although any machine may broadcast at any time, the rules of the CSMA/CD cause each sending node to test to see whether their message collided with another. If a collision occurs, then each node that contributed to the collision waits a random amount of time and then retransmits its datagram.
More practical characteristics of Ethernet relate to transmission speeds and cable types. Ethernet comes in several varieties. The two main divisions of Ethernet cable types are broadband and baseband. Broadband technology allows for using a wide bandwidth of transmission frequencies. Data transmission and video transmission may be performed over the same broadband cable. Further, separate, independent data channels may also exist on the same broadband. The characteristics of broadband Ethernet are defined by the IEEE standard 10Broad36. The more common form of Ethernet is Baseband Ethernet. Baseband Ethernet is limited to a single data channel. Baseband Ethernet may use one of several cable types, thick Ethernet (IEEE 10Base5), a thick rigid coaxial cable, thin Ethernet (IEEE 10Base2), a thin flexible coaxial cable, fiber optic cable, and recently, unshielded twisted-pair cable—the same cable used by telephone systems (IEEE 10BaseT). Ethernet's transmission speed and distance limitations and cost vary according to the type of cabling implemented.
Token Ring. Another widely used network protocol is called Token Ring. The IEEE defined this type of network in the 802.5 standard. All the nodes on a token ring network interconnect to form a logical ring. Token ring protocol use an access method called token passing to govern network transmission. A special transmission packet is passed from one node on the network to the next, and continuously cycles through all the nodes on the ring. A node may broadcast data only when it receives the token. It broadcasts by adding its data to the token as it passes it along the ring. Along with the data, the transmitting node marks the address of the intended recipient. Each node examines the token as it is passed to check whether or not there is any data marked with its own address. When the token arrives at the destination node, the recipient extracts its data and flags the token with an indicator that the data was received. The token is then available for additional data transmissions.
Token ring networks generally use shielded twisted-pair cabling. Each node on the network is cabled to a Multi-user Access Unit (MAU). As mentioned in the above discussion of network topologies, even though the logical organization of Token Ring networks is a ring, the physical topology is usually a star. The logical ring that the token cycles through resides in the Multiple Access Unit, not in the main cable system that connects the nodes on the network.
Proprietary. Besides the industry standard protocols described above, some network vendors may elect to use alternate methods. These nonstandard networks may perform well enough, but users of these networks will find it more difficult to interconnect with other types of networks.
Apple's LocalTalk is one of these proprietary networking implementations. Macintosh systems use a set of network protocols called AppleTalk that can be carried over Apple's own LocalTalk, Farallon's PhoneNet, or over Ethernet. Macintosh computers come with a built-in LocalTalk port, avoiding the need to install a network interface card. The LocalTalk cabling system uses shielded twisted-pair cables terminating with special connectors.
LocalTalk uses CSMA/CA access method protocol. CSMA/CA differs from CSMA/CD in that it uses a scheme to avoid collisions of data transmission on the network instead of correcting them after they happen. To avoid collisions, the network assigns each node a time sequence in which it is allowed to transmit. When a node needs to transmit, it must wait until its turn before it is allowed to access the network. Thus, there is a built-in delay for every transmission. LocalTalk transfers information through the network at about 230 kilobits per second, and requires a bus topology network. Apple's specifications limit the network to 32 nodes per zone and suggest a total cable length of less than 300 meters.
Farallon's PhoneNet network system has also become quite popular for networking Macintosh systems. This network takes advantage of the built-in LocalTalk ports in Macintosh, and uses the same CSMA/CA access method as LocalTalk, but provides interface devices that allow one to use standard telephone
wiring and connectors for the cabling system. In this way, Macintosh networks can be established in buildings without new cable installation by taking advantage of existing unused telephone wiring.
Ethernet may also be used to network Macintosh systems. Although this means that the network implementor will need to purchase an Ethernet NIC for each system and install an Ethernet cabling scheme, the resulting network will perform significantly faster than the LocalTalk or PhoneNet networks, and will have more options for communicating with other networks. More importantly, it allows Macintosh computers to use one of the well-recognized network standards described above.
This introductory chapter on network concepts certainly has not been an exhaustive discussion of the topic. Readers interested in acquiring additional information will find many books and articles devoted to LANs, ranging from general practical guides to in-depth technical works. The following publications listed sample some of the available literature.
Archer, Rowland. 1986. A Practical Guide to Local Area Networks. Berkeley, California: Osborne McGraw-Hill.
Bulette, Greg and Chacon, Michael. 1991. Understanding 3COM Networks. Plano, Texas: Wordware Publishing, Inc.
Comer, Douglas E. 1988. Internetworking with TCP/IP: Principles, Protocols, and Architecture. Englewood Cliffs, New Jersey: Prentice Hall.
Desmarias, Norman, ed. 1989. CD-ROM Local Area Networks: A User's Guide. Westport, Connecticut: Meckler.
Derfler, Frank J., Jr. 1991. PC Magazine Guide to Connectivity. Emoryville, California: Ziff-Davis Press.
Durr, Michael and Gibbs, Mark. 1989. Networking Personal Computers. 3rd Edition. Que Corporation.
Fortier, Paul. 1989. Handbook of LAN Technology. Intertext Publications. New York: McGraw-Hill.
Fritz, James S., Kaldenback, Charles F., and Progar, Louis M. 1985. Local Area Networks: Selection Guidelines. Englewood Cliffs, New Jersey: Prentice-Hall.
Hancock B. 1988. Designing and Implementing Ethernet Networks. Wellesley, MA: QED Information Sciences.
Laubach, Edwin G. 1991. Networking with Banyan VINES. Blue Ridge Summit, PA: Windcrest.
LaQuey, Tracy L., ed., 1990. The User's Directory of Computer Networks. Digital Press.
Madron, Thomas W. 1991. Enterprise-Wide Computing: How to Implement and Manage LANs. New York: John Wiley & Sons, Inc.
Marks, Kenneth E. and Nielsen, Steven P. Local Area Networks in Libraries. Westport, Connecticut: Meckler.
Martin, James. 1989. Local Area Networks: Architecture and Implementations. Englewood Cliffs: New Jersey: Prentice Hall.
Quarterman, John S. 1991. The Matrix: Computer Networks and Conferencing Systems Worldwide. Digital Press.
Ranade, Jay and Sackett, George C. 1989. Introduction to SNA Networking: A Guide for Using VTAM/NCP. New York: McGraw-Hill.
Sandler, Corey and Badgett, Thomas. 1990. Mac to VAX: A Communications Guide. Glenview, Illinois: Scott, Foresmann and Company.
Shatt, Stan. 1991. Linking LANS: A Micro Manager's Guide. New York: Windcrest. McGraw-Hill.
Sheldon, Tom. 1990. Novell NetWare: The Complete Reference. (Covers through Version 2.15.) Berkeley, California: Osborne McGraw-Hill.
Tanenbaum, Andrew S. 1988. Computer Networks. Second Edition. Englewood Cliffs, New Jersey: Prentice Hall.
Datamation: For Mangers of Information Technology Worldwide. Newton, MA: Cahners Publishing Associates.
LAN Technology: The Technical Resource for Network Specialists. Redwood City, California: M&T Publishing.
LAN Times: McGraw Hill's Information Source for Network Managers. Midvale, Utah: McGraw Hill, Inc.
Network Computing: Computing in a Network Environment. Manhasset, New York: CMP Publications.
Network World: The Newsweekly of User Networking Strategies. Framingham, MA: International Data Group.
Telecommunications. Norwood, MA: BPA.