Computers within a department or campus often communicate over a local area network, or LAN (see Figure 1). These networks are fairly small, in terms of physical distances, usually limited to a single building. The three basic categories of equipment that are found in a local network are workstations, servers, and gateways.
Users access networks through terminals and workstations. A terminal is simply an I/O device; the only processing it does is limited to presenting information for display on the screen. Simple (``dumb'') terminals do little or no formatting, so everything that is displayed is arranged by a computer system and every keystroke is passed back to the computer. For example, if you make a mistake while typing a line and hit the backspace key, a terminal just sends the keystroke to the computer you are using. The computer will update an internal buffer and send the terminal a command that will cause it to move the cursor to the left by one space. Until a few years ago terminals were connected directly to computers, but now it is common for them to be connected to a communications server, as shown in Figure 1. The server then uses the local network to connect the terminal to any machine on the network, giving users more flexibility.
An ``X terminal'' is almost a stand-alone computer by itself, with an internal processor and a few megabytes of memory. The processor displays windows and handles mouse and keyboard input locally, relieving a compute server from doing these low-level tasks. An X terminal rarely, if ever, runs an application itself, though; instead it just draws windows and menus and handles the user interface for applications running on other machines. The name comes from the X window system, a common protocol used by window-based applications (Section 2.3).
Workstations are single-user computer systems. They range in complexity from simple personal computers to advanced processors with sophisticated graphics and vector processing capabilities. Their common attributes include the fact that they can operate as stand-alone computer systems and usually have a ``point-and-click'' (window-based) user interface. Most often the person using a workstation will sit at the console and use the main keyboard and mouse. Unix workstations allow several users to log in to the system. In a networked environment it is not uncommon for additional users to log in by sitting at a terminal and using the communication server to connect their terminal to the workstation.
The second category of machines on a network are servers. As their name suggests, these nodes provide services to other nodes and are shared by many users. Printers are a good example: rather than connect a printer directly to one computer, and forcing each person to use that computer in order to print a document, a printer with a built-in network interface can be connected directly to the network and used by any system. Other common types of servers are file servers and compute servers. A file server is often just a workstation with extra disk space, but most workstation manufacturers build special server configurations for higher performance and better reliability. Mainframes and supercomputers that are likely to be used by computational scientists are usually compute servers on a local network. The term ``compute server'' reflects the fact that mundane jobs like editing files and reading mail are done on workstations and microcomputers, but users can run computationally intense jobs on these special purpose high performance systems.
The third category of network equipment is an interface to another network. These interfaces are either gateways or routers. A gateway translates information between the formats used by two different types of networks. In the early days of local area networking there were few standards and a wide variety of performance levels, and thus different protocols and services evolved in different settings. For example Apple microcomputers could be connected via an Appletalk network. If these machines were to be linked to a set of Unix machines, one of the Unix systems typically ran a software package that translated Appletalk protocol packets into the format used by the Unix systems.
A router simply forwards packets of information between networks that use the same protocols. In a packet-switched network each packet of information contains the network address of its destination [3][6]. The router monitors all network traffic and forwards all packets that are addressed for machines not on the local network. The most common use of a router is in building a hierarchical network. A local network within an academic department will have a router that connects to the campus-wide network, which in turn will have a router or gateway to a regional network. Routers are also used to improve performance within a local network. If network traffic is too high, the network administrator might split it into two separate networks connected by a router. Routers are invisible to users, so in most situations sending information to a machine elsewhere in the department or on campus is no different than sending information to the machine in the lab next door.
The remainder of this section is an overview of the types of services typically provided by a local area network and how they impact the way you work. The examples will typically be from networks of Unix systems because Unix is so predominant in university and government labs, but the main concepts also apply to most other types of local area networks.
