A communication circuit, as it pertains to a WAN applications, is basically a phone line, and is generally provided by the telephone company. The type of line can be anything from an analog dial-up line running at 9,600 bps to a 155,000,000 bps OC-3 fiber optic connection. Between these two extremes is a solution for nearly every requirement. This page will present the most common line types currently in use, and provide guidance as to what applications to use each type with.
Different types of lines provide different speed capabilities, and usually the faster the line, the higher the cost. However, it is not a good idea to base the decision of line type strictly on cost, as a system which operates at a slow speed may cause such a loss of productivity for the users that it actually ends up costing a company more than a higher speed line.
The first step in designing a WAN is to select a line type, and first step in selecting the line type is an analysis of what the WAN is actually going to do. It is not a good idea to match the application to the line type, rather it is better to start with the requirements of the application and select a line type which gives the best price/performance tradeoff. Often, the decision as to which type of line to use takes longer to make than the rest of the design.
Some of the factors which determine the bandwidth which a WAN will require are number of remote users, whether or not file servers will be installed at the remote sites, and where application programs will physically reside. All of these help to determine how much data the users will actually transfer across the circuit. It is also important to keep in mind "hidden" sources of traffic, such as network print jobs and protocol overhead.
An example will help to clarify this concept. Let us assume that a company has two offices named Main and Remote connected with a 56 kbps DDS line. The Main office has one file server, and the Remote office has three clients and one networked printer. All application programs physically reside on the workstations' hard drives, and the file server is used only as a place to store shared data files. At first glance, this network may appear okay. However, the users at the remote office are printing large proposals for clients. If a print job is 3 megabytes in size, then we have to send 24,000,000 bits of information across the line to the file server, which spools the print job until the workstation is finished generating it. The server then sends these 24,000,000 bits back across the line to a print server located in the printer.
Over a 10 Mbps LAN, this process only takes a few seconds. However, a 56 kbps line has only 1/178 the speed of a LAN, so the process takes 178 times as long to complete. Therefore, the print job will actually take over 14 minutes longer than if it did not have to go through the network. The other two users at the Remote office have to compete with this large print job for bandwidth on the 56 kbps circuit, and their access will be slowed considerably until the job finises printing. Also, their demands for bandwidth will slow the print job down. Therefore, every time someone prints a document, the entire network slows to a crawl for ten to fifteen minutes, and no one can get anything done during this period of time.
Funnel Restricting Flow Rate
Figure One shows this problem in different terms. Here, assume that the data needing to be sent is water, and that water must go through a narrow-necked funnel. The neck of the bottle is 3" in diameter, but the funnel's neck is only 1/2" across. To avoid overflowing the funnel, we must slow the rate at which we pour the water from the bottle, and effectively the bottle's neck has been reduced to the 1/2" size of the funnel. Eventually, all of the water will be transferred through the funnel, but it may take a long time.
Two solutions to this problem are readily apparent. The first solution is to upgrade the 56 kbps DDS line to a 1.544 Mbps T-1 line. This increases the bandwidth available between the sites by a factor of 27. Therefore, the 24,000,000 bits which need to be sent for the sample print job will only take 30 seconds to travel to the file server and back to the printer, as opposed to 14 minutes in the first scenerio. Using our earlier analogy, this can be thought of as "opening up the funnel." Also, it is unlikely that the PC generating the print job and the print server will actually process the data fast enough to saturate the faster link, so the other users of the system will not notice any slow down of the network while a print job is occuring.
A second solution is to install a second file server at the Remote office, and use it for data files and print jobs which will only be used at that office, and use the Main office's file server only for data which must be shared with the entire company. This keeps the print jobs from even being sent across the circuit in the first place. Since the load placed on it is now far lower than it was originally, the 56 kbps DDS line is now adequete.
Either approach is acceptable, and the choice as to which one to use depends on the cost difference between them. In other words, if the full T-1 line costs more over a certain period of time than adding a server would, then the second solution is preferable. If adding the Remote server would cost more, then it would be preferable to upgrade the line. Every situation is different, and needs to be investigated to find the optimum solution.
Since the first step to designing a WAN is selecting a type of line, we should touch on the types of lines which are commonly available. Please note that availability on these technologies varies by area, and the reader should contact his or her carrier to determine what is available in a specific geographic area.
Below are descriptions of various types of communication lines. This list is not all-inclusive, nor is intended to be an in-depth discussion of each. Rather, it is a very quick primer on the most commonly used and talked-about technologies available at this time.
Full T-1 lines can be rather expensive. However, they can also be a very cost effective way of interconnecting two offices, especially if some channels are used for voice and others are used for data. Such a setup can eliminate long-distance phone and fax costs between offices.
Some users do not need the full 24 channels, and can get by with a less expensive Fractional T-1 line. This is physically the same as a Full T-1, except the carrier ignores all channels except the ones the customer is paying for. For example, if the customer orders a 12-channel Fractional T-1, the carrier will forward any traffic on channels 1-12, and will not forward anything recieved on channels 13-24.
Frame Relay is a protocol used for a packet switched network. In this type of setup, each user's site has a single dedicated line between it and the nearest point of presence for the Frame Relay service provider. The user sends all inter-site data to the Frame Relay carrier, who forwards the data through its network to the office where the user's remote site is connected to the Frame Relay network. These connections are normally not distance-sensitive in terms of pricing. The user pays only for relatively short dedicated circuits ("Feeder Lines") to the Frame Relay provider for each site, and a rather low cost for the service itself.
Frame Relay is not automatically going to be less expensive than a dedicated line at the same speed. Generally, Frame Relay saves money in situations where either a long distance must be covered, or many sites need to be interconnected. It is generally not the best choice for short distance one-to-one connections, as the cost of the two feeder lines may be more than the cost of a dedicated circuit between the two sites.
Before settling on Frame Relay, it is a good idea to contact the carrier and obtain pricing for both a dedicated T-1 or 56 kbps DDS line solution and a Frame Relay solution to the same situation. Then, compare the two to determine which is more cost effective.
SONET is an acronym for "Synchronous Optical NETwork". It is a very high speed fiber-optic based technology running at speeds of OC-1 (51.8 Mbps) to OC-64 (3.31 Gbps). This technology is far faster than even T-3, but is also far more expensive. It is generally only affordable by the largest corporations, and is generally cost effective for only the most demanding applications.