In the previous chapter, I discussed the basics of setting up a local area network (LAN) and providing access and services for in-the-office and remote users dialing in by modem. This chapter discusses how to connect one or more LANs over a large geographic area to create a wide area network (WAN). Generally, a WAN is created when you utilize the phone company's data networking services to link LANs together. Some LANs are so huge that they would seem to be a WAN; I have worked on LANs that consisted of miles and miles of cabling, but here I am talking about actually connecting two or more LANs together with a phone company circuit, not just installing a large LAN. In some situations, you could connect two LANs together across a large university or hospital campus using two modems connected to servers. This could be considered a WAN, but for purposes of this discussion I am referring to the high-speed digital services usually offered by the local and long distance telephone carriers.
A WAN provides other LAN users in distant locations with the same information and services that your local LAN users have. If you have offices across town or across the country that must share services, such as inter-office e-mail or database information, you need to link the LANs over a WAN. Many companies have found that e-mail is a great way to pool resources, share ideas and distribute information about the company's business. If you maintain a central database or other applications, like shared documents that all personnel must access, you will need to set up a WAN, especially if you intend to expand your offices into other cities.
There are almost as many WAN types as there are LAN options. A very basic WAN is nothing more than two computers calling each other over a modem connection. A more complex WAN may involve connecting hundreds of offices together with high-capacity digital circuits. Usually a WAN takes the form of something between these extremes. The actual circuits and services used to build a WAN are determined by the type of applications being shared across the WAN link. The speed of the circuit used for wide area networking may be driven be the number of users sharing that line and/or the type of LAN traffic (that is, applications) sent over the line. The kind of circuit is usually based on the number of sites connected and the applications shared over the link. In the following section I discuss the most prominent WAN services in use today and describe the services in some detail.
The 56Kbps DDS has been the standard in WAN connections for many years. Many organizations have used this type of circuit to connect offices together. As the name implies, this type of circuit can transmit data bits (1s and 0s) at the rate of 56,000 bits per second. The 56K circuit is a point-to-point circuit; this means it connects one site to another or one LAN to another. It can be a permanent, or nailed, circuit between two sites, or it can be a switched circuit that is dialed up as needed to other compatible 56K services. This type of circuit is sufficient for passing files or e-mail and even for running applications on the server for a few users. However, if you have 80 users all accessing a database over this circuit and sending e-mail, the 56K circuit is too slow; users will experience long delays in sending and receiving data.
A T-1 (also called DS-1 or Digital Signaling-1) is a grouping of 24 64Kbps channels that create a 1.5-megabit-per-second circuit. Like the 56K circuit, a T-1 is a point-to-point service. T-1 circuits are widely used and make up a large percentage of both data and voice and video WANs. A T-1 is ideal for linking offices together at near-LAN speeds. If you are running an ethernet LAN at 10Mbps, the T-1 service will provide better than one-tenth of your LAN transmission rate. This may seem slow, but considering an average ethernet LAN only runs at about 20 percent or 30 percent capacity, the T-1 is actually closer to one-third of your LAN transmission speed. A T-1 can be nailed or switched, as with a 56K line. In addition, the T-1's channels can be divided among multiple sites. For example, your Washington, DC office could have eight channels in the T-1 going to Chicago and the remaining channels going to New York. This WAN effectively would tie all three office networks together.
A T-3 (also called DS-3 or Digital Signaling-3) is a grouping of 28 T-1s to create a 45Mbps circuit. This type of service is expensive and is employed primarily when a large-capacity data transmission path is required. In a typical WAN implementation, you will not need this service, but you may hear it discussed as part of a T-1 network implementation. T-1s are multiplexed, or bundled, together into a T-3 for easier routing through the phone company's network.
Frame relay is a service designed to operate on a 56K or a T-1 type circuit. It provides many virtual channels inside the circuit. You could take a 56K circuit and use frame relay to create several channels to other sites so that instead of one 56K circuit going from Washington, DC to Chicago and another going from Washington to New York, you would simply install one 56K frame relay in Washington with virtual channels to the other locations. This is one advantage to frame relay. Another advantage is that it is typically cheaper than standard point-to-point circuits.
The disadvantage is that the actual through-put in a 56K frame
relay circuit is less then 56K. The phone companies will offer
a committed information rate (CIR) to guarantee the minimum through-put.
You may be able to send bursts of data above the CIR but not for
sustained periods of time. You will always have the minimum transmission
speed guaranteed by the CIR rate you purchase.
EVN Records has offices in three states on the East Coast. These offices currently share demographics and sales information over the phone and through the postal system. It recently has begun working with a firm in California to set up its artists' touring schedules. Meanwhile, upper management wants to reduce the long-distance phone expenses among the three East Coast offices; with their new West Coast associates, the long-distance costs will skyrocket. Upper management is interested in using the Internet to set up web pages for their artists. The network managers have convinced upper management that the information among the East Coast offices can be shared electronically by e-mail and that linking their Windows NT networks over a WAN will cost less than EVN's current long-distance telephone charges. The California office only needs to share e-mail with the East Coast offices; it doesn't need to be linked to the WAN. Because the three East Coast offices only have a few users on each LAN, it is decided to use 56K Frame Relay to link the three East Coast office LANs. These three offices each have a Windows NT workstation, and these networks will be bridged together into one LAN using the NT server at the main office. The main office also will have a fractional T-1 connected to a local Internet provider for e-mail and putting the Web pages online. The other East Coast offices will share the fractional T-1 connection out of the main office to the Internet. The California office is already connected to a local Internet provider, and the Internet will serve as the WAN for the transfer of e-mail between California and the East Coast offices (see Figure 11.1). |
Figure 11.1 : A WAN implementation with Internet access.
Sonet (Synchronous Optical Network) is the latest in high-capacity fiber optic WANs. Sonet transmission rates usually start at the OC-3 (Optical Carrier-3 level), which is roughly the equivalent of 3 T-3s or a 155Mbps rayed circuit. Sonet can be deployed at the OC-12 level, which would be in excess of 600Mbps. Services like Sonet are deployed in environments such as medical imaging, intensive computer-aided design applications, or combinations of video, voice, and high-bandwidth LAN applications. There is little justification for using Sonet even if you are linking two 100Mbps LANs together, as the cost of Sonet versus the performance benefits is usually prohibitive unless your applications demand it. In any event, the deployment of Sonet service integration to your LAN is outside the scope of this document. If you have a requirement for such high transmission rates, speak with a network consultant about Sonet services.
SMDS (Switched Multimegabit Data Service) is somewhat like frame
relay in that it provides multiple virtual channels within a single
SMDS service so that each can be routed to a different location.
One of the main differences between frame relay and SMDS is that
SMDS transmission rates start at the T-1 level (roughly 1.5Mbps)
and go up to the T-3 level around 45Mbps. If you have a high-bandwidth
data application that is being shared among three sites at T-3
transmission rates, SMDS is probably a good choice.
Goodbody Hospital is a small hospital outside of the city. The two larger hospitals in the city have agreed to share their medical imaging staff expertise with Goodbody Hospital, but the doctors in the city do not have time to drive
out to Goodbody Community Hospital just to review CAT scans. Therefore, the doctors decide to deploy a high-speed WAN that will link the LAN at all three locations. The WAN service must support the transmission of CAT scan images from Goodbody over to
the other two hospitals' networks, where the images can be analyzed and compared against other CAT scans stored on the file servers at the two city hospitals. The CAT scan images are huge files that can be hundreds of megabytes or gigabytes each. A
T-1 would be far too slow for the efficient transmission of these size files. A standard point-to-point T-3 would be effective, but to link all three hospitals the doctors would need several T-3s, which is an expensive option.
In this case, the SMDS service was the solution. With this service, the hospitals could switch the full T-3 bandwidth of 45Mbps among the three sites as needed. The hospitals needed to purchase a router for each site to direct the LAN traffic over the SMDS network. The phone company is providing the SMDS access equipment as part of the service (see Figure 11.2). |
Figure 11.2: An SMDS network directs LAN traffic among three locations.
ATM (Asynchronous Transport Mode) is a transport service that will run over Sonet networks. In the previous chapter, I discussed ATM as a LAN technology; however, it makes the transition to WAN service easily. Indeed, using ATM to connect the average LAN workstations is somewhat like using a 747 jet to transport a letter to a nearby town: You could do it, but it isn't generally necessary. In large campus LANs, ATM provides a good backbone used to link multiple FDDI LANs together into one larger LAN. For those environments that need to extend the high-bandwidth capacities over greater distances than just a campus LAN, ATM is an excellent way to integrate and share not only LAN services but audio- (phone systems) and video-conferencing. ATM is still not being used widely for WANs, but many phone companies are offering some type of ATM WAN. In the future, ATM may become the standard delivery method for all computer communications, including transmissions headed into the home; television, Internet access, banking, shopping, and video phones may all be sent via ATM.
After you have determined that you need a WAN service, you must decide what applications will be running over the WAN circuit and how many people will be using them. If you have a Novel network of 200 users in one office and you want them to share e-mail and a database with a 50-user Novel LAN across town, you must figure the number of users accessing the database and sending mail simultaneously across the WAN. If only 5 users of the 50 in the remote office will be accessing the database at any given time, a 56K WAN circuit may be adequate. However, if all 50 users will be accessing the database at the same time, plus sending e-mail, you should plan for a T-1 or a fractional T-1, which is only 512Kbps or 768Kbps. There are no hard and fast rules about how fast a WAN circuit should be. It may require that you talk to the users of the applications to ascertain what their usage patterns are before you size the WAN circuit. A good general guideline is 128Kbps per 25 frequent users.
Connecting the WAN to the LAN is usually accomplished in one of two ways for data traffic that you will find in a LAN: bridging or routing. Both of these networking services are achieved by means of a hardware device installed on the connecting LANs. The bridge or router equipment is connected to your LAN just as the computers are connected to the LAN. The bridge or router also is connected to the WAN and acts as the "server" providing a path to the other LAN at the far end of the WAN.
A bridge is a device that passes all data on the ethernet, token ring, or whatever type of LAN you have over the WAN to the other LAN. Essentially, the LANs that are bridged over the WAN appear to be all one LAN. The advantage to this type of networking is that you can set up all the devices in both LANs as if they were all local to one another. All the numbering used to provide computers on the LANs with unique network addresses reflects a single LAN. Having a single network may simplify administration, as there is only one network to "look after." However, the disadvantage is that there may be data being sent over the bridge that should stay local, thus taking up valuable "space" on your WAN circuit. Bridges use filtering and a bridging algorithm to learn which network addresses are on the LAN and which are on the WAN, but there generally is more traffic passing through a bridge than with a router. Bridging is appropriate for small networks on either side of the WAN, but if you have large networks on either side of the WAN, you probably should use a router.
Routers are devices that are installed on the LAN much as bridges
are; a router connects to both the WAN and the LAN. The difference
between a router and a bridge is in the way it handles the data
it receives. In the bridging world, data bits on the LAN (called
packets) are passed across the WAN with minimum effort on the
bridge. The bridge doesn't look at the packets very closely to
examine the data, because it doesn't care what the data is; it
just passes the packets over to the other side of the WAN. Routers,
on the other hand, examine the data sent in the packets to see
whether it needs to go over the WAN or if it should stay in the
LAN. Think of a data application, e-mail for instance, as if it
were a letter being sent over the LAN. It is put into a Novell
or TCP/IP envelope (or whatever network you have), which is addressed
by Novell or TCP/IP to show you sent it and who the recipient
is. That envelope (or packet) is then "stuffed" into
an ethernet or token ring envelope (whatever type of network you
have) and is addressed again by the ethernet card to show from
which computer it came. A bridge doesn't care about the Novell
or TCP/IP addressed packet; it only looks at the ethernet or token
ring address. A router, however, will "open up" the
ethernet envelope to see the addressing on the packet inside the
ethernet packet. The router provides an additional criteria for
deciding where a particular packet should be sent. This is less
important in a small network, but in a large network this function
is critical to efficient use of both WAN and LAN resources. If
you are connecting your network to the Internet, you must have
a router. The network addresses you set up on your LAN to get
on to the Internet will be unique to your network and must be
routed as a separate network to the Internet, not bridged.
The ABC company makes widgets. ABC says it doesn't produce your run-of-the-mill widgets, but rather one-of-a-kind, top-of-the-line widgets. ABC's business has really taken off, and it has decided to open an office in Normal,
Illinois, to facilitate the growing orders for widgets in the Midwest. All the company's inventory and shipping information is stored on the Novell LAN file servers in the main office in Fort Lee, New Jersey. The new Illinois office has a Novell LAN
but needs access to all the information at the headquarters LAN. ABC has decided to implement a WAN so it can be linked to corporate headquarters.
The administrators found, after talking with upper management, that the ten-user LAN in Illinois will grow to more than 50 in the next six months. The users on the LAN in Midwest office will spend most of their time entering orders into the database in Fort Lee. The orders for widgets must be checked against the existing stock, ordered, and shipped using information stored in the central database with little delay. The administrators decide to purchase a point-to-point fractional T-1 from a long-distance company. The T-1 will provide a 256Kbps transmission rate initially and, with a simple call to their provider, the administrators can up the bandwidth to 512Kbps or higher when the additional employees join the Illinois office. Because the Novell network in Fort Lee is already set up, another Novell purchase is planned for the Illinois office. This new Novell network is a separate network with its own numbers so the administrators in Illinois can number the computers on their LAN with out having to check whether a particular address is already in use by a computer in Fort Lee. Because these are separate networks, they cannot be bridged, so administrators have selected a router with an internal CSU/DSU so they can connect directly to the WAN circuit. The router also supports Novell protocol routing (called IPX routing) (see Figure 11.3). |
Figure 11.3: A Novell network with a WAN in place.
To access any WAN service with a router or a bridge, you will need an access device. This equipment will vary with the type of WAN service you use. Each WAN service requires some hardware device, and some services, such as frame relay, requires software as well. The access equipment connects to the bridge or router with a V.35 cable, for example, and then connects to the WAN circuit using whichever cable the WAN service requires, like a fiber or twisted pair cable. When you purchase your bridge or router you must know the type of WAN service you will be using and make sure you have technical support for that service. Frame relay, as an example, can be delivered over a 56K circuit, which almost any bridge or router can support, but frame relay access requires software on the router or bridge; so you must match the equipment to both the circuit and possibly the service on that WAN circuit. Your WAN service provider should know how the WAN circuit will be delivered (that is, on fiber or twisted pair) and what access equipment you will need. Many phone companies will rent or provide the access equipment as part of the service. You then only need to supply the router or bridge with the appropriate WAN port option, such as V.35, RS-449, EIA-530 or RS-232 and software, if required.
For accessing a 56K circuit, use a DSU (data service unit). The DSU connects to your router or bridge and to the phone company's circuit. The DSU formats the data from the bridge or router so it can be transmitted over the 56K line. Each bridge or router that connects to a 56K circuit must have a DSU-type access device. The DSU can usually be purchased through the vendor that sells the router or bridge. Many routers and bridges come with a 56K DSU-type interface so the WAN circuit can be directly plugged into them.
To access T-1s, you need a CSU/DSU (channel service unit/data service unit). This device provides access to all 24 channels on the T-1, and like the DSU, the CSU/DSU formats the data from the bridge or router so it can be transmitted over the line. Many routers and bridges now come with internal CSU/DSU-type interfaces so you can plug them directly intothe T-1.
To access a T-3 you will probably need a fiber optic access device. Most T-3 services are now delivered on fiber optic cable, and the access device will be a fiber interface on the WAN side and more than likely a V.35 connection on the LAN side. Most routers and bridges do not support T-3s directly and require a separate access device.
Frame relay services typically are accessed over T-1s or 56K lines and require the appropriate DSU or CSU/DSU. Again, many routers and bridges support direct 56K or T-1 connections; however, the router or bridge you choose must support frame relay service in its operating software.
To access SMDS, you will probably need a fiber optic access device. SMDS is usually delivered over a T-3 or fractional T-3 using fiber optic cable, and it requires the appropriate SMDS/T-3 access device on the WAN side. Like frame relay, you need a router that supports SMDS in its software. You need the support of the SMDS provider, which may furnish the SMDS access device, in making this connection.
Sonet WANs also require a fiber optic access device. This device is specific to the Sonet WAN, and if it is not supplied by the Sonet access provider directly, your provider should know where you can get a unit that will interface with your LAN and the Sonet service.
ATM is usually delivered over Sonet in WAN applications, so you will need the Sonet access equipment described in the "Sonet" section. ATM has been deployed over T-3s in some environments, though, and in those instances, the appropriate T-3 access equipment is used. ATM is somewhat like the frame relay or SMDS services in that it is a networking service delivered using another networking technology. Accordingly, just as with frame relay or SMDS, you need a router or other system that provides the software support for ATM.
After you have decided on what kind of connectivity you want, what speed you need, and what kind of hardware and software you need to accomplish that speed, the issue remains: What do I do with a WAN?
The nice thing about a WAN is that, when it is set up correctly, it functions almost exactly like your LAN. You can set it up so that the difference between connecting to a computer across the country is not all that different from connecting to a computer down the hall. The only difference might be the speed of the connection.
For the purposes of your intranet, a WAN can help you solve several problems at once.
A WAN can help you to improve connectivity among the offices as well as a connection to the Internet. If you have two or more offices connected by high-speed data lines, only one needs a connection to the Internet. The other offices can connect through that single connection.
With a WAN, your intranet server can be located inside your own private network. Technically you don't even have to be connected to the Internet for several offices to share documents or communicate by a web or e-mail interface. Even if you have an Internet connection, your web server can be located inside your firewall and can be made accessible only to users on the WAN; in this way you can make your intranet information very secure.
It also is possible to use these lines for telephone and fax communications among two or more offices. It is possible that your organization will save more money in long-distance phone charges with your WAN than the data lines cost.
The bottom line is that when you have a working WAN in place and
are running TCP/IP, setting up an intranet Web site for those
machines is pretty easy-as easy as installing Web server software
on a machine and giving the computer a name. Instant intranet!
M&J had had a WAN up and running for some time. The Philadelphia office and the Washington, DC office had been connected by an ISDN line for more than a year. The WAN needs were twofold. First, to connect the Raleigh office to
the WAN, and second, to connect the three offices to the Internet.
To connect the Raleigh office to the WAN, the simplest and least inexpensive alternative was to run an ISDN line directly to the Washington, DC office. This option did not provide the best performance but was the least expensive. The performance was hurt because the intranet server, as well as the Internet connection, were both in the Philadelphia office, so to access either one, the Raleigh employees had to go through the Washington, DC office. It was decided that this was the best option because Raleigh had the fewest employees and was not likely to have too much problem with the speed of the connection. The Internet connection from the Philadelphia office was a little more difficult to set up. Finding a local service provider and getting the line installed was easy, but configuring the router was a little more difficult. It took several days to get the connection to work right with M&J's system. |
The SGAA's WAN needs were very simple: a T-1 connection to the Internet. Their Internet service provider (ISP) installed the CSU/DSU and their local telephone company installed the line itself. It took a couple trips by the ISP techie and a little tweaking from the consultants hired to build the intranet to get it to work right, but it wasn't too much of a problem. |
Anyone who maintains a central database or a central resource center and who needs to service offices that are located geographically apart from each other should consider utilizing a WAN. If you begin with a clear goal of what your WAN needs to accomplish, the WAN will began to build itself. Use the information in this chapter as a reference guide to making the most beneficial choices for your WAN. With a grasp of the framework and a knowledge of the basic connecting tools, your WAN soon will allow you and your users to perform your jobs more efficiently.