Network How-To

What every non-techie needs to know about networks

By Craig Lyndes

Craig Lyndes is the computer support technician at Champlain Valley Union High School in Hinesburg, Vt.

So you have a technology plan--complete with a proposal for a local area network (LAN)--and you've cajoled, finagled, and persuaded people to give you a budget. Now what?

If you're pinching pennies, or if you simply want to stretch your technology dollars a little bit further, check out the technology world's equivalent of sweat equity, and do some of the installation yourself.

The advice that follows is based upon my experience. In the last 10 years, our network here in Champlain Valley has evolved from a single computer and printer to our present network of more than 250 Macintosh and Windows PCs. In addition, all five schools in our district are connected to Champlain Valley through a leased-line wide area network, and we are connected to the Internet through a 256K fractional T1 line.

In the advice that follows, I've included my successes as well as my mistakes, and I've mentioned technologies others have told me are promising. Take what is useful, and feel free to disagree; a number of us do-it-yourselfers are finding new solutions to technology problems every day.

You can hire a contractor, of course. And at times I'll recommend that you do. But if the only thing that stands between you and a school or districtwide network is money for installation, you might want to consider some of the strategies I mention here. Simply break out the punch-down tools (more on this later) and begin.

Network basics

The whole goal of a network is to get information from one computer to another. That information--whatever it might be--travels in packets, which are like envelopes with addresses placed on them.

Say, for example, that you're sending e-mail to another computer on a different network. Once you hit the "send" button on your e-mail program, your e-mail application calls the networking software running on your computer and says, "Here, deliver this message to Joe in Alaska." The networking software is given the address of the computer in Alaska; but at this time, all your computer can do is communicate with other devices on its local network.

However, the networking software knows there is something called a router on the network, and it knows the router can send stuff anywhere on the Internet. So the software sends the whole mess in an envelope addressed to the router and lets it figure out how to get the message to Joe.

When the router gets the envelope, it tears off the outside envelope and throws it away. Looking at the inner envelope, it sees it needs to find the address of the router on the network connected to Joe's computer in Alaska. The router then puts the envelope with the message and Joe's address into another envelope addressed to the router in Alaska.

When the router in Alaska receives the envelope, it tears it open and reads the address on the inner envelope. This router knows the address of the network card in Joe's computer, so it puts the message in a final envelope and sends it to Joe.

When Joe's computer receives the envelope, it takes the message out of the inner envelope and puts it in Joe's mailbox.

All of these envelopes have to be understandable to the computers that receive them. Computers, after all, are very literal; they don't do well with unclear instructions. So for this system to work, you need clearly understood rules for forming and addressing packets. These rules are called protocols--and they come in several "layers."

The lowest layer--called the physical layer--actually transmits data over the wire. The next level up is called the data-link layer, which is a way of wiring computers together. (Different kinds of data-link layers use different hardware in computers and in hubs, which are devices that send any data they receive to any computer connected to them. Different kinds of data-link layers also run at different speeds.)

The next level is the network layer. These are the rules the networking software uses to assemble the packets and addresses--sort of like the language computers use to communicate with each other. (Some examples are Novell's IPX, Apple's AppleTalk, and the Internet's TCP/IP.) These upper-level protocols can be put in any of the lower-level envelopes; for example, you can connect two Macintosh computers with any of the hardware-layer protocols; the messages and addressing will always be AppleTalk no matter how they are transported.

You can also have different protocols zipping around on a single network. Here at Champlain Valley Union High School, for instance, our network has AppleTalk, IPX, and TCP/IP packets passing between machines on the same wire. The hubs don't care, because they are only looking at the outer envelope. The computers and routers are the only devices paying any attention to what is inside.

The twists and turns of networks

So how do you put the routers, hubs, computers, and the like together? There are a variety of ways you can configure your network, but something called Ethernet is my suggested choice for schools today, for a variety of reasons. At less than $50 per computer ($100 for Macintosh), Ethernet is much less expensive to install than token ring, which is a rival "layout" for a LAN. (One caveat: Prices are volatile--and that makes giving specific figures difficult.) The hubs used with Ethernet are also less expensive than the media access units (MAUs) token-ring layouts use.

Once you've chosen Ethernet, you have another decision to make as well: what kind of Ethernet to use. My experience is that Thin Ethernet is the least expensive option open to schools--but it's not an optimal technology. The reason: Thin Ethernet does not use hubs--it links computers in a daisy chain. Thus, a single failure--whether it's from a bad connector or a bored student--will take your whole network down.

I prefer 10 Base-T Ethernet, in which computers are arranged in a star configuration, with a dedicated wire connecting each computer to a hub. With this configuration, you don't have to worry about a single workstation failing and taking out the whole system. (Plenty of other things can kill any network.) Another advantage: The hubs used in 10 Base-T Ethernet are relatively inexpensive; a so-called unmanaged 12-port 10 Base-T hub costs less than $300.

Anyone interested in putting together his or her own network also needs to know something about the wiring involved. 10 Base-T uses a type of phone wire to connect the computers to the hub. This wire has four "pairs" of wires, and each pair is twisted together. (The twists are very important because they enable the wire to carry data without interference.) Each wire in turn is graded by how much data can pass through it with an acceptable amount of loss and interference. Category 3 wire is OK for most of the Ethernet networks used today, which typically run at a speed of 10 million bits per second (10 Mbps). But on the near horizon are several competing standards that will multiply that speed by 10, and these 100 Mbps network technologies operate on Category 5 wire.

Wiring a building is very labor intensive--and therefore very expensive. For that reason, I'd advise you to install Category 5 wire at the outset. That way, your wiring will be useful for a few years to come.

When it comes to wiring, you can make the most of your technology dollars in other ways as well:

The wire you install to connect computers on the network usually goes from a hub to a wall outlet where the computer will be located. 10 Base-T Ethernet uses an eight-wire phone connector (called RJ45 connectors) to link the computer to the wall jack. But you need to remember: These connectors also are rated by how fast data can pass through them. So if you are putting in Category 5 wire, make sure you also install Category 5 wall jacks.

You can hire a commercial vendor to connect the computers to the wall jacks. But connectors generally come all color coded--telling you where to put which color wire. If you have a handy telephone repairman's tool called a punch-down tool, you can make these connections yourself. For some reason, these tools are relatively expensive--between $30 and $60, to be exact. But they are simple to use and quickly pay for themselves. For example, a commercial company would have charged us $130,000 to install 250 networks and five hubs; instead, we had the commercial company pull the cable and terminate the classroom end, at a cost of $24,000, and I did the rest.

You can save money at the hub end of your wiring, too. Hire a commercial vendor to make the connection, and chances are he or she will install hundreds of dollars of "termination equipment" so you can make changes to the system without using a punch-down tool. What is customarily done is to "terminate" the wires on a "patch panel," which is a group of RJ45 connectors (usually in multiples of 12) that are directly connected to enough punch-down connectors for all 12 of the incoming wires.

Commercial installers will terminate the wires on a patch panel. They will then connect the hub to another patch panel, then use patch cables with RJ45 connectors at each end to connect the two patch panels to each other, thus connecting the computers to the hub.

If you're doing this yourself, you can do without the patch panels and instead use a punch-down tool to connect the computers to a punch-down block. That way, you can save a lot of expensive termination equipment. (Take a look at the phone system; that's the way the phone company does it every day.)

The hub of the matter

Hubs--the stand-alone devices that send any data they receive to any computer connected to them--are another key element in your network, and they come in many different sizes. The least-expensive hubs are called work-group hubs, and they come in eight, 12, 16, or 24-port models. (An eight-port hub has connections for eight computers, a 12-port hub has connections for 12 computers, and so on.)

Next come stackable hubs. These come in basically the same sizes as work-group hubs, but they can be stacked on top of each other and connected so they behave like one big hub. That way you can start with 12 computers connected to a single hub and add more hubs later, as you need them.

Finally, there are concentrators, which accept cards or some other type of module. Each time you add a module, you can connect more computers. Concentrators have another advantage as well: They can support different types of networks (10 Base-T modules in the same concentrator as Thin Ethernet or token ring or fiber). Concentrators have more advanced features, support more computers, and are more expensive.

A major difference among hubs to keep in mind is whether they are "manageable." A manageable hub is set up so that it can send information about itself to a computer on a network that is running network-management software.

Say, for example, that you get a call telling you the principal's computer has just gone dead. If you have a manageable hub, you can check the network connection even before you hang up the phone, even though the connection is in the janitor's closet down in the basement.

Actually, a manageable hub is handiest when you have a persistent problem that has no easy answer--something that usually occurs when you make major changes in a network. (You might, for example, have two or three connections that are not quite connected, so that every time someone tries to use the network from those computers, the transmit signal bleeds over to the receive wires, effectively jamming the network. At the same time, you might have an electrical wire that's too close to a bundle of wires coming from a computer lab, adding static to every one of those connections. And your new server might have a defective network card that starts jabbering incoherently every five minutes.)

All of these problems might be occurring simultaneously, masking each other. But a good manageable hub will allow you to nail all of these problems from your desk.

When you're just starting out with your network, you might not realize how important it is to have diagnostic tools built into your hub, because you might not have that many computers on the network to begin with--and people might not be that committed to using the network anyway. However, my experience is that networks rapidly become "mission critical." People come to depend on them, and when the network does not work, everyone wants your head. Your best bet: Buy hubs that can have network-management software added at a later time.

One final word on hubs: Hubs are repeaters. That is, whenever a computer transmits data, the hub repeats that data to the receiver in every other computer connected to it. Generally, hubs can repeat the data very quickly, too. Still, there is a limit to how many repeaters you can send data through before that information takes too long to get to its destination. That limit is four. When you start a network with one hub, this is not a concern. By the time you are networking a whole building, though, you usually have to take this limit on the number of repeaters into account.

This is where 10 Base-T Ethernet has problems. 10 Base-T Ethernet can only go over 300 feet of wire before the signal gets too weak. If you have a hub in the middle of a 600-foot hall, you probably can reach all of the rooms on that hall. But if you have a school with two floors--and 800-foot halls--you've got a problem, and you'll have to place your hubs around the building in such a way that no computer is more than 200 feet from a hub. (The rule of thumb is 300 feet of wire--but since the wire is twisted to keep out interference, that 300 feet of wire is closer to 200 feet of space.) With Category 5 wire, my experience is that 200 feet of hall is about as far as you want to push your luck--or pull the wire.

Among your options if you can't observe the 200-foot rule:

You can use an Ethernet wiring scheme that daisy-chains devices on a single wire; this is known as a backbone. Originally, Thick Ethernet (10-Base 5) was the norm for this arrangement, and could carry a signal for up to 800 feet. But this wire is expensive to purchase and install, and the electronics that must be connected to it are more expensive than the alternatives.

Thin Ethernet (10-Base 2) uses a coaxial cable that's slightly smaller in diameter than the cable used in cable television and can carry a signal up to 680 feet. Because this is the least-expensive type of Ethernet backbone, it is what I would suggest for a school.

Thin Ethernet uses BNC connectors on the end of each wire. These little guys are more of a problem than you can imagine. Unless you are very experienced at putting these on, I strongly recommend getting a professional to do it. Yes, it is possible to buy good connectors and the crimping and soldering tools and to get good solid connections that will last for years. However, it is also incredibly easy to put on a connector that will work fine for the first year, then start adding static and interference at odd times (usually times of high use) later on. The only way to solve such a problem is to go through the whole backbone, hack off the connectors, and replace each one. No fun. Have a pro do the BNC connectors.

One final note: Whenever you discuss wiring your school, someone will say, "What about fiber?" At this point, fiber-optic cable is more expensive than copper. Therefore, I think it is only practical where copper won't work. (Our whole school is networked using copper; however, when we wanted to extend the network across the parking lot to the district office via buried conduit, we had to use fiber-optic cable.) One advantage fiber has over copper is that it can carry a signal far greater distances. The connectors used with fiber also require special tools and supplies, and at this point are not the type of thing an amateur can do.

Another reason to use fiber is to get network speeds in the 100-Mbps range. Soon there will be 100-Mbps Ethernet over Category 5 copper, but if you need something that fast right now, you will probably end up with FDDI over fiber.

A word about routers

Remember your friend, Joe, in Alaska and how your router depended on his router to get your message through? Routers, as you will recall, are special computers that connect two or more networks. One example: You have some Macintosh computers that are connected using localtalk cables, the relatively slow .2-Mbps network that is built into every Macintosh. But say you've just purchased some of the new Macs that have Ethernet. How can you get these Macs to talk to each other? You simply set up your Ethernet network, and at some point, where you have both an Ethernet connection and a localtalk connection right next to each other, you set up a router. This router sits there telling each network to send it any packets that need to be sent to the other network.

When the router receives a packet from the localtalk network, it throws out the localtalk envelope and reads the AppleTalk address on the packet. The router then takes the information and stuffs it into an Ethernet packet, writes the Ethernet address of the recipient on the Ethernet packet, and sends the packet out over the Ethernet. All the sending computer had to know was the address of the router, which was saying it could figure out the final address.

This routing process works the same for other protocols. (It simply doesn't matter whether you're routing Novell's IPX between Ethernet and token ring or if you're routing the Internet's TCP/IP between Ethernet and a wide-area network protocol like SLIP or PPP.)

Unfortunately for the do-it-yourselfer, routers are not fun to set up. Especially nasty things happen to a network when a router starts advertising a route to a network that doesn't exist or when two routers claim they are the path to a certain network, and one is lying. Other problems occur when you have multiple protocols on your networks. Then a router has to say something like, "I am the gateway to TCP/IP network number 198.114.149, and IPX network 122EC4, and AppleTalk network 200-210." A router that is routing multiple protocols like that has to keep an awful lot of information in its routing tables, and unless it is very fast, it can really slow down the traffic that has to pass through it. For this reason, I believe it is best to purchase routers that are designed from the ground up to be routers and to stay away from using regular computers running routing software.

I'd also advise hiring a network professional to work with you the first few times you configure a router. Having access to someone who has done it before can greatly reduce the time it takes to learn this new skill--and the stress on the person responsible for keeping the network running while the router is being installed.

So try your hand at these approaches. You'll find Bob Vila doesn't have a corner on the do-it-yourself market.