Hubs, Switches, Repeaters, Bridges and Routers

The biggest difference between using coaxial cable and twisted-pair cable is that when you use twisted-pair cable, you also must use a separate device called a hub. Years ago, hubs were expensive devices — expensive enough that most do-it-yourself networkers who were building small networks opted for thinnet cable in order to avoid the expense and hassle of using hubs.

Nowadays, the cost of hubs has dropped so much that the advantages of twisted-pair cabling outweigh the hassle and cost of using hubs. With twistedpair cabling, you can more easily add new computers to the network, move computers, find and correct cable problems, and service the computers that you need to remove from the network temporarily.

A switch is simply a more sophisticated type of hub. Because the cost of switches has come down dramatically in the past few years, most new networks are built with switches rather than hubs. If you have an older network that uses hubs and seems to run slowly, you may be able to improve the network’s speed by replacing the older hubs with newer switches.

If you use twisted-pair cabling, you need to know some of the ins and outs of using hubs:

  • Because you must run a cable from each computer to the hub or switch, find a central location for the hub or switch to which you can easily route the cables.

  • The hub or switch requires electrical power, so make sure that an electrical outlet is handy.
  • When you purchase a hub or switch, purchase one with at least twice as many connections as you need. Don’t buy a four-port hub or switch if you want to network four computers because when (not if) you add the fifth computer, you have to buy another hub or switch.
  • You can connect hubs or switches to one another, as shown in Figure below; this is called daisy-chaining. When you daisy-chain hubs or switches, you connect a cable to a standard port on one of the hubs or switches and the daisy-chain port on the other hub or switch.

Figure 3: You can daisy-chain hubs or switches together.

Be sure to read the instructions that come with the hub or switch to make sure that you daisy-chain them properly.

  • You can daisy-chain no more than three hubs or switches together. If you have more computers than three hubs can accommodate, don’t panic. For a small additional cost, you can purchase hubs that have a BNC connection on the back. Then you can string the hubs together using thinnet cable.

The three-hub limit doesn’t apply when you use thinnet cable to connect the hubs. You can also get stackable hubs or switches that have high-speed direct connections that enable two or more hubs or switches to be counted as a single hub or switch.

  • When you shop for network hubs, you may notice that the expensive ones have network-management features that support something called SNMP. These hubs are called managed hubs. Unless your network is very large and you know what SNMP is, don’t bother with the more expensive managed hubs. You’d be paying for a feature that you may never use.
  • For large networks, you may want to consider using a managed switch. A managed switch allows you to monitor and control various aspects of the switch’s operation from a remote computer.

The switch can alert you when something goes wrong with the network, and it can keep performance statistics so that you can determine which parts of the network are heavily used and which are not. A managed switch costs two or three times as much as an unmanaged switch, but for larger networks, the benefits of managed switches are well worth the additional cost.

Network Repeaters

A repeater is a gizmo that gives your network signals a boost so that the signals can travel farther. It’s kind of like a Gatorade station in a marathon. As the signals travel past the repeater, they pick up a cup of Gatorade, take a sip, splash the rest of it on their heads, toss the cup, and hop in a cab when they’re sure that no one is looking.

You need a repeater when the total length of a single span of network cable is larger than the maximum allowed for your cable type:

Cable Maximum Length
10Base2 (Coaxial) 185 meters or 606 feet
10/100BaseT (Twisted Pair) 100 meters or 328 feet

For coaxial cable, the preceding cable lengths apply to cable segments — not individual lengths of cable. A segment is the entire run of cable from one terminator to another and may include more than one computer.

In other words, if you have ten computers and you connect them all with 25-foot lengths of thin coaxial cable, the total length of the segment is 225 feet. (Made you look! Only nine cables are required to connect ten computers — that’s why it’s not 250 feet.)

For 10BaseT or 100BaseT cable, the 100-meter length limit applies to the cable that connects a computer to the hub or the cable that connects hubs to each other when hubs are daisy-chained with twisted-pair cable.

In other words, you can connect each computer to the hub with no more than 100 meters of cable, and you can connect hubs to each other with no more than 100 meters of cable. Figure below shows how you can use a repeater to connect two groups of computers that are too far apart to be strung on a single segment.

Figure 4: Using a repeater.

When you use a repeater like this, the repeater divides the cable into two segments. The cable length limit still applies to the cable on each side of the repeater. Here are some points to ponder when you lie awake tonight wondering about repeaters:

  • Repeaters are used only with Ethernet networks wired with coaxial cable; 10/100BaseT networks don’t use repeaters. Actually, that’s not quite true: 10/100BaseT does use repeaters. It’s just that the repeater isn’t a separate device. In a 10/100BaseT network, the hub is actually a multiport repeater. That’s why the cable used to attach each computer to the hub is considered a separate segment.
  • Some 10/100BaseT hubs have a BNC connector on the back. This BNC connector is a thinnet repeater that enables you to attach a full 185-meter thinnet segment. The segment can attach other computers, 10BaseT hubs, or a combination of both.
  • A basic rule of Ethernet life is that a signal can’t pass through more than three repeaters on its way from one node to another. That doesn’t mean you can’t have more than three repeaters or hubs, but if you do, you have to carefully plan the network cabling so that the three-repeater rule isn’t violated.
  • Repeaters are legitimate components of a by-the-book Ethernet network. They don’t extend the maximum length of a single segment; they just enable you to tie two segments together. Beware of the little black boxes that claim to extend the segment limit beyond the standard 185-meter limit for thinnet or the 100-meter limit for 10/100BaseT cable. These products usually work, but playing by the rules is better.

Network Bridges

A bridge is a device that connects two networks so that they act as if they are one network. Bridges are used to partition one large network into two smaller networks for performance reasons. You can think of a bridge as a kind of smart repeater.

Repeaters listen to signals coming down one network cable, amplify them, and send them down the other cable. They do this blindly, paying no attention to the content of the messages that they repeat. In contrast, a bridge is a little smarter about the messages that come down the pike.

For starters, most bridges have the capability to listen to the network and automatically figure out the address of each computer on both sides of the bridge. Then the bridge can inspect each message that comes from one side of the bridge and broadcast it on the other side of the bridge, but only if the message is intended for a computer that’s on the other side.

This key feature enables bridges to partition a large network into two smaller, more efficient networks. Bridges work best in networks that are highly segregated.

For example, suppose that the Sneetches networked all their computers and discovered that, although the Star-Bellied Sneetches’ computers talked to each other frequently and the Plain-Bellied Sneetches’ computers also talked to each other frequently, rarely did a Star-Bellied Sneetch computer talk to a Plain- Bellied Sneetch computer.

A bridge can partition the Sneetchnet into two networks: the Star-Bellied network and the Plain-Bellied network. The bridge automatically learns which computers are on the Star-Bellied network and which are on the Plain-Bellied network.

The bridge forwards messages from the Star-Bellied side to the Plain-Bellied side (and vice versa) only when necessary. The overall performance of both networks improves, although the performance of any network operation that has to travel over the bridge slows down a bit.

Here are a few additional things to consider about bridges:

  • Some bridges also have the capability to translate the messages from one format to another. For example, if the Star-Bellied Sneetches build their network with Ethernet and the Plain-Bellied Sneetches use Token Ring, a bridge can tie the two together.
  • You can get a basic bridge to partition two Ethernet networks for about $500 from mail-order suppliers. More sophisticated bridges can cost as much as $5,000 or more.
  • If you’ve never read Dr. Seuss’s classic story of the Sneetches, you should.

Network Routers

A router is like a bridge, but with a key difference. Bridges are Data Link layer devices, so they can tell the MAC address of the network node to which each message is sent, and can forward the message to the appropriate segment.

However, they can’t peek into the message itself to see what type of information is being sent. In contrast, a router is a Network layer device, so it can work with the network packets at a higher level. In particular, a router can examine the IP address of the packets that pass through it.

And because IP addresses have both a network and a host address, a router can determine what network a message is coming from and going to. Bridges are ignorant of this information. One key difference between a bridge and a router is that a bridge is essentially transparent to the network.

In contrast, a router is itself a node on the network, with its own MAC and IP addresses. This means that messages can be directed to a router, which can then examine the contents of the message to determine how it should handle the message.

You can configure a network with several routers that can work cooperatively together. For example, some routers are able to monitor the network to determine the most efficient path for sending a message to its ultimate destination. If a part of the network is extremely busy, a router can automatically route messages along a less-busy route.

In this respect, the router is kind of like a traffic reporter up in a helicopter. The router knows that the 101 is bumper-to-bumper all the way through Sunnyvale, so it sends the message on 280 instead.

Here is some additional information about routers:

  • Routers aren’t cheap. For big networks, though, they’re worth it.
  • The functional distinctions between bridges and routers — and switches and hubs, for that matter — get blurrier all the time. As bridges, hubs, and switches become more sophisticated, they’re able to take on some of the chores that used to require a router, thus putting many routers out of work.
  • Some routers are nothing more than computers with several network interface cards and special software to perform the router functions.
  • Routers can also connect networks that are geographically distant from each other via a phone line (using modems) or ISDN.
  • You can also use a router to join your LAN to the Internet. Figure below shows a router used for this purpose.

Figure 5: Using a router.