Cable and wiring for LANs

Over the last three years, network speeds have increased by a factor of ten, with another tenfold increase in 18 months or so. There has been a definite
Author:
Publish date:

Over the last three years, network speeds have increased by a factor of ten, with another tenfold increase in 18 months or so.

There has been a definite increase in the amount of networking technology installed in television facilities around the world. The reasons for this are clear. Over the last three years, network speeds have increased by a factor of ten. Technologies on the horizon promise another tenfold increase in 18 months or so. At the same time, consumer demand for these technologies has driven the prices down dramatically. 10Base-T hubs are now in the $20 range, 100Base-T hubs can be had for $125, and Gigabit Ethernet hubs are now available for around $1500. This low-cost, high-speed network technology has moved networking into the fast lane.

Basic topologies

Ethernet is by far the most common technology in use today. Fibre Channel and ATM are also finding their way into the television plant. While Ethernet and Fibre Channel are used widely within a facility, with a few exceptions, ATM is generally employed to connect one facility to another.

There are four basic topologies employed in computer networks today. These are point-to-point, star, Thinnet, and switched fabric.

Point-to-point

Figure 1 illustrates point-to-point topology where each node is directly connected to the next without any intervening hardware. The most common systems using this topology are fiber channel arbitrated loop (FC/AL) and token-ring.

Point-to-point topology is low-cost, but it has a big drawback. It is subject to many single points of failure. Since each computer is connected to the loop, a failure of any cable, any computer, or any network interface card can cause all computers to be unable to communicate. A common way to deal with this problem is to employ two loops. Another way that designers deal with this problem is to move to the star topology depicted below.

Star

Figure 2 illustrates star topology where each node is connected to a hub or concentrator. This is the basic building block of most Ethernet deployments. It is also used in Fibre Channel networks when designers want to protect the user from the problems described above.

A disadvantage of star topology is its somewhat higher cost. A big advantage of this topology is that all most all hubs can isolate a faulty node so that this connection does not affect other computers attached to the network. Of course, the hub is a single point of failure, and steps should be taken to minimize risk in this area.

Thinnet

Figure 3 illustrates Thinnet topology in which all computers connect to a common cable. Thinnet is based on RG-58 Coaxial cable, “T” connectors and 50-ohm terminators. This topology was very popular a number of years ago, but is now largely obsolete.

Thinnet is inexpensive, but if connection fails, it will affect all computers on the network. Also, termination of cables is time-consuming and prone to error. Furthermore, Thinnet typically costs more per connection than other technologies available.

Switched fabric

The fourth topology is switched fabric. (See Figure 4.) Switched fabric topologies are commonly used in switched networks such as Ethernet or Fibre Channel and ATM.

Switched networks can provide the highest aggregate bandwidth of any topology currently available short of dedicated point-to-point networks. Switched fabric networks will be the basis of most high-speed networks of the future.

The party line

With the exception of switched fabric topology, as the number of nodes on the network increases, performance decreases. This is because these networks connect everyone together on one big party line. As with an old-style telephone party line, if more than one person talks, it can be very confusing to determine what is being said.

In the past, network designers resolved this problem by keeping individual network segments small. They connected a number of segments together with bridges to build larger networks. (Bridges isolate the chatter on one network segment from another and only send messages across segments when required.) This solves the party line problem to some extent, but it does not fix it entirely. If you are unfortunate enough to share your segment with a fileserver, graphics workstation, or other high-bandwidth user, you will still notice a significant decrease in performance whenever these high-bandwidth devices are active.

The advantage of switched fabric

So what is the difference between a switched fabric and other topologies?

The biggest difference is that a switch can give you a dedicated full-bandwidth segment all to yourself. This means that if you are running a 10-, 100-, or 1000MB network, you will have access to all that bandwidth (less overhead, of course).

Another big difference is that, once you go to a dedicated segment, you are able to use full duplex communications — that is, you can send and receive data at the same time. Remember that in other systems, one person talks while everyone else listens.

Furthermore, in particularly heavy usage situations, you can install more than one network connection to a device to provide simultaneous dedicated full bandwidth connections.

Figure 5 illustrates how you might construct a switched network to provide very high bandwidth connectivity to a server and a graphics client while providing conventional connectivity to typical desktop users.

Note that the server is given two 100Mb full duplex connections to the switch. Because it is the only computer on these nodes, it has the full 100 megabits available on each link. It is possible to add more than two cards to a server, increasing the available network bandwidth even further. (There is no point in increasing the network bandwidth if the server has I/O or other limitations.)

Second, note that the graphics workstation has a dedicated 100Mb full duplex link to allow it to exchange very large files with the server.

Third, note that the Ethernet hub provides desktop clients access to the same high performance server, but over much slower 10Mb shared links.

Finally, note that the graphics workstation can consume all its available network bandwidth to the server without affecting any of the desktop connections.

Wiring

Most networks today are based upon unshielded twisted (UTP). Fiber is also gaining in popularity, but that is a subject for another article. UTP is popular because of its low cost and ease of installation. However, there are some things to be aware of. If you are planning a new network installation, be sure that you build it with CAT 5E components. (CAT 5 was the rating for network installations up to 100Mb/s. Cat 5 has been Extended for use up to 1000Mb/sec or one Gigabit. The “E” in CAT 5E stands for Extended.) CAT 5E assures that your wiring will be capable of supporting Gigabit+Ethernet speeds. If you use CAT 3 or conventional telephone cable, you are asking for problems. I once had a network installation that was built with CAT 5 components, all except for a jumper panel that used some flat ribbon jumpers. The jumpers were not CAT 5 rated. As a result, the network was unpredictable, and in some cases, would not function at all. Replacing the non-rated jumpers with CAT 5 components fixed the problem. You might save some money by purchasing less expensive wire, terminating blocks or patch panels, but in doing this, you may be exposing yourself to problems down the road. When you look at the total cost of a network installation, wiring and components are frequently a small part of the total cost of the system.

As network speeds increase, new technologies building on this speed emerge. Storage Area Networking or SAN is one example. SAN is shared storage where the storage in a computer or server is physically separate from the computer itself. Multiple computers can connect to the same storage across a high-speed network. One can easily imagine a high-speed network with SAN to provide access to a pool of shared storage. Nonlinear editors, graphics workstations and other systems could share the storage. Such designs were not possible just a few years ago.

New networking equipment and techniques are being developed all the time. One thing is for sure — broadcasters will be at the forefront of this technology employing high speed networking in video applications.

Brad Gilmer is executive director of the AAF Association and president of Gilmer & Associates, a broadcast consulting firm.

Send questions and comments to: brad_gilmer@intertec.com