Switches: Not the universal solution
Russell Driver's and Zoltan Nadj's article, "Networking hardware technologies and techniques," published in the July 2000 issue (see page 13), has a serious flaw
Russell Driver's and Zoltan Nadj's article, "Networking hardware technologies and techniques," published in the July 2000 issue (see page 13), has a serious flaw. The section on switches would lead any novice to believe that switches are the answer to all bandwidth problems. This is false. Switches work well when segregating traffic moving within different segments of a network, or when the intent is to create workgroup areas on the network.
Switches do not help when:
- All traffic is going to a single entity, such as a server or Internet connection. For example, take a network with 50 computers and a 56-kbit frame-relay Internet connection. A switch will not change the fact that the 56-kbit connection is substantially undersized when it comes to supplying the Internet demands of those 50 computers. You can install switches until hell freezes over, and the 56-kbit wide-area-network link will still be the bottleneck.
- There is very little traffic on the network. In my travels through network land, I have found that the network technician/consultant with traffic monitoring equipment is very rare. Yet most technicians and their consultants will not hesitate to recommend installing switches on any network. There are many network managers who do not have a network traffic-monitoring package as part of their arsenal of tools. I have seen too many networks with Ethernet utilizations below 5%, yet switches were installed to "improve" performance.
- The network design is poorly thought out and the switch is badly located. The location of the switch in the network can make a huge difference on its overall value to the network.
Don't forget the fiber
Frank Murawski's article on the flattening of the cabling market (see "Cabling market growth slowing dramatically," July 2000, page 43) was quite interesting. He focuses on unshielded twisted-pair (UTP), and I think he may have overlooked a very important factor: fiber.
While the copper-cabling vendors are confusing everybody trying to sell three different types of cables (Category 5E, Category 6, and Category 7) all at once, with two of the three being proprietary systems (or, as they say, "pre-standard"), fiber has quietly been sneaking in and taking more of their market.
Many more users are discovering that a centralized fiber network eliminates the wiring closet, reducing costs by $5,000 to $75,000 per closet. That makes fiber the cheaper alternative in many cases.
If our sales of fiber-optic testers and toolboxes are any indication, building fiber sales are approaching triple-digit growth rates!
Jim Hayes, President