Energy Transformation Systems, Inc. (ETS) is the pioneer... Cart Contact
1 2 3 4 5

Future Networks for Business

People have been installing Local Area Networks (LANs) for over 20 years. The first LAN of note was ARCnet. The second to achieve a large installed base was AppleTalk. The watershed event for Apple was putting network connectivity directly into their printers.

I have not included IBM systems on this list of pioneers, because those were not true LANs. The IBM systems could not communicate peer to peer but relied on the central computer to control things. IBM already had million of dumb terminals when everyone else was just starting with LANs.

Then a funny thing happened. Network operating systems that relied on servers (the client/server paradigm), took over and became first choice. Servers are like central computers in that you go to them to pick up your mail. What was driving this? Cost. PC's with LAN's were a lot cheaper than main frames with terminals, especially for e-mail.

Ethernet was first conceived as a bus system. It announced in 1979 by Xerox, Intel and DEC. All the computers were attached through plenum transceivers to the same continuous cable. It was conceived for spanning a building, so the cable was bulky and inflexible, and signals could traverse 1500 feet. The system (10Base5) was dubbed ThickNet.

Because this system was expensive to install it was followed a couple of years later by ThinNet (aka 10Base2, colloquially known as "CheaperNet"). The big differences were that ThinNet used a readily available, flexible coaxial cable and could only accept 32 terminals on a segment. This quickly became the workgroup LAN of choice.

In the meantime, some people left Xerox PARC, intending to make fiber optic networks and founded Synoptics. But they quickly decided to put Ethernet over telephone wire as a quick way to build a base of revenue. This design was such a hit that the IEEE started a project for standardizing Ethernet over telephone wire. This standard (10BaseT) was based upon using two pairs (transmit and receive) from each user to interconnected hubs to send the signal from any station to all the other station. This system was an instant success with everyone except IBM.

In the early 80's IBM rejected Ethernet for technical reasons: it was not controlled and hence could not guarantee timely delivery of data packets. So IBM created and supported Token Ring as a competing LAN standard. The first implementation was 4Mbs, which grew to 16Mbs. IBM's support for Token Ring in conjunction with trashing the basic tenets of the Ethernet design resulted in several years of good growth for Token Ring. Many large companies listened to IBM. In fact, in 1989 many industry "experts" publicly predicted that by 1994, Token Ring would be the dominant LAN. Well, it didn't happen. By 1994, Token Ring was on its way to oblivion. The primary problems were cost (much more than 10BaseT) and the availability of 100BaseTX, dubbed Fast Ethernet.

Advances didn't stop there. Gigabit Ethernet was confirmed as a standard two years ago for fiber and last year for copper. All of these Ethernet speeds, 10,100 and 1000Mbs run over telephone wire. Of course, we no longer call it telephone wire, it has been redesigned for frequencies up to 100 MHz and using sophisticated techniques has been made useable up to 1000-megabits. Most people know this wire as Category 5 (Cat5) wire. Please note that I have used two terms, MHz and Mbs. MHz (megahertz) refers to the bandwidth of the wire while Mbs (mega bits/second) refers to the rate at which data bits flow over the wire. Special encoding allows more bits per Hertz. For example, the basic telephone bandwidth is less than 4KHz (accommodates a phone requirement of 300 to 3000 Hertz), but modems send data to the central office at 33KHz over that same channel. Note that data can be sent from the central office facilities at higher speeds by bypassing a filter that limits the bandwidth. All common phone lines have protective filters.

The IEEE Ethernet standards committee is currently working on 10-Gigabit Ethernet. It will only be on fiber. Transmission losses over copper wire at the higher data rates prohibit the use of conventional Cat5 wiring, making a 10Gbps over copper standard too costly.

There are a lot of forces creating the 10-gigabit Ethernet sweet spot. Today, long distance transmission uses Sonet networks. Last year 10-gigabit transmission over Sonet became very economical. Today, more and more fiber transmission is moving up to 10 gigabit from 2.5 gigabit. With high volumes comes lower costs, and 10-gigabit will quickly become the preferred medium.

A problem with transporting Ethernet over the Sonet network is the need to encapsulate the Ethernet packets. With the development of 10 gigabit Ethernet, transmission will require fewer overheads, again reinforcing it as the choice over Sonet. In the near future, 10 gigabit Ethernet will be the system of choice for wide area Internet connectivity.

You might well ask what this has to do enterprise networking. My point in this abbreviated history is the continued, rapid development of capabilities. I would have been accused of inhaling, if in 1980, I had predicted 1 gigabit, let alone 10 gigabit LANS today. You are reading this because you want to know what your best options are, so as not to make expensive mistakes.

The structure of government is much different in a small town than in a big city. That is also true of networks. A major reason for installing increased bandwidth is driven by increased message traffic at any given point in the network. A network with 15 active users will not degrade significantly with 10BaseT. In fact with normal users (as opposed to power users), you can have up to 100 nodes with only minor collisions. What many people do is pay a small premium and buy 10/100 Network Interface Cards (NICs). These cards automatically switch between 10 and 100Mbs based upon what the rest of the network can handle. If the 10BaseT hub is replaced (as dictated by changing network requirements) with a Fast Ethernet (100BaseTX) hub, the NIC cards do not need replacing. Also, if everyone except a couple of power users are satisfied with the 10BaseT system, the power users can be migrated to 100BaseTX hubs, along with the corporate servers.

Suppose there are a couple of hundred users making the network too big. You can change from hubs to switches. Hubs share their total bandwidth among all concurrent users. Switches allows several messages to proceed to their destinations simultaneously at full bit rate. This cuts the bandwidth bottleneck way down.

If you are not a large company you will probably not need gigabit LANS in the immediate future. That means that the basic technology you need for your future LAN already exists and is getting cheaper by the day. What you do today is use Fast Ethernet hubs with selective placement of switches in the network to eliminate bottlenecks.

You might say, "Wait a minute, what about fiber?" Show me the application. Fiber is not heavily used in LANS because it is too expensive. When you come up with an application that makes sense for fiber to the desktop we can revisit it. The most common internal use for fiber is in server-to-server connectivity.

Again you say, "Wait a minute, what about video?" In the context of business (not homes or stores) there is not much use for TV, in spite of many years of opportunity. Most video will be over IP (Internet Protocol). By the time it is digitized and compressed (with compression techniques getting better every day), good quality video now requires less bandwidth than 10BaseT provides.

I cannot foresee the need for anything more than Fast Ethernet switches for most small business. One caveat: 10 years ago I (or you) could not have predicted the popularity of the Internet. That is because the browser and hyper-linking did not exist. Absent a major invention, I stand by my analysis.

© Energy Transformation Systems, Inc.

Web City Pages