About 35 years ago Bob Metcalfe, an engineer working for Xerox Corp., came up with a transmission scheme that has survived and thrived to this day. It was called Ethernet because of its similarities with radio transmission schemes that broadcast over the "ether."

The original intent of the system was to transmit computer data over a shared coaxial cable.

According to the plan, a number of users would ride the same cable, and transmission speeds (originally) would be 3 Mbit/s. The system would serve a local area network (LAN) -- a networkd covering a small physical area like a home, office, of small group of buildings. Schools and airports come to mind as good examples.

The "shared" requirement, of course, presented a significant challenge. What if two users started sending a message at the same time? Clearly there would be a collision. But then what? Do the terminals know there has been a collision, and if so, what do they do about it?

The Ethernet situation isn't much different from what happens, say, at a dinner party when two people start talking at the same time. They hear each other, and instantly fall silent (don't we wish some of the talking heads on TV would occasionally do the same). There is a brief pause, and then one or the other re-starts the conversation.

Because the two are looking at each other they are usually able to avoid a second collision. But what if they can't see each other? (This happens on TV when the two speakers are thousands of miles apart, and the signal being transmitted is via satellite. Often there is an uncomfortable time lag in the conversation as the two wonder if it is their turn to speak). The result is a second collision.

So what happens with Ethernet? A station with a message to send will wait until the line is idle, and then start transmitting. Other stations, hearing that transmission, will refrain from sending any message. However if another station started transmitting at almost the same time, there is a collision. The two messages are on top of each other. Both of these transmitting stations detect this collision (they are not only transmitting, they are listening). Both will immediately stop sending their message, and then send what is called a "jam" signal. This signal simply extends the silent period to the minimum packet time so that all stations know that there has been a collision. Incidentally, this entire process is called Carrier Sense Multiple Access with Collision Detection, or CSMA/CD.

But now what? To get back on line each station goes to a little formula (it is called the truncated binary exponential backoff) that tells it how long it should wait before re-transmitting its block. If by chance there is a second collision the formula (with different input numbers) will be re-applied, and the station will try again. If, after a predetermined time, the station is still unable to successfully transmit its message, the process is aborted. That's where the "truncated" comes in: the process doesn't go on forever.

Now the important part: what's in the message?

First of all, recognize that we're dealing with packets: little bundles of data. In this case they're called Frames. We're not sending one long stream of data making up an entire message.

Each frame is made up of a series of octets: 8 bits of data.

Up front, on every frame, is a preamble consisting of 7 octets (each 10101010), followed by a start of frame signal (one octet of 10101011). These frames are stripped off as the frame moves down the line.

Then comes the address. This is built into 6 octets, and is followed by the return address, consisting of another 6 octets.

Next the system must be told how long the message is going to be. This is done with 2 octets.

And finally comes the payload: anywhere between 46 and 1,500 octets.

At the end of the frame is a housekeeping item, consisting of 4 octets.

Finally there is an interframe gap of 12 octets.

The transmitted information, therefore, is anywhere between 64 and 1,518 octets.

And what's the speed of this entire operation? As we mentioned, the first Ethernet system was designed for 3 Mbit/s. But we've moved beyond that, and on paper there are at least a dozen versions. Many, of course, are little used or completely obsolete.

The original standard called for a single coaxial cable into which you literally tap a connection by drilling into the cable to connect to the core and screen. (Remember, it was designed to be a coax system.)

Then came 10 Mbit/s Ethernet. It also used coaxial cable.

Next came the system called 10BASE-T, which operates over 4 wires (2 twisted pair) on a Category 3 or Category 5 cable. It is 10 Mbit/s.

A system called Fast Ethernet came next. The most popular system is the 100BASE-T system, operating at 100 Mbit/s. It utilizes Category 5 cable (using 2 out of 4 pairs)

Next is Gigabit Ethernet, operating at one gigabit per second. There are 3 working versions: 1 using Category 5 unshielded twisted pair copper, 1 using multi-mode fiber, and 1 using single-mode fiber.

Finally there are 5 systems making up 10-gigabit Ethernet. We will see more of these in the future.

So the system designed more than 3 decades ago continues to be the workhorse for local area networks. 

What’s your take on this subject? Leave a comment and get the conversation going.