December 1999, Issue 113

Thin is In:
Clients, Servers and Systems

Start • Old Friends New Technology • The Chicken or the Egg • The Software Skinny • Administering Thinsystem • Sources & PDF

Back in days of leased hardware and software that filled rooms the size of football fields, a relatively obscure company called IBM was successfully implementing what we would know later as thin client architecture.

The user terminal of choice was called a 3270 workstation and consisted of a CRT and keyboard. The CRT/keyboard combination was the perfect picture of a dumb terminal.

All of the smarts were contained in a coax-attached external terminal controller called a 3274. The 3274 was a cluster controller capable of handling up to thirty-two 3270 devices.

Impact printers with 3xxx designations could also be attached to the 3274’s BNC ports. Parallel channels using massive cables passed data between the central processor, the 3274, and its 3270 parasites.

At that time, the smallest of these processors was the size of a large doublewide refrigerator. Typically, though, the processor footprints were large enough to take a little bit of time to walk around.

Practically all of the application smarts were contained at the processor level. All of the programs were run on the remote processor that resided in the glass house. The 3274 was only there to convert the raw data and application interactions to and from the 3270 terminal protocol.

This processing architecture still exists today, with the 3274 and 3270 controller/terminal combination giving way to smarter devices at the terminal end. In fact, the 3270 green screen terminals gave way to the 3279 high-resolution color units later on. This was (and is) a large implementation of massive processing power capable of simultaneously serving thousands of users anywhere in the world.

To sum up the technology, relatively small packets of data from terminals were bounced between a processor (server) and terminal (client) with most of the heavy-duty application processing occurring at the server side. It has been debated for years as to the real value of having such a behemoth system when today’s miniature desktops can do the same job. I really don’t want to get in that fight.

It’s funny that later on, the powers of the computing world decided that this was not the most efficient way to compute or do business. So, they decided to "distribute" the processing by offloading the processor (server) applications to the clients.

Thus, the clients (terminals) got smarter and the processors (servers) got smaller. IBM’s first smart terminal answer was the 3270PC, which was no more than an IBM AT with some special BIOS hooks and a 3270 ISA adapter card.

At the time, the IBM AT ran at a mind-bending 6 MHz on top of a 30-MB hard drive! They even extended this PC/AT technology to run the large OSs like VM (Virtual Machine).

That was then, and this is now. Thin is in…again.