How to tell a mainframe from a minicomputer from a micro

I worked at a company in the early 1990s that ran its business on an IBM System/36. It was basically a System/360 but one tenth of the size and expense! This was the only minicomputer I ever encountered, and I was not responsible for it. I was there to work on the "micros" (Windows PCs). I remember the System/36 programmer making comments about PC techs not being "real" computer operators or programmers. She was nearing retirement at that time, and if she is still alive I imagine she is enjoying a comfortable pension.

> a mainframe isn't really a computer, it's a whole cluster of closely-coupled computers, many with dedicated functionality.

This is a great description. I think we're moving back to the mainframe architecture, but the "cluster" now includes all the machines in a data center, plus the mobile and PC clients connecting to services.

Tangentially: I talk about how programming models have evolved from the mainframe to today here: https://gridwhale.medium.com/gridwhale-and-a-brief-history-o...

> So as someone else commented, a mainframe isn't really a computer, it's a whole cluster of closely-coupled computers, many with dedicated functionality. Quite possibly all implemented with different architectures, instruction sets, programming languages, bit widths, everything.

Nearly all computers more powerful than a simple microcontroller look like this in 2022.

If you find this fact surprising, you'll probably want to take a look at this talk from USENIX ATC '21: https://www.youtube.com/watch?v=36myc8wQhLo

If there is someone phoning you or knocking on your door when you turn it off, it's not a microcomputer.

I'd make a point that the entire line of experimental computers as developed at MIT and later commercialized by DEC, with an emphasis on realtime and graphics, doesn't really adhere to this pattern. E.g., we may call the PDP-1 the first commercial workstation. (Single user, realtime, graphics, used for exploratory work, emphasis on I/O capabilities with a large array of input and output devices. But the PDP-1 saw also use as a small mainframe and was a testbed for time sharing.) In a sense, minis began conceptually as downgraded workstations. (Which became viable only as memory cycles shrunk, thus allowing for smaller word sizes.)

Originally a HN reply posted here: https://news.ycombinator.com/item?id=32969692

> Separate processors in storage, in terminals, in networking controllers, in printers, in everything. Typically the machine cannot actually drive any output or input directly (e.g. mouse movements, or keystrokes, or anything): peripherals do that, collect and encode the results, and send them over a network.

Sounds exactly like USB.

Oh -- thanks!

It is just a repurposed HN reply, which is why I did not post it myself.

Yeah, sort of, but it's more like this:

First, there were computers. There were some that were made to be small enough for a group (IBM 1620, for example), and some that were large and intended as the central device for a large organization (IBM 7090). IBM referred to both as a "data processing system". These were delivered as batch-oriented systems.

Large computers like the 7090 often had smaller, accessory, general-purpose computers that handled things like card reading/punch and printing. For example, the PDP-10 had a PDP-11 as its front-end controller.

In the late 60s the term minicomputer appeared. A little while later, it became popular to refer to the large computers as "mainframe" computers. This probably was marketing speak to imply that, while minicomputers can be good for somethings, you wouldn't mini-frame your house, so you had better main-frame your company. I like to think about this time, which was also the heyday of database innovation, as the time when companies thought there would be many integrated applications running on the same data base [two words at the time] operated by a large, reliable computer and accessed by peripheral terminals. Through the 80s, people in companies often referred to "the computer" rather than considering multiple computers as the norm.

Microcomputers appeared with the development of the microprocessor. Certainly, the Altair 8800 started the wave of using a fully integrated microprocessor as the basis for a general-purpose computer. They grew out of calculators and, tellingly, from the project to develop an intelligent display terminal.

Micros often looked like minis in their use - they might be operated by terminals and punched tape, saving onto hard disks, floppy disks and magnetic tape. In my opinion, the distinction was if you used a discrete ALU, like the VAXes or the Xerox Alto, you were still using a minicomputer. Many minicomputers used ECL* to achieve high speeds that were not reachable by NMOS or CMOS logic at the time. Eventually, minicomputers also became microcomputers, for example, the MicroVAX. I think that these articles get it right that the microcomputers are different because they are not considered to be the orchestrator of a system of disks, tapes, terminals, printers, etc. They are at best part of a network with specialized microcomputers helping share resources when needed.

There is trend of data ending up in silos that happens with minicomputers and microcomputers. Applications get written for their own data, they aren't given direct access to other applications' data, and the data itself is geographically distributed.

As time has gone on, we now have the datacenter and virtualization. Data has been moving out of silos and into large, shared databanks. The computing unit is no longer made out of hardware, but rather represented by resources that can be provided in many virtualized ways.

* Emitter-coupled logic, a high-speed discrete logic often used in supercomputers and fast general-purpose computers until CMOS caught up with better power and heat characteristics.

mainframe: building sized

minicomputer: room sized

microcomputer: desk sized

So now microcomputers are mainframes, but with all that extra hardware virtualized and emulated.