If a PC is like an amoeba then a mainframe is like a vertebrate.
At this point, I'm sure any biologist could be justifiably displeased with such an analogy, given the complexity of the amoeba, and the fact that we continue to discover new things about this basic life form which has so many aspects to it.
However, if you will indulge me, I think this is a good way to introduce the hardware of the mainframe and draw a distinction between it and the computing hardware that most of us are used to.
My first introduction to what might be called a PC was the Apple ][+ computer: a consumer-focused, self-contained computer designed to have all of the most basic essentials together. Conceptually, this is very much like a single-celled organism, with all of its functionality bound up in a single place. And, while things such as the screen and other peripherals were external to the computer case, everything was about a single user at a time and served by a single processor and operating system instance.
Next, IBM introduced their PC, which became the genus of the wintel species that has come to predominate the consumer and small business computing world. Like the Apple, the IBM PC was a stripped-down, affordable, consumer-oriented machine that had everything in, or attached to, a single box.
While processor speed, memory and storage capacity, and complexity of operating systems, applications and interfaces have grown over the years, the PC has continued to be the conceptual equivalent of a single-celled organism - everything self-contained in a single place serving a single user or purpose at a time, all directed by the CPU.
To which I hope you are responding, "But how about networking - home, corporate, Internet, etc. - and multitasking and virtualization?"
The first half of this question, for me, can be answered by saying that the network can be seen as just another interface, like the keyboard, screen, mouse, joystick, and disk drive: it's an outside source of input and destination of output for the PC. In our analogy, it's part of what might be called the environment.
Multitasking and virtualization are, of course, primarily software issues, so I don't want to stretch the analogy too far. Briefly, though, re multitasking, let's not forget that there are many processes happening concurrently inside an amoeba as well - movement, digestion, reproduction, and so on - but they're all focused on the behavior of that single cell. I should also mention that PC's are at their most functional when they have no more than one high-demand task running at a time.
Virtualization also doesn't change the essential nature of the PC - it just allows for one or more instances of the PC environment to run generically on one or more PC hardware configurations.
This is where the difference between PCs and mainframes begins to resolve: no matter how many PCs I have, each one is, in its essence, a generic PC. I can have a million PCs connected together, but that doesn't make them into even a single mainframe any more than a million amoeba together spontaneously become a vertebrate.
The difference is a highly-ordered, functional structure that allows different parts of the mainframe to have specialized tasks that work together, much like a vertebrate with a brain, internal systems and organs, a skeleton, and limbs. As with the PC, software is an important part of how these operate. Unlike the PC, however, the mainframe hardware was designed from the very beginning according to this structure, which enabled even the earliest mainframe processors to be effective, despite having very small capacities.
Unlike a PC, then, the mainframe's brain or central processors are able to focus their power and capacity (which continue to be leading-edge) primarily on doing the work that pays for them.
Like bodily systems and internal organs, the mainframe has controllers to deal with the vast amount of data that passes through the mainframe.These "sub-computers" free up the central processors from focusing on the other activities of the mainframe, and they are not simply mini-mainframes - they're functionally designed to perform support roles other than application processing.
The limbs, then, would be analogous to the actual devices attached at arm's length to the mainframe via the controllers, including: vast amounts of disk and tape storage, high-speed printers, network-connected users and other computers.
There are other hardware-based aspects that are more analogous to the autoimmune system, enabling IBM's statement of integrity to the effect that unauthorized application programs, subsystems, and users are prevented from bypassing mainframe security. This is also a significant differentiator because it means that mainframes are secure all the way down to the bare metal, as distinct from PCs which were designed for simplicity and consumer pricing, with security as an afterthought at best.
This structure, with its skeletal consistency and solidity, is essential to enabling the mainframe to reliably process vast amounts of data without spending all of its cpu time just inputting and outputting data.
Of course, as part of an environment designed to be highly robust, all of this hardware is of the highest quality, and not merely commodity as one is prone to get with PCs. While that means it costs more, it pays for itself every nanosecond of every day of every year with a mean time between failure ("MTBF") measured in decades. And it means that the next layers on top of it can be optimized for this most reliable of environments to bring even greater benefit from it.
As I think of my own species of vertebrate, and all the things we've built using our bodies, such as language, culture, art and science, I can appreciate the value of being able to optimize for a highly-functional, reliable hardware platform.
Of course, the world needs all different species, including such inveterate invertebrates as the amoeba. Likewise, the world of IT needs single-purpose-oriented computers that allow for the conceptual simplicity of having a platform all to yourself. Just like there are more amoeba in the world than there are vertebrates, it makes sense that there should be more PCs than mainframes.
But when only a mainframe will do, it's nice to know that this highly-reliable, robustly-structured, leading-edge, proven hardware platform continues to be the backbone of the world of business IT.
Now, speaking of quality and reliability, I'd like to take a moment to express appreciation for the comment on last week's blog entry from Jim Michael, a friend and valued mentor of mine who continues to offer me greatly-appreciated support and guidance, and was a significant factor in my increasing involvement with SHARE.
Next week, I intend to talk about the next layer up from the hardware: the operating systems that make the mainframe great.