what mainframe computer is and how it works
In: 33
When technology was new a company could only afford one actual computer, called a mainframe. Today this would just be called a server, but back at the beginning it took up an entire room and was expensive.
Every employee used a “dumb terminal” like a Digital Equipment Corp VT102 which had literally just enough logic inside to provide a text interface “on” the mainframe over a serial line (dial-up or hard-wired “nullmodem”). So instead of a network cable there were serial cables running to every desk. When actual desktop computers became a thing, everyone connected into the mainframe with a Terminal Emulator app over the same old serial lines.
So it was very similar to current “cloud computing” where everything actually happens on a server somewhere else and cheap low-thought “terminals” can then provide a “screen” but do no actual tough processing themselves. So then they don’t need a lot of RAM or a fast complicated CPU of their own and can be very inexpensive and simple to replace (user environment lives completely on the server, if the access device breaks you simply replace it with another clone, no personal data need be transferred or backed up, etc). The only real difference is that Remote Desktop provides full GUI while mainframes were text-interface only.
Mainframes are both an important part in the history of computers in general, and a specific, and rapidly disappearing type of modern hardware. Back when computers were first invented, they were huge, power hungry and very expensive. There was no way everybody was getting one on their desk, so organizations would get a small number of computers, put them in dedicated rooms / facilities, and their employees would sign up for time slots to run their programs on the computers.
Eventually computers got sophisticated enough that multiple peolpe could use them at once, and everybody got tired of carrying around stacks of punch cards, so terminals were invented. There were many different types of terminals, but all of them were essentially some variation of a way to display text and a keyboard, connected to a computer in a different room, but still fairly nearby using wires. Now, you have a mainframe.
As computers advanced, and hardware became cheaper, more powerful, and smaller, development reached a crossroads. Either you could use these advances in technology to cram more and more processing power into a centralized computer, or you could make a ton of smaller computers. Personal computers spawned from the latter path, but mainframe development never really died out.
With the market for personal computers becoming so much larger than mainframes, economies of scale kicked in and it very quickly became cheaper to get the equivalent amount of computing power out of a lot of smaller computers than one big one. The servers that power almost every site you visit really aren’t fundamentally different than the computers you’re used to using, they just have a lot of redundancy, and are in a form factor that’s more conducive to being in a datacenter with tens of thousands of other computers.
However, mainframes still have their applications. Imagine you’re running a stock exchange. It’s really, really important that trades that interact with each other are processed in the correct order. This is fairly simple to achieve if you just have one server. However, what happens when you have more traffic than a single server can handle? If you were doing something like streaming video, you’d just add another server, but if you try to add another server to your exchange, what happens if two conflicting trades get sent to different servers?
Certainly, you can make the servers talk to each other and synch up, but that’s both a lot of work, and very slow. What really would be great is if you could just have one big server, more powerful than any personal computer, and capable of handling insane volumes of real-time data all in one place. Well, that’s where a modern mainframe comes in. They are used for tasks such as handling airline reservations, and credit card transaction processing.
Why are they dying? Because regular servers are always getting faster, and the software industry is always getting better at making them work together. Servers based on PC hardware will always be a lot cheaper than mainframes, so if something can be done on 100 servers rather than 1 mainframe, eventually it’ll be converted.
I think that a lot of ‘explanations’ here never really address the difference between a mainframe and just like a regular beefy (set of) x86-based server(s) (filled with Xeons or Epycs).
From what I understand is that they are tuned for extremely high volume transaction processing although I can’t explain how this is achieved.
Also the redundancy / resilience is extreme with support for hot-swap CPUs! To prevent down time.
Somebody must be able to explain this better than me.
The important part of Mainframe is the word “frame”. Back in the day when a large computer would have a room dedicated to it, the “frame” was what was inside the cabinets and was used to attach storage and other other peripherals. Think of a classical computer room from the 60’s and 70’s.
As computers got smaller and more powerful, the term drifted to simply mean a large scale, powerful computer. Typically with a lot of storage and usually located in a different place than the users.
Weirdly, mainframe computers were not necessarily “faster” than smaller computers, but they typically were designed to input/output large amounts of data and had a lot of storage. They would process transactions for banks, insurance companies etc. The computation is easy, but the amount of data moved was quite large.
They were the first broadly used machines to handle I/O in hardware. Smaller computers might have used software to manage network data, disk I/O, etc. A mainframe would have external hardware to handle that data being moved, freeing the computer to only have to do the actual computation. This allowed a mainframe to outperform a smaller computer of similar CPU power, by 10x or more.
Today, all computers, even phones, delegate the I/O to dedicated hardware, so the term mainframe really just means a large computer that does a lot of data-processing that is usually located in a different location than the users.
In the early 80s a mainframe was a computer that had a CPU that wouldn’t fit in a single cabinet (processor, memory, and input/output were often separate), a minicomputer fit the CPU in one cabinet, and a microcomputer fit everything in one cabinet.
At my job we have a bunch of machines producing goods that inspectors pick off the machines and package. Each machine has a windows computer at the end the inspectors use to log what they took and the quality of the product.
Now we have a lot of these computers; it’d cost a lot of money to buy a windows image for every computer so instead we have one large linux computer with UNGODLY amount of resources running virtual instances (virtual machines) of the windows os and then sending the gui to the machine terminals.
To answer the question, the computer with ungodly resources simulating other computers and passing the screen to another machine is the mainframe.
What are some examples of IBM Z in use today? IBM Z is used by 44 of the top 50 banks and all top 10 insurers worldwide, as well as a large number of government, healthcare, airline, and retail organizations and other companies that require the highest standards for performance, security, reliability, and availability.
IBM zSystems has its own architecture which is designed to work with huge caches, and it has instruction sets to handle enterprise workloads by using IBM’s own S/390x chip architecture. The S/390x architecture is supported by several popular Linux operating systems including SUSE, Red Hat, and Ubuntu, as well as a large number of open source applications.
IBM zSystems runs operating systems including z/OS, z/VM, z/VSE, z/TPF, and Linux. In fact, it is common for multiple operating systems to run on a single mainframe. LinuxONE is an IBM zSystems system that’s specifically dedicated to running Linux; this combines the benefits of the Linux OS with the capabilities and strengths of the mainframe like outstanding data security, availability, performance, efficiency, and cost savings.
Imagine your desktop computer was all your company / school could afford to buy.
Now imagine that you could get really long wires for the keyboard and screen.
Now imagine you could plug two keyboards and two screens in, but have those in different rooms.
Two people could use the same computer at the same time.
It’s that.
Fun fact, the current computing infrastructure of large VMware servers are in fact mainframes all over again.
Lets say in your job there are occasionally really hard math problems that need to be solved. So the company has someone really good at math sit by your desk and solve the problems when they come up. But people who can solve really hard math problems are hard to find and get paid a lot. Since the problems only come up occasionally, the math person isn’t doing much most of the time.
So your company decides that instead of having a math person at every desk, they will set up a room full of math people. Instead of having your problems solved at your desk, you will send them to the math room where they will be assigned to someone to solve. Since multiple people can use the same person they need less math people and the ones they have will be working most of the time. Plus, when you have a really really hard problem multiple people can work on it together.
That is basically what a mainframe is. A computer designed to handle a large amount of operations as quickly as possible for things like data and transaction processing.
Latest Answers