What are Mainframes?

Mainframe computers, also known as “big iron,” power things from credit card processing to airline ticketing. How do they work, and what makes them different from other large-scale devices like supercomputers?

TunnelBear message: TunnelBear is the easy-to-use VPN app for mobile and desktop. Visit http://tunnelbear.com/linus to try it free and save 10% when you sign up for unlimited TunnelBear data.

Follow: http://twitter.com/linustech

Join the community: http://linustechtips.com

Thanks to Connor Krukosky for his assistance with this episode.

License for image used: http://creativecommons.org/licenses/by/2.0/legalcode


46 thoughts on “What are Mainframes?

  • For some reason it is impossible to explain a journalist the difference between a supercomputer and a mainframe. They claim that the public is unable to understand the difference and as a consequence they have an obligation not to understand it either.

  • I am glad you created this video. I used to work on a pair of mainframes at a large global car manufacturer and I assisted there with the mainframes SNA connectivity amongst other things (SNA is kind of like TCP/IP but for mainframes). Mainframes are often kicked in the teeth because PC's obviously outsell them, are ubiquitous and few have true experience with them in order to make a realistic comparison to a PC. As you say, they are architecturally almost completely different and built from the ground up to work with a different kind of load to what is PC is meant for. Because of a mainframe's architecture, it can support hundreds if not thousands of users with a 1/10th of the RAM a PC would need for it to battle to cope with just one neurotic high-end user. But of course, a mainframe ain't built for making pretty pictures like a PC is. These machines live in a different but parallel dimension

  • Nice of you to assume I want a mainframe for playing games. I'd instead want one, perhaps, for use with an MMO, working with Eve Online-esque cluster servers.

  • I used to work at a company that WAS NOT IBM, that built mainframes, some of them were the large rack type that were 6 or 8 feet tall, and FILLED with boards and processors, other mainframes were small rack mount types that were used in aircraft, and ships that like you said were for redundant functions!! We even built the related boards and hardware that went into those mainframes as well!!! And part of my job was "box building" which was actually putting the wire harnesses and the metal cabinets themselves together!!
    Sorry for those folks who hate people who edit their posts….but I had to add this here!!! Linus you forget mainframes are NOT only used at stores or other traditional businesses! Mainframes IN FACT are still a major part of places like NASA sure….BUT they also run Google and Amazon as well as Ebay, so there are a lot of places that use mainframe computers, and at a different company I worked at we built printers for companies like Target retail stores, and we had a room in the build known as the "IT Jungle" that was about 20 foot by 20 foot room with probably 30 mainframes STUFFED in it, that were all daisy chained together for our "in house" computer network AND print function back up!! In fact ONE of the mainframes was set up as our own personally "in house" server, and with a few clicks of a mouse on ANY computer, they could update ALL the software to the mainframe that ran all the other networked printer test computers!!

  • Video idea: why didn't cpu building companies just make CPUs larger to cram in more transistors to make them powerful?

  • This is awful shrieky at times. As for the premise, that you need super expensive, large boxes from IBM to do all that I/O and deal with the sort of transactional rates and databases used by banks is wrong.

    I used to work a low with machines like this in the telecommunications business. At one point, it was the only way of handling big problems like itemised billing, and the bazillions of different things required to keep vast and ever bigger networks and telephone systems going. Nowadays its different. There are ways to architect systems that don't require such machines. That, and the capabilities of modern system based on cheap commodity processes are often more than these IBM monsters. Indeed, IBM themselves make them. They just aren't so lucrative as a wholly proprietary architecture.

    In the meantime, boxes running UNIX, LINUX and Windows pioneered things like Fibre Channel, gigabit and then 10 gigabit Ethernet, and infiniband. When I left, our massive multi-Petabyte systems were running on UNIX or Linux systems using boxes with the power that massively outperformed the mainframes of just a few years before. The datawarehouse machines were moving multiples of gigabytes a second. Meanwhile, the old mainframe apps got maintained, but only for legacy reasons as nobody was prepared to re-write thm

    I can guarantee that mobile phone networks aren't run by these things, no those massive online market systems, such as eBay and Amazon. Indeed, when it comes to all those card transactions, then they don't get processed by mainframes in the first instance.

    I'm also reminded that when a UK bank (Natwest) had its biggest melt down a few years back which denied access to accounts for many for many days, it was the mainframe system that was at the centre of this, albeit shockingly shoddy off-shored operational support compounded the problem.

    The reason that mainframes are used in many banks is not that it's the only way of doing it. It's because of the legacy software that they dare not, or will not rewrite for more modern platforms.

    As for this stuff about I/O cards having their own processors well, of course. That happens on every decent sized machine out there. Indeed, a modern decent desktop PC is going to have processors in some of its I/O cards. It's just they've all shrunk.

  • In the late 90s the Cleveland airport still had vintage plain text computer monitors showing flights to travelers and large dot matrix printers at gates. Y2K and 9/11 really sped up upgrades, but in truth that old stuff was probably working just fine.

  • IBM Mainframes supported Graphics Terminals using software called GDDM. The API was similiar to OS/2 Presentation Manager GPI. Actually GDDM came first, so GPI was probably based on it. Mainframe graphics were not very widely used.

  • Nothing against Linus, but this is a pretty horrible explanation at what mainframes are. Remove the drama…remove the over-excitable talking, and just explain it simply without trying to create hype to keep someone's attention….he lost mine.

  • 8,000 (working) Linux servers on one mainframe? Yes.
    Re-entrant code. IBM's VM software allows each server to run off one, (1) copy of the OS.
    VMWare tried this but it is complicated.
    Yes we need COBOL programmers. If you want to make 70K a year and up, start learning.
    As for DBAs and system programmers – salaries are higher but it is very complex.

  • Long time ago I worked on a mainframe, it was about 4 feet high and 8 feet long and 4 feet deep, not counting tape drives. It had core memory boards that were a foot square and each had 1K bits. It ran at about 300K bits per second. The drum storage held about 600B bytes of memory.
    About 20 years ago I got to work with a Cray. Much faster.

  • So hacking into a mainframe is possible but I have to be physically able to access the machine
    Or I could just put some techno music on and let the sounds do the work

  • I once worked for a place that had a Stratus computer. It had three banks: Memory, CPU, and I/O. Each card was paired. (There were like a dozen of more cards in each bank.) If any of the cards died, say a CPU card, its partner was already doing the same processing.The computer would automatically call Stratus Corp, who would overnight you a new card. It would be on your desk in the morning. You take it to the Stratus. Open the big door. Find the card with the red light. Pull it out. Slide the new one in. In literally just a few seconds the green light comes on and you're done! (Well, close the door, too.) Zero down time. The users never know.

    I find this totally amazing. Can you swap your CPU out with the machine still up and running?

    Very cool. But also very expensive. I guess this was like mini-mainframe, but it's primary mission was fault tolerance. Everything was in pairs except the backplane and the tape drive.

    Disclaimer: I do not work for Stratus. I do not get any compensation whatsoever for posts like this — or from anyone, AAMAF.

Comments are closed.