Talk:Mainframe computer

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Anachronistic photo[edit]

#History has a picture labelled An IBM 704 mainframe (1964). There may still have been 704s in 1964, but any 704 or IBM 709 still in operation would be considered to be a museum relic; more realistic machines for 1964 would be, e.g., Burroughs B5500, GE 635, Honeywell 800, IBM 7074, IBM 7080, IBM 7094, UNIVAC 1107.

There should be a photo for a 1950s mainframe, but the photo in IBM 704 has a more reasonable date, and photos of earlier mainframes, e.g., IBM 701, UNIVAC I, UNIVAC 1101, are available.

Would it be TMI to include a photo for each decade? Shmuel (Seymour J.) Metz Username:Chatul (talk) 17:07, 12 May 2021 (UTC)[reply]

Big Iron[edit]

Doesn't the term "Big Iron" refers to the class of overall large commercial computers? [1] [2] Galzigler (talk) 21:22, 3 May 2022 (UTC)[reply]

Those two One of those sources appears to be contributor-based (i.e. anyone can add a definition), however [3], [4], and [5] are decent WP:RS if anyone wants to add further clarification of the term to the article with appropriate citations. - LuckyLouie (talk) 21:53, 3 May 2022 (UTC)[reply]

z/OS does not provide virtualizing facilities on IBM Z systems.[edit]

The introductory discussion of mainframes states that "Software upgrades usually require setting up the operating system or portions thereof, and are non-disruptive only when using virtualizing facilities such as IBM z/OS and Parallel Sysplex...", but z/OS does not provide this facility; it is IBM's Hypervisor OS, z/VM that provides it. [1] --Zvmphile (talk) 18:25, 25 June 2022 (UTC)[reply]

z/OS has some facilities for supporting predecessors, back to OS/360. It is supposed to be that programs compiled and linked on OS/360 will run on z/OS. Sometimes that counts as virtualization, and sometimes not. There is support for 31 bit and 24 bit addressing space, for such programs. But yes, the usual virtualization is z/VM. Gah4 (talk) 20:39, 25 June 2022 (UTC)[reply]
I would not consider that virtualization; I'd call it "backwards compatibility". The only thing making it different from, say, programs developed for and built on older versions of macOS running on newer versions, and 32-bit programs running on 64-bit platforms on various OSes, neither of which are considered virtualization, is the switch from OS/360 and single-address-space OS/VS to MVS, but even that is more like "look, by magic your partition/region always starts at the same address and there are no other partitions/regions you can poke around in, but most code doesn't depend on that" than virtualization. Guy Harris (talk) 22:47, 2 June 2023 (UTC)[reply]
IBM has offered a feature called z/OS Container Extensions (zCX) since z/OS V2R4. I'm not sure whether it qualifies as full-fledged virtualization. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 14:27, 2 June 2023 (UTC)[reply]
OK, I get how Docker and other containers work when the OS running in the container is the same as the OS on which the container is running - it's not full-blown virtualization, but it does run the code running in the container in a "kernel namespace" separate from the main namespace[2], and it's apparently called OS-level virtualization, so I guess that's virtualization of a sort.
But if the containers are Linux contaiers, with the OS on which the containerized software is running is Linux but the OS on which the container itself is running is z/OS, there's presumably either a way in which z/OS handles Linux system calls made from within the container, which sounds like the same way version 1 of Windows Subsystem for Linux works (it has a Linux system-call handler running in Windows), or some way in which z/OS is acting as a (type-2?) hypervisor for Linux, which would be full-blown virtualization. Guy Harris (talk) 22:59, 2 June 2023 (UTC)[reply]

References

  1. ^ IBM IT Infrastructure: https://www.ibm.com/it-infrastructure/mainframe-os
  2. ^ McCarty, Scott (February 22, 2018). "A Practical Introduction to Container Terminology".

Secure platform includes software[edit]

@Guy Harris: a recent edit added the template {{Vague|date=June 2023|reason=Presumably that's IBM Z machines running z/OS or z/VSE or z/VM, rather than running the Linux later credited with "thousands" of vulnerabilities - or is Linux on Z better than Linux on x86?}} without a talk link. I agree with the comment, and believe that the issue of security of software versus hardware should be discussed here. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 14:20, 2 June 2023 (UTC)[reply]

Whoever wrote this has a warped perception of how terminology gets agreed upon"[edit]

Saving time: 1) "mainframe" as the hand-wavy term that it is, is most strongly defined by the COMPANIES AND CUSTOMERS that CONTINUE to develop and purchase them, obviously in addition to the historical record. IBM, or their competitors are, in this case, the "authorities" whose notion of where to draw the lines is the reference. 2) As it stands, product offerings exist for HPC mainframes that can be easily referenced with face-palm obvious examples. Drug discovery (an industry that strangely depends almost as strongly on low-error, auditable computational power as the banking or securities industiries). There are at least a few mainframes marketed to drug discovery companies that are based on custom silicon, with custom software, and deployed within a custom operating system environment. These are miles short of "supercomputers", with the largest systems fractions of the size of the oldest LINPAC systems many steps away from "clusters" or any other term. THe same is true of contemporary non COTS based Neural Network hardware offerings, which *notoriously* include non-standard and completely bizarre custom software ("Wafer scale single chip" system anyone?). Mainframes are most importantly characterized and distinguished by being PRODUCT OFFERINGS and not just generalized schemes for enterprise scale systems. This is a hard reqirement. ALL mainframes are products developed by company and sold or deployed as either a discrete one-time purchase or service contract. Usage requirements are even more general than transaction-heavy. THey must be high-uptime with low error, auditable or with a comparable functionality, and come equipped with high levels of security. Those factors overshadow EVERYTHING else even compared to high speed or high computational power. You could walk away with that defintion: A commercial enterprise scale computer system that is designed to deliver highly reliable low error performance with high security, among other characeristics that combined toghether cannot be met with COTS based system architecture nor software67.165.123.62 (talk) 19:12, 22 August 2023 (UTC)[reply]

Meaning of "frame": cabinet or...?[edit]

There's currently a little boring drama on Wiktionary about the origin of "mainframe", and whether it really refers to a solid cabinet, or something else. [6] I would really appreciate any evidence from experts, or very old nerds, regarding the etymology. Hit me up. Equinox 04:29, 21 February 2024 (UTC)[reply]

Not sure what a solid cabinet is. I believe many have thin metal removable sides, where the frame is solid, but the sides are not. The sides are decorative, and not part of the structure. I suspect that kitchen cabinets can be designed where the sides are part of the structural integrity, and others that are not. Gah4 (talk) 04:59, 21 February 2024 (UTC)[reply]
@Gah4: The point is what is the "frame". (See Wiktionary page history at "mainframe".) Are you old and experienced enough to tell for sure? I'm passing middle age but "frame" to me means that piece of shit you'd put your webring in. Lordy. Equinox 06:15, 21 February 2024 (UTC)[reply]
Historically, the term mainframe included computers that ran a single job at a time. The frame for the CPU was typically a metal cabinet with a height of 1-3 meters. Typically the 1960s frames used wire wrap backplanes to connect multiple printed circuit cards. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 14:10, 21 February 2024 (UTC)[reply]
I was trying to compare against cabinet. A kitchen cabinet commonly is solid on all sides, though one is an openable door. The frame for a mainframe is commonly metal rectangles that hold the parts inside together. Even with everything closed, there is need for ventilation, so some sides will have holes, or otherwise openings, or might be completely open. Gah4 (talk) 01:33, 22 February 2024 (UTC)[reply]
Power and coolant connections are commonly on the top or bottom. Also, it is common for a door to include wiring. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 15:56, 22 February 2024 (UTC)[reply]
Mainframe is/was a computer housed in a large cabinet [7]. Early units either consisted of or roughly conformed to the framing of a 19-inch rack. - LuckyLouie (talk) 16:17, 22 February 2024 (UTC)[reply]
I worked for IBM for two decades beginning in the late 1970s designing integrated circuits, many of which were for IBM System/390. In the early days of computers, the electronics were heavy and had to be mounted on heavy steel frames that were strong enough to handle the weight. There might or might not be decorative covers. The engineering prototypes of System/390 were operated without covers.
When the electronics became lighter, small computers were mounted in 19-inch racks, a pre-existing standard standardized by AT&T about 1922. Computers mounted in these racks were known as minicomputers. Large computers took advantage of lighter electronics by increasing the number of circuits, and making the cooling more sophisticated, so the heavy steel frames were still needed to mount them.
A possible source of confusion is that mainframe families had a range of performance; while the highest performance family members had steel frames, the lowest performance members might use 19-inch racks or even fit on a desktop. I'm not sure what the structure of the most recent high performance mainframes is. Jc3s5h (talk) 17:27, 22 February 2024 (UTC)[reply]
The IBM z16 still uses a 19"[1] frame. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 21:04, 22 February 2024 (UTC)[reply]

References

  1. ^ }}cite book | title = IBM z16 Technical Introduction | id = SG24-8950-01 | edition = Second | date = April 2023 | section = 2.2 Frames and cabling | section-url = https://www.redbooks.ibm.com/redbooks/pdfs/sg248950.pdf#page=31 | url = https://www.redbooks.ibm.com/redbooks/pdfs/sg248950.pdf | publisher = IBM | series = Redbooks | access-date = February 22, 2024 }}