from the There's-nothiing-new-under-the-sun? dept.
http://www.winestockwebdesign.com/Essays/Eternal_Mainframe.html
In the computer industry, the Wheel of Reincarnation is a pattern whereby specialized hardware gets spun out from the "main" system, becomes more powerful, then gets folded back into the main system. As the linked Jargon File entry points out, several generations of this effect have been observed in graphics and floating-point coprocessors.
In this essay, I note an analogous pattern taking place, not in peripherals of a computing platform, but in the most basic kinds of "computing platform." And this pattern is being driven as much by the desire for "freedom" as by any technical consideration.
"Revolution" has many definitions. From the looks of this, I'd say "going around in circles" comes closest to applying...
-Richard M. Hartman
A funny thing happened on the way to the future. The mainframe outlasted its replacements.
[Ed. Note: This story submission was my first exposure to the linked essay. Though dated from 2013, I found the essay eminently readable as well as making insightful observations of how dramatically the concepts and capabilities of mainframes have persisted for so many years. --martyb[
(Score: 1, Interesting) by Anonymous Coward on Tuesday September 17 2019, @04:37AM
I made my living with IBM MINIs. They sames out fo the IBM 360/370 line.
They started as: IBM S/3 models 10 (single "user" and 15 "tri-user" mainly. 8bit with 16bit addressing card systems, used in mid-range insurance and nuclear site construction control.
then came the S/32, simple single user OS with 8x40 char screen and line printer builtin in front of keyboard. Think Large long desk with 220v power plug.
then came the S/34. merge of S/3 model 15 and S/32. Multi-user (28 workstation directly attached and 32 remotely attached). Same simple OS as S/32. Damn thing - IBM tried to kill it 3 times and kept going and going and going. Near indestructible.
then S/38 - now that was left turn! Program call stacks, Even more display types, and still the ever loving card reader from S/3. Was real next set for S/3 model 15, since the CPF "time-share" system nor the LARGE disk platters were not supported directly on S/32 or S/34 - thought there are storiies of guys in L.A.
then S/36, same indestructible build as the S/34 and a read next home of all the software that was written. Also 4 models from size of IBM AT - Bady/36 (5364), a PC tower (5363) and small two-drawer file cabinet (5362) and Large washer/driver combo (5360)... all with 1M of memory. Then a 5360-7 7M of memory.
Then came the AS/400 - S/38 OS made friendlily of the S/36 crowd. But to costly to get S/36 market.
Then the family line up: AS/400-236 - only ran S/36 OS, but with new processor. AS/400-436 ran both AS/400 partition AND S/36 partition to finally get the rest of S/36 converted to new machines. Along the way merged in AIX platform so can run 3 OS at one smae at the same time AIX, S/36 and AS.400, Then added Linux as OS of choiuce and dropped S/36, since AS/400 could run most all that code nativity now.
The "workstation" I ever seen and IBM p with 64 cores and 4T of memory, sitting next to desk at the NOAA. With a PC stock PCIe16 graphic card and PCIe1 USB card for keyboard and mouse. Oh, the actaul size is 2 big Refridators with a 650lb UPS on the top.
S/36 on also supported Intel cards, so they could also run Windows inside the case with host OS. Also cards for PC to hard wire to backplane, so the disk drives are in the IBM "mini" (DAS?).
IBM Minis (mid-range) was as small as PC and as large as small main-frame - with overlap!. Now the fastest computer in the world has thousands of the Power chips running them.
(Score: 2, Funny) by Anonymous Coward on Tuesday September 17 2019, @04:38AM (5 children)
TFA:
Actually, there is only one natural kind of computers: embedded ones. A mainframe is a computer embedded in a building.
(Score: 0) by Anonymous Coward on Tuesday September 17 2019, @04:55AM (4 children)
Except at this point IBM mainframes are about the size of a refrigerator.
(Score: 3, Interesting) by NotSanguine on Tuesday September 17 2019, @05:32AM
Actually, IBM mainframes [wikipedia.org] are of varying sizes depending on the model.
In fact, I was installing S/390 cards [wikipedia.org] into RS/6000s back in the mid 1990s.
The irony was that while the S/390 processor was an MCA [wikipedia.org] card installed in the ~30-40kg RS/6000, the tape drive (IIRC, a 3480 or 3490 cartridge tape unit) for the processor card weighed ~400kg. I actually had to contact the building managers to make sure the machine room floor (not in a data center -- long story) could handle that monster.
Heck, the bus and tag cables [wikipedia.org] for the tape drive connection were several times heavier than the S/390 card.
No, no, you're not thinking; you're just being logical. --Niels Bohr
(Score: 1) by khallow on Tuesday September 17 2019, @03:11PM (2 children)
Ah yes, the portable mainframe.
(Score: 0) by Anonymous Coward on Tuesday September 17 2019, @03:36PM (1 child)
Yes! As/400 P2. Size of a laptop!
(Score: 2) by driverless on Tuesday September 17 2019, @07:30PM
Not a laptop, a flattop.
Friend of mine worked with an IBM laptop mainframe emulator but he called it "the flattop" because it was a laptop the size of an aircraft carrier. Took him several days to find a backpack large enough to carry it around in.
(Score: 2) by DannyB on Tuesday September 17 2019, @02:46PM (5 children)
Okay, so modern server rack farms in air conditioned data centers are "mainframes".
But not really. The word "mainframe" carries a lot of baggage. Baggage which is decidedly NOT part of modern server racks.
First would be the backward compatibility with late 60's early 70's mainframes.
Second would be the "mainframe vocabulary", a set of terminology that nobody uses except IBM, their customers, sales people, and organizations that are built around supporting IBM equipment. I would begin with the term "core memory".
Third would be that traditional mainframe design is quite different from modern server racks. Mainframes try to make everything ultra reliable with some redundancy. Modern server racks try to use reliable equipment, but the basic design thesis is that commodity hardware will fail in fairly predictable statistical amounts, at scale. Eg, you can absolutely expect to have to replace XX failing hard drives, every single day of the year. No need to make all the hardware hot swappable, because individual racks or even specific server units in a rack can have their workloads migrated elsewhere within the data center, then be powered down, replaced / services, and then powered back up again.
Fourth would be the secrecy, proprietary nature and monopolist sourcing of mainframes. Commodity modern server racks have multiple sources for everything. Open source is the norm. Even open data center design with published rack designs, power supply designs, etc.
So while there is a surface resemblance of modern data centers to mainframes, it is only a superficial resemblance, IMO.
The lower I set my standards the more accomplishments I have.
(Score: 0) by Anonymous Coward on Tuesday September 17 2019, @03:51PM (2 children)
There's also a big architectural difference.
Mainframes were (once you ditch various fripperies) optimised for moving bits without involving the central processing system at all. In IBM it took the form of channel architecture, but there were other, analogous answers from other manufacturers. Present (bad) reinventions of the mainframe don't do a good job of that. They're like the bastard offspring of old supercomputers (with worse interconnects) and mainframes (with worse general IO).
(Score: 2) by DannyB on Tuesday September 17 2019, @04:04PM
Now that YOU mention it, I realize that I should have mentioned that.
Channel architecture was a significant part of what was considered a mainframe. That being yet one more big difference between traditional legacy mainframes and modern server racks in data centers.
I almost wonder if TFA is really an attempt, in the author's own mind, to tie legacy mainframes to the sexiness of modern server farms.
The lower I set my standards the more accomplishments I have.
(Score: 0) by Anonymous Coward on Tuesday September 17 2019, @04:12PM
But you missing the point of write up. It would be better centralizes processing power / timeshare and decentralized/distributed processing over AN AREA.
With that view AWS is a main frame / super computer. It is centralized and restricted access and should be the TOP co”computer” on the top 500
Yes IBM’s channels were and are great to queue the I/O, feed the processor. Just airlines of old had series 1 to handle the terminals and feed only completed requests to the centralized system Or DECs being I/Ovfor Cray 20’s. Or JavaScript to make sure CC# entered is mathematically correct. Airlines learned that smarter front saved money (about 70%) on main frames since validating at the ends is cheaper than testing everything aon mainframe. Ie processing
(Score: 2) by NotSanguine on Tuesday September 17 2019, @07:30PM (1 child)
Personally, I'd begin with this [anvari.org] myself:
No, no, you're not thinking; you're just being logical. --Niels Bohr
(Score: 2) by DannyB on Tuesday September 17 2019, @07:41PM
I assume that the non-blank characters (as well as the blank ones) are in EBCDIC.
While the rest of the computer loving world uses ASCII.
ASCII stupid question, get a stupid ANSI.
The lower I set my standards the more accomplishments I have.