"Today’s System z, itself an evolution of the original System/360 and 370, can still run many of the original programs, unmodified, from 50 years ago. This is a testament to 2 things, the wide adoption of the IBM systems, and the forward thinking of IBM."
Or rather the crazy levels backwards compatibility and effort to build what is basically a hardware emulator for a 50 year old machine. Put another way, IBM isn't making native ficon disks, they build a fancy controller that uses SAS/etc disks and presents them as ficon, which runs on the same physical transport as fiber channel, rather replacing much of the protocol stack. The resulting MOD3 or MOD9/etc disks appear like 3390 model 9 disks circa the late 80's early 90's. AKA instead of a room sized disk they are basically virtual with all the overhead that implies.
The CPU's are of course native, but like all the other ISA's reflect a more modern design style of basically cracking the CISC instructions into an out of order superscalar CPU and only using "micro code" for the really complex instructions.
while the iSeries (think AS/400) has a shorter history many programs compiled to earlier versions run on later versions due to the separation of hardware layer from operating system layer. The system can recompile existing programs without human usable/retrievable source due to the method use to encode them.
there is code from the eighties running in slightly updated form on our z and even the i series systems have twenty year old code. both systems have web facing extensions that directly map to older code (as in, green screen 3270/5250 that has no change to its code) pretty much seamlessly. new code is often employed as necessary but the idea that stuff that worked well that is a decade or more old still works today is what amazes me about the various IBM platforms
>""Today’s System z, itself an evolution of the original System/360 and 370, can still run many of the original programs, unmodified, from 50 years ago"
Wasn't this the value proposition of System 360, that IBM guaranteed that software for it would be supported in perpetuity basically? This is why there are still so many IBM main frames around in government running COBOL.
Here's an interesting top 10 list of the oldest system in the US Federal government. Look how many of them are IBM Mainframe/COBOL:
That would have been pretty useful during my college years. Most of the classes were on a IBM 370. I still have my assembly book and the assembly banana book. XEDIT was an interesting editor.
It was an interesting era since pretty much everyone trudged over to the computer lab to use the terminals to do their assignments. It does generate a bit of community in groups of people going through their degree at the same time.
It would have been interesting if IBM had got Motorola to second source the chip and used it instead of the 8088. Would have been a very different PC era.
Nick gave an excellent talk about this chip at the 1995 Microprocessor Forum (which was around the 25th anniversary of the IBM/370). It was really impressive that IBM could make the PC using the 8086 architecture because there was such corporate alignment around the 3x0 architecture for all computing that was IBM. Nick mentioned that in some ways the PC division felt threatened by the notion that he had created a way to have the 370 architecture go "from top to bottom" on their product line.
The PC was created due to the consent decree. IBM was forced to use other people's processors, OS, hardware interfaces, etc. If it wasn't for the consent decree the world would look very different.
The /370 vision of a single chip mainframe was realized, I think the article ends prematurely. The first card I've actually seen as such is the P/370 and that may have been a productized version of this chip. There were some other skunk works projects that might predate that http://www.garlic.com/~lynn/2002d.html#4.
As others pointed out "real" S/390s also switched to CMOS microprocessors in the 90s. There's some good reading in http://ps-2.kev009.com/ohlandl/P390/s390_i-o_continues.txt on the evolution of the I/O subsystems because I/O is what makes mainframes interesting :)
I had the joy of using the XT/370. It was extremely cool to see the "mainframe" run in a box. But it was dog slow compared to a single user session on the mainframe (which IIRC was a 4341).
The 360/370 is/was an incredible architecture. It supported virtualization decades before anyone outside a small group of people even knew what that was. The instruction set was rich and easy to use which was important as a lot of stuff got written in it directly those days. Unlike Unix which was mostly written in C most of the 370 OSes of the time were written in assembly.
That's neat as heck. A mainframe on a M68K for under four grand? I'd have believed the emulation but not the price if someone told me. The RISC workstations were so expensive. Add the word mainframe to a lower price than that & I'd assume you were selling me a lemon. Haha.
Note that the IBM 9672 was the first actual mainframe with a CMOS microprocessor circa the early 1990s. Now all mainframes are powered by microprocessors.
"Not all processors are designed this way, some designs,known as random logic, the op codes control the actual hardware directly"
An interesting note is that the name "random logic" comes from the fact that when you look at a photomicrograph of a hard-wired CPU design you don't see any discernible geometric patterns like you do on a microcoded CPU.
Right, minicomputers were the fridge-sized Vaxen and their ilk.
(Speaking of fridges, that was the era of 'computer rooms' with raised floors and positive pressure air conditioning at a fairly small scale - might have one serving one or two dozen people. Comms were poor.)
Or rather the crazy levels backwards compatibility and effort to build what is basically a hardware emulator for a 50 year old machine. Put another way, IBM isn't making native ficon disks, they build a fancy controller that uses SAS/etc disks and presents them as ficon, which runs on the same physical transport as fiber channel, rather replacing much of the protocol stack. The resulting MOD3 or MOD9/etc disks appear like 3390 model 9 disks circa the late 80's early 90's. AKA instead of a room sized disk they are basically virtual with all the overhead that implies.
The CPU's are of course native, but like all the other ISA's reflect a more modern design style of basically cracking the CISC instructions into an out of order superscalar CPU and only using "micro code" for the really complex instructions.