Slashdot Mirror
O'Reilly On The Importance Of The Mainframe Heritage
theodp writes "After exchanging e-mail with mainframe software pioneer Mario Morino, Tim O'Reilly writes 'It's important for the open source community to look more at the software heritage of the mainframe era.' O'Reilly might want to take a look at how Marino's own MICS software has been used since the 80's to automatically charge IBM mainframe users for printed material that could be ordered from PC clients with a single action by using billing and shipping information that was previously stored on a Mainframe server. The whole process might seem oddly familiar."
I'm sure many minis and possibly fewer mainframes are sitting in storage and waiting for a trip to the dump. The companies waiting to dispose of these monsters would probably love for you to haul it away, rather than them discarding it and having to pay the tipping fees.
:)
I could have gotten my hands on an old MicroVAX in this manner, but when I figured out the performance (or lack thereof) I would be getting, the lack of any real support resources when/if the thing broke, and most importantly -- the estimated electricity usage -- I smiled and declined.
It _would_ have been cool, though.
probably quite a bit... maybe a few ten thousand dollars? remember people are still using these beasts...
In IBM land the generations go like this:
S/360 series.. the OG of the mainframe world, these are what you think of when you think of government and scientific computers of the 60's. They are notable because they are the first "modular" computers, meaning they weren't custom made 1-off jobs, but you could still upgrade the proccessing and i/o power as you need and afford it.
S/370's replaced the s/360's and were used up thru the late 70's 80's and still a few operate into the 90's. At work we use s/370 era hardware and disks. You know tape reels, 200 meg 3480 18-track tape drives, IBM "infowindow" and 3178c terminals, and stuff. These things were built like tanks and have been running practicly 24/7 for 15-20 years.
S/390's are the modern versions, which is what we use for the actual computer power... All modern mainframes from IBM are these.
SO... you could probably get one fairly cheap as far as IBM mainframes go, but good luck finding parts and your going to need to a power supply (from the power company) on the same level as a small apartment building. Oh, don't forget the several thousand dollars a months for the OS's and such, but if you got a s/370 or later you can run linux on it!!! (still have to run a IBM OS though to run the linux OS).
Pretty expensive for the same proccessing power of a 486 PC. Put I/O is still unparralelled for the modern stuff .
A lot of the programmers and project managers put in place by IBM and Sears were old school mainframe guys (and gals) and they did some groundbreaking work, including adapting a TPF system (Transaction Processing Facility for airline reservations) for delivering content and email messages. They built a huge place in northern Westchester county to handle an anticipated ten million members. I have to give the rank and file people at Prodigy a lot of credit for going beyond mainframes and adapting to Unix, AIX, TCP/IP, etc. and making it work pretty well. A series of incompetent managements sank the good ship Prodigy. The people rowing the oars always did their best.
Set up a Vax system in his house as a prank, as I recall.
why not think about an old iSeries (AS/400)? They go for 1000$+ on ebay, are relatively modern, the manuals are on the web and they can do email/webserving (i think the more recent ones run linux, too). just be sure to get a risc model(preferably with the os V5Rx or at least V4Rx), and not one of the old grey cisc-"i can't do anything anymore"-models...
Rather than increase your electricity bill further, you could try a 370 emulator, I know there are commercial ones about, but a quick google revealed the one below.
http://www.conmicro.cx/hercules/
Stu
I'm a toolie from way back with a few contributions of my own, SEARCH, a flat file database query tool (used Boyer-Moore string search to make it real fast), and REACC/QUACC, a command that let you determine whether a R/O CMS disk had changed and needed to be reaccessed.
Also some that never got off the ground. I had this idea to emulating temporary files without doing actual i/o to disks. Couldn't call it virtual i/o since there was already a mainframe i/o method called that. I described it to a friend who said oh yeah, unix has those, they're called file pipes. !!? This is the late 70's or very early 80's when unix was basically unknown at IBM. So a first for unix in that case. File pipes were cool and I was probably the only one who had them on mainframes at that time.
So yeah, other than that, we mainframe guys invented everything first. But we never believed the stories told by those ex Future Systems guys. They claimed they invented everything first.
At least the IBM mainframe and CDC projects I worked on were not open source.
They were huge though, we often had more than 400 programmers working on the one system, each working on their own little corner (well you prayed you didn't overlap). The banks and insurance companies had even bigger teams.
I am vaguely aware that the DECUS (DEC user group) kept VMS going on their own while Digital then Compaq then HP tried to decide if it was profitable or not. I think the obstinate customers who insist on something reliable have swayed HP now. Corporate policy is something like: "as long as you don't cost us any money, you can do what you want". They might change their minds if they see a profit in it again.
Not necessarily; it all depends how old the machine is. I remember a time (back when the Earth was newly-cooled, and Real Programmers got their assembly-coding done while trying not to get eaten by dinosaurs...) a shop I worked at decommissioned a Burroughs B3700.
The contractors were only too happy to take it away, as there was enough 3/4 inch-thick silver cable under the floor to fill a small truck. I'm pretty sure they would not have lost money on that operation.
Ya, but the iseries aren't realy mainframes. Their midrange servers...
There realy is a big difference. A server is designed for serving other computers, providing services and small to large files... a mainframe is a standalone thing, maybe with a couple supporting PC's to aid in network functions and stuff like that.
Mainframes are dinasours. Tiny brains, big bodies. You have the central housing for the cpu, but then each device has a disk controller that controls and operates each peice of equipment...
Think of it like a video card, but instead of crunching 3-d stuff it's crunching I/O, each device (the controller for a bank of tape drives, or disks) operates itself and the central computer just tells the bits were to go, it stays out of the loop as much as possible.
A example would be:
This 3,000,000 phone numbers get added to existing records and sorted to the zip code, and sent to this disk volume, then gets divided by state, then each state volume gets backed up into tape drives for backups... Then a week later a client wants at least 4000 records from wisconson that represent people who are male, make over 30,000 dollars and are bow hunters. You find 1200, but then they want more so you add some zip codes searches from surround communities and find enough records. They pay for the records and then you zip the volumes into a single file print out a hardcopy of it's contents and it gets ftp'ed to the client's print company for mass mail flyers.
Don't need much cpu power for those jobs, but you need tremendous thrurougput to do this and 8 other jobs like that at the same time. Mostly it's shuffling data around and appending records.
That is the world of mainframes...
Even the brand new s/390's (t-rex stuff) getting stomped on in terms of proccessing power compared to the power4 servers....
Supercomputers on the other hand can do tremendous amounts of I/O AND have lots of proccessing power. so they can do complex mathmatics with large amounts of data on a tremendious speed. Both a high-end server and a mainframe would choke on that.
So that's the difference between a high-end server(lots of cpu power), a mainframe(lots of I/O power), and a supercomputer(mega lots of both).
Linux should always be questioned, just like the use of every other tool. Linux is fabulous, but it is not all powerful or appropriate for every situation. You should always consider the situation before making a decision about the tool to use.
... ITS LINUX!! IT RULEZ!! No discussion needed d00d.
Knee-jerk Linux advocacy isn't any better than unthinking opposition to Linux. I actually think its worse. When someone simply opposes Linux, there is probably another solution that either already works, or could be reasonably expected to work on Unix or Windows. An unthinking Linux advocate may push a solution which is unworkable.
I recently had a discussion with a coworker who though that Linux would be the best OS to use on an Itanium server. I found to be an astonishing view. For our industry there is a growing amount of vendor software that runs on Itanium.... under HP/UX. Hundreds of packages run under HP/UX, but I can only think of about 2 that run under Linux. Linux would clearly be a poor choice for us, and yet that was the automatic answer from my coworker. After some additional discussion it turned out that he didn't really know about the applications, but basically assumed that Linux could do it. I think that is way too common among Linux users.
Frankly, contrary to you, I think Linux will have made it when it is questioned (like every other IT/engineering solution should be) and chosen as the best solution for the problem, and not because
early 70's ( of course no orders from PCs - didn't exist that time but ). Working in an insurance company, our customers, banks, city/county fleet managers, shipping companies, etc.. were able to order print jobs and were billed eletronically based on information we had of them. Small jobs could even be routed to their RJE printers - slow, or sent on tape. Later on at start of 80's I had mainframe customers selling print services in their IBM 3800 laser printers ( 215 pages/min ). It's impressive to see 4-5 of those printing almost 24h / day. Almost all the orders came over network ( BSC, x.25 or SNA still at that time ) and were billed on their bank by paper type, pre-processing requirements, layout requirements and post-processing, etc.. All the customer information was, of course, in mainframe databases. The real challenge was the world wide requirements in a bank to get the print to follow the user independent of the printing subsystem, IBM, Tandem, Honywell, Prime, Dec, whatever and independent of the protocol, BSC, X.25, SNA and later IP. To authenticate a user and printer in foreign country can be interesting and is not possible without very detailed information of user, network(-s), routes, equipment, etc. and of course of the paying organization or department for inside jobs.
I recall the huge uproar when IBM decided to withdraw access to the source. It was called their "Object Code Only" (OCO) policy and users were outraged. I still have some of the coke can wrappers passed out at a users group meeting making fun simultaneously of OCO and New Coke. (From a Google search, found this which references Feb. 8th, 1983 as the date of the OCO announcement.
There was a very active community within the major IBM users groups such as SHARE, sharing modifications to the systems. The best collection at the time was the CBT Mods Tape which was originally assembled by a Systems Programmer at Connecticut Bank and Trust. I guess it doesn't suprise me that it still exists (Thanks Google!) as it was an invaluable tool back when I was still involved with mainframes.
A generalisation: some machines were engineered that way, more commonly as LSI came in during the 1980's. But the customers weren't entirely ignorant about it: it wasn't primarily a technical issue, but a matter of contracts and resourses - factors that members of /.'s army of part-time boy programmers may eventually understand when they've had to earn their living in the trade for a few years.
Sure - if they had their own engineering organisation, trained to maintain the machines they used, and were willing to take full responsibility for any and all unanticipated consequences of what they did. You needed to be a pretty big organisation with pretty unusual requirements for that to be worthwhile.The vendors charged what they reckoned the market would bear. Big customers did better than small ones. Another aspect of the mainframe era that is being repeated today - I'm sure I don't have to give any clues about the particular platforms and products involved.
Dr Hu - who worked with mainframes as recently as the early 1990s.