IBM's Mainframe Dinosaur Turns 40
theodp writes "According to an SFGate.com article, PCs were supposed to kill off the mainframe, but Big Blue's big boxes are still crunching numbers, posting sales of $4.2 billion in 2003. First unveiled on April 7, 1964, the IBM mainframe computer celebrates its 40th birthday this week with a sold-out party at the Computer History Museum." The SFGate article also reveals: "Doug Balog, an IBM vice president, noted that 70 percent of the world's data are still housed in mainframe computers."
Yes. Yes it does.
COBOL is still in wide use. It is even being used with .NET, just to give you some idea of how widespread it is.
libertarianswag.com
That's because you are the one that is wrong. Any and every dictionary I've ever seen has data as the plural of datum. Maybe no one is paying attention to you because they're tired of explaining it to YOU.
My sig is blank, I typed this by hand.
I appear to have a blog. Odd.
Data is the singular. DATUM IS THE PLURAL
:)
Merriam-Webster begs to differ: Etymology: Latin, plural of datum
Although I don't think I've heard many people ever talk about datum either... maybe because it's always always plural. When was the last time you had only one piece of datum?
Place sig here.
At my first startup, one of my first multipeople multiyear Java projects was a mainframe screen scraper ( TN3270 using AWT - example ). I was fresh out of college & totally unaware that mainframes still ruled the planet. Those two years & the huge revenues it brought led the startup to be acquired and made a lot of people really rich ( minus moi, ofcourse :(
Lots of money to be made in desktop-mainframe connectivity.
Mainframes and Minis will be around a long time. To get PC based systems up to their level of reliability, ease of use, and maintainability would turn the PC based system into a MINI.
I have 75 iSeries (As/400) that I oversee. You want to know how much time I spend per week checking up on them? Only an hour or so. I receive reports from the machines when they have problems. If one has a fault it is usually hardware and rarely does the downtime pass a few hours.
Meanwhile the network group (read : uses PC based technologies) is always fixing something and has 5 people dedicated to it compared to two for the iSeries boxes. That doesn't count the PC-support group which supports desktops...
We have 3 mainframes as well, some of the code from these machines has been in use since the early 70s. Some of the code migrated to the iSeries with little but header changes.
But the best, the iSeries has been on 64-bit PowerPCs natively for 10+ years. Didn't have to recompile or change 99% of our code to do it. How long has the PC base world been struggling to get there?
* Winners compare their achievements to their goals, losers compare theirs to that of others.
I guess this depends on how you define "data". The Economist recently described a Berkeley report that 3.5 to 5.5 *Exabytes* of data were produced in 2002. If you believe the unlikely proposition that Blue Glue is holding 70% of that new data, then you have to wonder why IBM only made $4.2B in selling mainframes to store and process that data.
I think they are talking about mainframes that run Linux as a guest operating system on a virtual machine. The real operating system is VM. VM allows you to create a large number of virtual machines, each of which can run Linux or another operating system.
Mea navis aericumbens anguillis abundat
The most widely used flying command and control platform is the AWACS designed by IBM and Boeing back in the 70s. The USAF,NATO,JDF, and saudi's are all based on the same dual IBM 360 platform (named 4-pi). These mainframes all have been upgraded in memory and converted from tape drives to hard drives. We still develope the software in JOVIAL and assembler.Info
Science is the Real TRUTH!
This is the 40th anniversary of a mainframe: the System 360. The 360 was a darned important machine (amongst other things, it was the first computer with a byte-addressible memory), but it was hardly the very first mainframe. True computers had been around for about 25 years -- and technically speaking, all computers were mainframes before integrated circuitry made minicomputers and microcomputer feasible.
The simplest way to think of these two classifications is that
- "Supercomputer" refers to processing speed and is defined differently in different contexts (i.e. Apple calling its G4 400 a supercomputer because of an outdated US Customs document).
- "Mainframe" refers to large systems that many users are going to use at the same time, typically via dumb terminal interfaces. Most importantly, mainframes have IO architectures which blow any desktop/workstation out of the water. A good mainframe can be talking to 500 terminals while printing 1000 different bank statements to 100 different high-speed line printers without even breaking a sweat.
Hope this helps. Any other fun definitions to add?
The CB App. What's your 20?
Mainframes:
General purpose machine.
Tons of IO bandwidth.
Substantial processing power.
Highly redundant and fault tolerant.
Flexible and scalable architecture.
Their OSes are very secure and support thousands of users.
Supercomputer:
Specialized scientific machine.
Tons of memory and/or interprocessor bandwidth.
Loads of processing power, especially vectors.
IO speed may not be important.
Redundancy and fault tolerance not as critical as with mainframe.
Architectures tend to change more frequently.
OSes not geared for business use.
Take a good look at the SunRay terminals that Sun is offering. Rather than hack and patch Windows, they simply made a few modifications to X, most of the client-server tech was already in place.
Thin Client Windows has been a nightmare, and it's only getting worse. One of the original incarnations, WinDD hosted by a Tektronix-modified version of Windows NT 3.5, wasn't so bad... Windows was simpler back then. But all of the "ease of use" and "zero administration" crap Microsoft and Citrix have built up since then has made thin client Windows a miserable beast to deal with. I know many administrators who swear a building full of plain PCs and a good Norton Ghost setup is easier to maintain.
The "problem" with mainframes is not so much that they are old, but that most of the applications didn't use relational databases. If the applications used relational databases, then one could much more easily slowly replace COBOL applications with a more pleasant language of implementation in piecemeal.
Mainframes are Turing Complete, so that any software can be made to run on them if the tools are built. Thus, things like limited-length file names can be transitioned to longer names in a way similar to how Windows allowed one to move to long file names. A mainframe could make an ideal web server because of its security and multi-processing capabilities. If this is the case, then why is it not done often?
Companies seem to have trouble doing this because of data sharing issues. They must keep using the old data while the conversion takes place to newer conventions. But this would mean having Java and PHP apps accessing data stored in the likes of IMS (navigational) databases. But this would mean one had to keep using IMS even after the conversion. (There are IMS-to-relational translation techniques, but they are hokey for the most part and it is tough to get decent normalization because of the different philosophies.)
Thus, the "problem" with mainframes is not the hardware, but the database conversion. The live data cannot easily be in two kinds of databases at once.
Table-ized A.I.
If the speed is measured in gigaflops, or it looks fancy and new, it's a supercomputer. If it can interface with teletypes, chain printers, reel to reel tape drives, or punchcard readers, it's a mainframe.
Supercomputers are all about speed. Large size is optional, but it must be able to do at least a billion floating point ops per second.
Mainframes are always huge, and are all about reliability. They run great, because the current ones were designed in the 1970s, and have had nothing but bug fixes since then.
AS400 .NE. mainframe. System 390 .EQ. mainframe.
He never said AS/400 ws a mainframe. He talked about both mainframes and minicomputers. In fact, if you look at his post again, he said his business has 75 AS/400s and 3 mainframes.
And for those that don't already know this: even a big Unix server is still a Microcomputer. Takes more fault tolerance and funky system architecture than what a Sun has to be called a Mini.
Which brings up the question: is an HP/Tandam NonStop Himalaya a mini or a mainframe?
For those who don't know what a PDP-11 is and are afraid to ask, it's a minicomputer (mainframe, mini and micro, a desktop is a micro) from DEC.
PDP-11 FAQ
-- Would it be acceptable to just put my name on my sig?
In English it's neither plural nor singular. Data is a mass noun - like "water" or "air" - you don't count how many of them you have without specifying a container or a measurement of some sort. Just like it is nonsense to say "I have 3 airs here", but you could say "I have 3 bottles of (or litres of, or cubic feet of, or kilograms of...) air here. It's nonsense to say "I have 3 data here." That doesn't mean anything. Now, "3 Bytes of (or pages of, or databases of, or integers of, or strings of, or columns of...) data, now that makes sense. The singular or plural designation goes on the measurement noun, not on the mass noun.
Don't label something "offtopic" unless you know the topic well enough to tell what's on topic.
Before the /360, IBM had the 1400 series (1401, 1410, 1440, 1460, 7010) that was byte addressable, the 702, 705 Models I, II and III and 7080 (one series) and the 1620 models I and II and 1710 (another series). These were all byte (or character if you want to claim bytes have to be 8 bits - which wasn't the standard until later) addressable: that was the common wisdom about how to do machines oriented toward commercial processing. Most of the other computer companies had byte addressable systems, too.
John Roth
Doesn't it make you feel good to know that our freedoms are protected by politicans, lawyers and journalists.
Yes, but don't forget the Google cache.
That said, I've been talking to IBM about Linux on the mainframe recently and while I don't have an actual figure handy, I wouldn't be surprised if the number your source cited were true, and in fact there may be even more movement in the Linux-on-mainframe area than that figure suggests.
IBM is marketing Linux on the mainframe primarily to existing mainframe customers who want to further leverage their investments there. Remember that mainframes tend to be very modular and upgradeable ... you need not replace the thing to see performance gains or new functionality. You can just buy some new parts.
So IBM is selling a version of Linux that will run under zVM, its mainframe virtualization technology, as well as hardware modules that are basically PowerPC G5 units you can add to the base hardware for the explicit purpose of running Linux. (I don't think you necessarily need the add-on modules to run Linux, I just know that they're available.)
This doesn't really have any benefit at all if you're running a compute cluster or any other application where the Linux boxes are running at high utilization all the time. The main purpose for this is consolidation of lightweight servers. Let's say you have a farm of a hundred Linux Web servers that mostly sit around idle, and the heaviest lifting they need to do is to hand off transactions for processing in the database on the zSeries mainframe. IBM suggests that you instead roll all those servers into virtual machines on the mainframe itself.
Note that we're usually talking about a mainframe that's already in production use, here. You don't need to wipe your mainframe and start over with Linux. You can run Linux instances and z/OS instances at the same time. You gain the following advantages:
- You can now use the same staff to maintain those Linux "boxes" that you were already using to maintain the mainframe
- VM makes it pretty easy to provision new virtual servers as needed, and keep their configurations consistent
- You get the benefit of increased I/O -- the Linux instances think they're communicating over TCP/IP to some remote database, but really all the I/O happens using the in-memory channels on the mainframe
Are these advantages compelling enough to make a lot of companies run out and spend the money on a mainframe? Probably not, especially with today's economy so focused on short-term gains instead of long-term ROI. But if you've already spent the money it could be pretty attractive.From my understanding, IBM doesn't really have a whole horde of customers yet, but I bet a lot of mainframe customers are evaluating the option.
More information on this, as well as mainframe topics in general, in last week's InfoWorld: here, here, and the full PDF special report on mainframes here.
Breakfast served all day!
The IBM mainframe architecture was well designed and well implemented...
Indeed. In fact, there are many now-old innovations in it that "newer" technologies still don't completely get, like a true virtual machine architecture. Such capabilities, relatively trivial to add if designed into the hardware from the beginning, are painful and inefficient to emulate if not.
Then again, I don't miss hex-based floating-point!
Of course they have made some improvements over the years...
One of the more amazing at the time that I saw was a workable subset implemented in the '80s on a PC card. It turned an early IBM PC into a desktop mainframe for some applications.
"God, yes. You hardly ever see [MRAM] anymore, and [Millipede] are being phased out right and left."
DATA
data ( P ) Pronunciation Key (dt, dt, dat)
pl.n. (used with a sing. or pl. verb)
Factual information, especially information organized for analysis or used to reason or make decisions.
Computer Science. Numerical or other information represented in a form suitable for processing by computer.
Values derived from scientific experiments.
Plural of datum.
[Latin, pl. of datum. See datum.]
Usage Note: The word data is the plural of Latin datum, "something given," but it is not always treated as a plural noun in English. The plural usage is still common, as this headline from the New York Times attests: "Data Are Elusive on the Homeless." Sometimes scientists think of data as plural, as in These data do not support the conclusions. But more often scientists and researchers think of data as a singular mass entity like information, and most people now follow this in general usage. Sixty percent of the Usage Panel accepts the use of data with a singular verb and pronoun in the sentence Once the data is in, we can begin to analyze it. A still larger number, 77 percent, accepts the sentence We have very little data on the efficacy of such programs, where the quantifier very little, which is not used with similar plural nouns such as facts and results, implies that data here is indeed singular.
I don't know about that. There's quite a lot of big old mainframes running weather tracking and analysis software, for example. The USGS, I believe, has a number of mainframes that collect several terabytes of weather data per day... and they keep all of it. Forever.
There are quite a lot of such obscure applications out there (especially in the earth and space sciences) that gather titanic amounts of data. Even if Google cached all five billion web pages, and each web page was a megabyte (which is probably way overestimating), that's 10 petabytes of data (5 petabytes each for the pages and the cache). Now think about the thousands of mass-data-collecting computers there are out there, that (between them) archive more data than that every day.
"Destroy science and religion. Science would re-emerge exactly the same; but not religion." - Penn Jillette, paraphrased
> Think of it as a gigantic array of pointers
But it's not. They keep a cache copy of almost all the HTML and other text, and thumbnails of all the graphics.
Chris Mattern
The S/390 port of Linux will run natively in a zSeries logical partition (or LPAR -- a builtin virtual machine facility). You can define between 15-30 LPARs in your complex, regardless of the number of physical processors that are present. I run twin z/OS images and one SuSE Linux server on my single-processor system, without the benefit of z/VM.
There is no "PowerPC G5 unit", though you may be referring to a so-called "IFL" processor. This is a CPU that is only licensed to run z/VM or Linux. Since z/OS is charged on a per-CPU basis, you can save on software costs if you purchase additional IFLs instead of full-function processors. (This is only a licensing trick; both types of processor still run S/390 code.)
US Democracy:The best person for the job (among These pre-selected choices...)
CICS is a neat idea that deserves a new look. It's a "transaction processing OS". Think of it as an OS whose purpose in life is to run CGI programs efficiently. In its simplest form, each incoming transaction starts up a new program which reads the transaaction, connects to the database, processes the transaction, and exits, typically within a fraction of a second. The operating system is optimized for starting and running those transactions.
CGI processing under Linux is inefficient, and hacks like mod_perl are needed so that a new process isn't created for each transaction. One could do better. Transaction programs under CICS are started, run up to the point that they need input, and stopped. When a transaction comes in, a copy of the stopped transaction program is forked off, used to run the transaction, and terminated. So there's no way for data to leak between transactions. All transaction programs run in a jail, allowed to talk only to the database and to reply to their incoming message.
With better OS support for transactions, web servers could have a cleaner, faster interface for their transactions.
Even smaller IBM servers (like AS/400) have built in UPS. So unplugging for a short period of time won't hurt it.
Besides you can find even PC servers with redundant, hot pluggable power supplies. In mainframe every piece of hardware is hotpluggable including processors.