I recently was talking to a Chief Architect of Fujitsu about all the latest non-mainframe technology including grids, virtualisation, blades and so on. He said we're just trying to get back to where we were with mainframes 20 years ago, and we're not there yet. Mainframes make so much more sense than the hundreds of smaller servers we see with all their installation, integration and huge heat/space problems. IBM didn't need to do anything. Except of course one thing - stop charging ten times as much as anyone else.

Quick math: $1,000,000 / 1500 = $666.66. Kind of cheap for an x86 server, eh? I don't think that I've seen anything that cheap and can be rack mounted. Of course that's what Google probably pays for their boxes, but they just buy what fell off the back of a truck and chuck them on a bench.

Reboot!

the CG textured one.

Demise of the mainframe?

But that's proper computing!

Z EOD <enter>

I'd like to see that!

Think how much space they'd take too. Plus power, cables, aircon.

A key feature of the modern mainframe _is_ virtualisation. I asked recently, and some folk are running upwards of 1000 Linux guests on their mainframes.

They're good at I/O (Gbytes/sec), but don't expect to see one mentioned at top500.org any time soon.

Can we expect a reghardware review then?

That'll be the number of the really big beast then.

I did like "sweeping the floor of the data centre", I look forward to iRobot producing a 64-way Roomba to compete.

"There's no mention of virtualisation ... which many forecast ..."

Uh ... a technology which IBM first offered as a commercial product in the early 1970s on the mainframe (my first interesting job was systems programming for VM/370 R1). And which IBM is now pushing it quite vigorously on both mainframe and high-end UNIX servers.

As Deep Though pointed out, the only thing that threatens the mainframe - today as in the old days - is IBM's addiction to extreme markups.

That's Apple's weakness too, come to think of it.

R.

A decade ago I told the boss "That mainframe will be obsolete" implying that it would be replaced by servers. Well, I guess I was right -- that one will be obsolete and be replaced, but by another smaller, faster mainframe. Go figure...

If a few hundred x86 or Power systems tied together can make top500, think of what tying together 4 or 8 z10s as a cluster would do.

Cray would be proud for IBM showing that his "2 oxen vs. 1024 chicken" analogy was correct. Of course, you'd expect that with technology upgrade rates and Moore's law, we'd have 1024 oxen boxes by now; but most IT managers went for the Intel chickens anyway.

I only wish that I had the money to buy one ... though that "mid-range 100k" mainframe is only twice the starting price for a Sun 25K ... if I had that money, I'd sure go for the big iron! Though I'd buy anything, as long as it's not x86 or Windows based.

BOTTOM&50 (ISPF's version of the "nodding duck" so my session won't time out ;)

As Sophisticated and as Complex as the Industry may configure itself, it will still only need Plain Text Input to Program the Programs that they crunch. Simple Dealogue between Virtual Global Operating Devices...... Planning the Future.

Daniel B.: Have you cracked open the covers on the new mainframes? I think you'll find that rather than one big oxen, it's a big chicken coop with 256 chickens inside.

Re another comment:

"Mainframes make so much more sense than the hundreds of smaller servers we see with all their installation, integration and huge heat/space problems."

It seems to me that the sloshing of computing power back to centralized mainframes will favor the consolidation of governmental power.

It's not too hard to imagine a world in which PCs will be confiscated (or simply no longer be sold) and replaced by state-approved thin clients in an effort to remove "weapons-grade" encryption capability from the grubby hands of hoi polloi.

It's my considered opinion that most of us IT folks have been busy building the gilded cage that everyone will eventually inhabit. Regardless of our personal scruples, most of us need the work and a job's a job.

There was never any need for the Boys from Brazil. The will to power (like the poor) will always be with us though the balance between it and the traditionally irrepressible anarchy of human behavior is swinging towards fascism, notably facilitated by the technology we develop and implement.

* Traceability of physical location by commercial activity (credit and debit card purchases)?

Check.

* Traceability of physical location by voice communications?

Check.

* Traceability of physical location by Internet communications?

Check.

* Transparency and data mining of free email?

Check.

* Voluntary and even cheery disclosure of socal network connectivity over and above what can be deduced from email?

Check.

* Concentration of the media facilitating cultural and historical amnesia?

Check.

Way to go!

@amanfrommars

Indeed, plain text input is the hallmark of IBM mainframes. Why would anyone need anything beyond:

//SYSPRINT DD SYSOUT=D

//SYSIN DD DUMMY

//SYSUT2 DD SYSOUT=(S,,CHKS),DCB=(RECFM=FBA),COPIES=1,

// DEST=(BAR1TN06,XEROX)

I'm going for a check-up you're starting to make sense!

Erm, not sure of this is the right place to ask, but has anybody actually seen a mainframe fall over, aka easter egg, aka blue screen of death, aka green stripe of death? As opposed to being brought down in orderly manner by a techie, that is. Was that in person, or was an urban myth type story from a man in the pub?

microsoft and intel's greatest trick was to convince us to run only one app per box (or group of boxes)

Rackham, you just gave me an IBM flashback. *shudder*

IBM has the worst operating systems imaginable, but at least someone documented their woes in the book "The Mythical Man Month", a classic that is required reading for most CS students.

The only danger I worry about with people pushing mainframes is that many of them also push the idea of Thin Clients. That is where you really start to disempower the user and kill off the ability to create new software outside of IBM or SUN or Oracle.

A $1 million mainframe? Who are they kidding? IBM sells P-series for that much. A mainframe rarely goes out the door for under $3 mil.

Once you buy your $1 million mainframe, you need another $2 million of upgrades ot get it to where it needs to be. Then a couple of million for the software to run it, another mil for CA management software, then there's the cost of mainframe-specific storage, mainframe-specific backup hardware and software and mainframe-specific office of grey haired old geezers to run the box. And that's after IBM Global Services has gained their share of the consulting pie to get the gear to the point where you can run it.

Everything that it touches becomes burdoned with the cost of a non standards-based approach and worse still, you now have vendor lock=in: IBM has you by the short and curlies.

This is why many of those customers that have mainframes are trying to get off the box to save multiple millions per year. If you're not locked in with home-grown software, it's compelling, you're not locked into a single vendor approach.

[Penguin, because VMware Hypervisor is written in Linux.]

I think you will find it is 64 larger birds (Emu's maybe?) under the covers. It says so in the article.

I worked with mainframes, they fail just like any other large server. Their is no unique or differentiating hardware or software in them anymore, with the E10K, Sun overtook the mainframe in scaleability about 10 years ago. A 64 way mainframe is a medium sized Unix enterprise server. Many large Unix servers have over 100 cores. Amdahl and Hitachi built larger and often better mainframes than IBM. They saw that other companies would overtake the mainframe and got out of the business.

The mystery concerning mainframes comes from the fact that you cannot find anyone to explain them except IBM. Thus not many people can refute IBM claims.

MIPS = Meaningless Indicator Of Performance. This is not a useful measurement of performance or metric.

The term mainframe is missused enormously, any large UNIX server is a mainframe.

If you run a mainframe you are totally locked into IBM, you cannot have an open bid to upgrade it. There is only one bidder IBM, you cannot go anywhere else for the s/w, IBM supply z/OS. Most TCO/ROI calculations will be done by IBM Global Services, who have more people than you company so you will not have a chance or the resources to analyse their cost model or question every detail.

For any new application being installed on the mainframe calculate the cost of exit. You have nowhere to go. With Unix mainframes you have several companies that will supply you a platform and you can truely compare costs & ROI. And move to another supplier if you are not satisfied with your Unix Server.

Mainframes are good, but so are many large Unix servers/mainframes.

Caveat Emptor, buyer beware.

@ MD Rackham

Please don't forget the invese logic of the COND=(0,GT) clause. Although personally I always prefer it to the new fangled //IF things!

Are they comparing this mainframe against 1500 server in terms of power consumption?

ho yes. two main types of BSOD; green dot as the power to the console goes and the frozen screen. Also had a IBM VM Hypeavisior just do ramdom hex dump/count to the MCC on a load.

The King is dead

Long live the Mainframe.

I pick Paris as REXX was a bit of a dog too.

No, mainframes are not best for everything. But, if you need real reliability, performance, security and easy to operate at the same time nothing beats z/OS (aka MVS) in mainframe. And yes, even a mainframe can go down. There has been some OS problems which made it unusable, of course, no electricity - no computer, an operator error and a catastrophic hardware failure (which usually leaves it running but with so much reduced power that it is basically unusable - seen one and have rumors of other.) About name mainframe, it can be argued but when it comes to raw information, not computational, power I don't think that there are other (today) than a real z/OS based system from IBM. It is unfortunate that Hitachi, Fujitsu, Amdahl, etc are not any more competing, it was an interesting time. About real ROI - if running hundreds or even thousands of Linux (business or development) servers in a mainframe there is no question that it is much less expensive in every way than separate servers. And it is a myth that mainframes are more difficult to manage than Unix, Windows, etc. When an installation has a good systems programmer taking care of the systems I still can teach the administrators (used to be called operators,heh) in a couple of weeks and after six months expertise they will be fully fluent. And development is same as in any other platform except the developers don't have to care make files(?), libraries, versions, etc - all that was taken care (in a good installation) already in 70's. JCL ? what is that, in no organized mainframe shop using source/configuration management to manage the environment has required the developers, etc to know except basics - mostly how to fill your credentials, run conditions, output requirements and maybe accounting information, the system takes care of rest and has done that 30 years. Now - add VM to that. Each group and even a person has own mainframe or maybe own Linux to develop, to test, to QA, etc just by starting it (once again - if the installation is even a little designed and organized right!) Mainframes just got a bad name for a while because of the same reasons other systems have/will get a bad name - business didn't understand how to use them and IBM has to take some blame on that, they moved to marketing instead of their at one time excellent education and training.

"There's no mention of virtualisation ... which many forecast.

As Deep Though pointed out, the only thing that threatens the mainframe - today as in the old days - is IBM's addiction to extreme markups.".... Virtual Threat? By Rippy Posted Tuesday 26th February 2008 15:15 GMT

Rippy,

IT is only a Virtual Threat .... [and whenever the Reality on the Ground is Provided/Dictated to (by Virtualised Services*), it is Virtually Real, and therefore a Quantum Communications** Command and Control Environment which QuITe Literally TelePorts one to a Stage Managed Perception which Units accept as the Status Quo Temporal Position at Present/just always only a Global Snapshot of Life/Shared Human Perceptions and Ideas, presented by Media and Communications ..... aka ICT .... and misleadingly called Reality, whenever in actual Fact, it is a Fiction created for AI Virtual Reality. Of course, that Fact is not freely Shared given that the Reality which is created by Media Moguls/ITControllers/Great Games Players is so Bad ..... with them Providing so much Destruction and Divisive Play/Inequity rather than Collective Constructionism and Social Instruction in the Selfless Provision of Virtualised Services.

One would wish to think that that Failing/Omission was due entirely to their Ignorance of the Readily Available Facility rather than a Selfish Arrogance in ITs Abuse ..... as only the Former would/could Rightly Excuse the Latter Surely. Anything else would indicate a Prior Knowledge of events which they were complicit in supporting via their reporting. And with Imagination available to supply any Beta Program for Sharing/BroadBandCasting/Virtualisation, QuITe why we would continue to accept Sub-Prime Perverted Destructive Content/Output whenever all that IT takes is the Simplest of Conscious Decisions to Provide AI Beta Imagination from Beta Imaginative Sources, if Current Present Source Providers are not Fit for ITs Future Purposes.

And you may like to Consider and Ponder, for more than just a Moment, that the present Status Quo Position/Establishment Controls would hardly be likely to be Immune or Free from ....... Encouragement to Change to ITs Beta AI Virtualisation Control Protocols.] ... if it is Perceived as an Attack rather than Welcomed as a SurReal Opportunity.

The Choice made by the System will decide whether it pays a Heavy Cost with its Collapse or IT Creates AI Wealth you need only Imagine. And the Interface which will Best and Transparently reflect the Choice made, will be the Markets in Love and Banking on Printed Paper Money rather than Generating Real Wealth from Virtually Nothing, which is hardly an Alien Concept if Abused so widely in Ponzi Fractional Reserve Circles/Cabals for Personal Profit at Public Expense.

The Old System is Dead [in case you didn't notice] Long Live the NeuReal System.

Comprendez, Kemo Sabe?

* TP2TP ..... Transparent Peer to Transparent Peer

** A Communications State capable of Yin Yang Resolution with the Provision of an Alternate Reality Game .

Register Interest and/or Disbelief here .... but do not Deny yourself the Reality for that would be QuITe Delusional in the Face of the Colossal Facts just Shared.

Whats with the nostalgic moments of using Mainframes, those machines are old.. outdated, have really bad UI and suck when it comes to usability!. Think its about time you guys retire and read mainframe stories to your grandchildren.

We need a fresh breed of IT people.. GO HOME!

"has anybody actually seen a mainframe fall over, aka easter egg, aka blue screen of death, aka green stripe of death? "

Oh hells yes. Mainframes are not special magical computers, they work just like anything else, just on a different scale. Whatever else you can say about your i386 Windows / RedHat server from Dell, it probably didn't crash and burn for 24 hours because a 50pence rubber seal broke on it's water cooling system's pump.

Scene: 1993 - A Classroom to learn about the New Apple computers that the company is putting on people's desks. Myself and a colleague, 2 MVS System Programmers are the only students.

Instructor: (Takes us through the basics, how to switch it on etc). "And here is the reset switch in case it locks up on you".

Locks up??!? We stare at each other.......

Yes and No.

In a well set up SYSPLEX cluster several machines can crash - even a complete site can go -- but as far as the end user is concerned the system is still there. -- Sun E10K etc, cannot come close to this level of availability.

JCL -- its different but certainly no worse tha bourne shell or DOS .bat.

UI -- matter of taste really -- ISPF is probably the best character terminal based environment there ever was and its editor is one of the best ever (equal first with vi). If you really insist you can run X-Windows with emacs on.

Standards compliance -- the Unix System Services feature is the most POSIX compliant unix environment I have eveer worked in. Anything you can complile and run on a unix box will probably complie and run here -- the only caveat is that if you pick "EBCDIC" as the character set for your shell things like "c == x30 " wont work as expected.

I predict that someone will be predicting the death of the mainframe in 2028.

We were running AS (4GL), DB2, TSO, Real-time refinery control system with SRTOS (real-time extension to MVS developed for the moon launch) on MVS second level on VM/XA on a 16Mb (Mb not Gb) single cpu 4381, when the real-time statistical analysis process we were developing in the real-time region overwrote part of Nucleus (kernel).. what a thrashed MVS did can only be described as remarkable..

With a corrupted kernel and 100 times physical memory in page files, MVS detected the corruption (through Nucleus buffer ckecksums), check-pointed and shutdown surviving services. Watching the system console as the OS fought through dead dependencies (JES2 was gone) was like a day at the races.. but the best was yet to come.. after restart.. going back to the recovered unsaved PL/1 source file with no code lost.

Not the same feeling seeing a Sun E10k crash every domain when “hot swap” memory failed or seeing a DataSynapse Broker thrash with 5 blades working in a 500 blade grid

We've been hearing about the imminent demise of the mainframe for fifteen years or so now. Big iron is still around, and we're still hearing about how its extinction is just around the corner.

Client / server architecture was supposedly going to kill them off. Then it was clustering that was going to make them a distant memory. And still, we're right back where we were in the early nineties: Mainframes are good for some applications, poorly suited for others. Yes, they centralize resources, and can lead to a single-threaded world of fail if they go Tango Uniform on you. Then again they tend to be exceedingly stable, consume less power than a room full of servers, &c. &c. yaddayaddafishsticks.

Vendor lock-in? Yes. Undoubtedly. But you tend to have that with Micro$oft and other vendors, too. It's not strictly an IBM problem.

Fifteen years from now, we'll be hearing yet again how NEW! SHINY! PLATFORM! is plotting the mainframe's imminent demise. And IBM will be introducing yet another line of oxen / chicken coops / emu flocks / jet-powered pterodactyls.

Paris, as there's no 'Tempest in a teacup' icon.

"The term mainframe is missused enormously"

Ahem! Does that mean that the term "PCs" was mistered, mastered, or mustered?

Drizabone and Akubra, thanks.

We all know that biggest isn't always best, its almost without fail IBM prove it at every point...

Unisys mainframes can be run with just a handful of staff. Trying pulling that stunt with anything IBM, and you'll find you need a whole army of freaks you've never seen near polite society to maintain it.

Paris, because she's been f****d as many times as anyone with an IBM contract!