UK Government Department Still Runs VME Operating System Installed In 1974

Qedward writes: The UK government's Department for Work and Pensions is on the hunt for a new £135,000-a-year CTO, with part of their annual budget of £1 billion and responsibility for DWP's "digital transformation" to oversee the migration of the department's legacy systems which are still run on Fujitsu mainframes using the VME operating system installed in 1974.

Comments

Modern Technology

[Galaga88]

How many modern systems can anybody imagine still working and apparently doing what we need them to 40 years from now?

Re:Modern Technology

[TheRaven64]

I am always amazed how buildings constructed thousands, or even hundreds, of years ago are still standing although often in a state of disrepair due to neglect

That's what's known as survivor bias. The only examples you see of thousand-year-old buildings are the ones that didn't fall down. The ones that collapsed within a decade are long forgotten.

A modern-day castle might survive a century whereas the castles throughout Europe remain or at least remnants of their existence survive to this day

And yet, in the village where I grew up, and near countless other villages in Britain, there was a hill with a raised mound on top, which was the only remaining evidence of the castle that stood there 900 years ago.

Re:Modern Technology

[TheGratefulNet]

http://www.eevblog.com/forum/t...

nice old classic tek test gear. highly in demand by collectors and those who appreciate good old fashioned engineering and build quality. the last of the 'repairable' tek scopes, pretty much (and even this is borderline repairable, with many custom chips).

still, a few new caps, a new battery backed nvram module and you have another 20 or 30 yrs left on this scope.

search that same forum for other old test gear (power designs (brand) power supplies are also built like tanks and run forever. I have 4 of them at home in my lab and they date from the mid 50's to early 60's. still hold their precision and would cost $5k to $10k today if you could even buy them.

I have audio gear that I personally built in the 70's and 80's that still runs fine (hafler amps, etc).

today, its hard to find things built to last, but it USED to be the norm "before your mother was born", so to speak.

Re:Modern Technology

[gstoddart]

Give me what that system cost in 1974 inflation adjusted dollars and I'll be happy to flip out a modern system every year.

Sorry, I'm calling complete and utter bullshit.

I've worked on enough legacy systems to know they didn't start off with some astronomical budget. They built it based on a set of requirements, coded it in house, and then it gradually expanded over many years of service.

Mainframe applications aren't sexy or glamorous, they're built on relatively simple interfaces, and slowly expand in scope over time.

They keep running because eventually they're woven into fabric of every other business process you have until they become something you can't trivially get rid of ... because every other damned thing relies on it even if it isn't obvious to the user. You end up having to replace everything

My experience with migrating from legacy apps says you'd churn out a half asses solution, which isn't compatible with the existing stuff, and which can't be made so, and which would eventually be abandoned as untenable.

You'd produce some solution which might be good if it didn't depend on throwing away every other system which touched this.

The vast majority of people who claim they could produce a functional replacement for legacy software in a short period of time have never been involved in that kind of process.

If it was easy, they'd have replaced it by now.

The problem with looking for a "track record of transitioning a large enterprise from ageing mainframe technologies to next generation web, social, mobile cloud, Big Data and deep learning technologies" is that it's a set of requirements written by idiots who don't want to replace the system, they want something completely different which will involve re-tooling everything else that touches this existing system.

Put your money where your mouth is, apply for the damned job.

Re:Modern Technology

[mrbester]

Because, as a student, if you hit it and it breaks you did something dumb and reduce the number of units for the class to use. However, as an instructor, if he hits it and it breaks, it was due for replacement.

Re:Modern Technology

[csnydermvpsoft]

What basis do you have for the claim that we "couldn't even begin to recreate" those structures? There are certainly some ancient structures for which we haven't figured out how they were constructed with the technology available at the time, but nothing that we couldn't reproduce with today's technology.

The sticking point isn't technology—it's economics. A large portion of recent development has been around cost-effectiveness. This is why we're able to have so many more material possessions, even in the face of stagnant wages (for most classes). Of course, many (including myself) would argue that we've gone too far in this direction at the expense of durability, but that's an economic choice we've made. Look hard enough, and you can find any product that meets your durability specifications—if you're willing to pay the higher price.

That being said, I do agree with the sentiment that there is more than survivor bias at work. My house was built in 1916, and has an unusually open floor plan for its age. Lacking CAD, the builders accomplished this by massively overbuilding—the floor joists (with are already quite thick) rest on beams comprised of four 2x10's laminated together. Despite its age, this house feels more solid than just about any other wood-framed building I've been in. I have no doubt that if it were placed alongside a newly-constructed house and both left to nature, that the 99-year-old house would remain intact longer.

Re:Modern Technology

[gstoddart]

Well, the problem happens when some technology evangelist or manager who doesn't know a damned thing about the existing system claims it's easy to migrate it to modern tools.

And neither the customer, nor the guy saying it's easy, has the barest clue about just how many other things depend on that system, and nobody can fully enumerate the functionality and corner cases.

And then you end up trying to shoe-horn a purpose built piece of software which has ran fine for decades into a modern paradigm, and realize you are failing utterly.

Because the modern tools usually simply can't accommodate all of the rules and logic in that system. They can't be cajoled into having enough flexibility, or simply can't do the same task.

People consistently underestimate just how well these systems do their job, and just how many little corner cases and integration points have been woven into them over the years. The platform is no longer elegant, or easy to explain, but it just keeps working. But dozens of other things rely on it, and if you change the underlying thing you rebuild everything else.

I've been on several projects trying to replace stuff built in the 60's and 70's -- and I wouldn't go near another one without very loudly saying how much risk is involved. Hell, even a system which has been around only since the 90s might be non-trivial to migrate away from -- precisely because in the 90s people were still building much more purpose-specific software.

It's a catch 22 ... they get increasingly difficult to maintain, but they sometimes are impossible to replace.

As I said, if it was easy to replace these systems, it would have been done already. Discovering just how difficult this can be has been the downfall of many a naive person who claims it's an easy thing to do.

Re:Modern Technology

[jeremyp]

Without knowing *why* the castle is gone, I have no idea what your point is.

Usually, if it was a stone castle, the building materials were robbed out to make new dwellings. The reason that people could do that is because the owners abandoned them as being shit places to live.

With buildings, as with other man made items, technology moves on. Generally speaking, a house built now will be more comfortable, easier to heat and more suited to modern life styles than a house built 50 years ago. Who cares if the old ones fall down? If you are going to knock it down and replace it with something better, money spent making it last a millennium is wasted money.

old != bad

[AndroSyn]

My money is on this VME system being around for another 20 years while the mess of Java and Oracle(you know they're going to use Oracle). It'll be overpriced, late and won't actually work.

Just because something is old, doesn't mean it needs replaced. In short, why not just upgrade the mainframe?

Re:old != bad

[Shinobi]

Nono, like other big IT projects in the UK, it will be using "the very latest in Agile know-how", and cost 3 times as much as any clusterfuck that involves Oracle, take 50% longer, and spread 300% more blame on "old fossiles"....

Disclaimer: Had to interface with a EU project under UK IT auspices last year.... Painful....

It is called good coding.

[www.sorehands.com]

Many people are shocked that computers/systems for 20 years still run, but is says a few things:

1. That people are used to crap code that can't keep running.
2. That people are used to crap products that can't last for more than a couple of years.

If it ain't broke, why fix it? They sent man to the moon on less CPU horsepower than my Nexus 6. Voyager has been running for more than 35 years in the harshness of space.

Re:It is called good coding.

[Todd Knarr]

They have. But they didn't do it overnight, they did it small bits at a time and those 40-year-old systems were patched or updated and debugged with each change. The result is a twisted nightmare of code that works but nobody really understands why and how anymore. And the documentation on the requirements changes is woefully incomplete because much of it's been lost over the years (or was never created because it was an emergency change at the last minute and everybody knew what the change was supposed to be, and afterwards there were too many new projects to allow going back and documenting things properly) or inaccurate because of changes during implementation that weren't reflected in updated documentation. As long as you just have to make minor changes to the system, you can keep maintaining the old code without too much trouble. Your programmers hate it, but they can make things work. Recreating the functionality, OTOH, is an almost impossible task due to the nigh-impossibility of writing a complete set of requirements and specifications. Usually the final fatal blow is that management doesn't grasp just how big the problem really is, they mistakenly believe all this stuff is documented clearly somewhere and it's just a matter of implementing it.

Orange Leos?

[sysjkb]

I wonder if they are running "orange Leos"? Here's a post from alt.folklore.computers in 1998. Terribly impressive. I'm not sure his age estimate is necessarily accurate, though: the final incarnation of the Leo ceased to be manufactured in the latter half of the 60s, so it may be a bit younger.

From: Deryk Barker (dbarker@camosun.bc.nospam.ca)
Subject: Re: Multics
Newsgroups: alt.folklore.computers, alt.os.multics
Date: 1998/11/09

[...]

When my wife was working for Honeywell, in the 1980s, one of the
customers she had dealings with was British Telecom.

BT, at one location, had what they called the "orange Leos".

Now, for those who don't know this, the LEO was the world's first-ever
commercially-oriented machine (1951). Even more amazingly, the Lyons
Electronic Office was designed and built by the J Lyons company,
best-known as manufacturers of cakes and for their nationwide chain of
corner tea shops.

Anyway, an "orange Leo" was an ICL 2900 mainframe (they came in orange
cabinets), emulating an ICL 1900 mainframe, emulating a GEC System 4
mainframe emulating a LEO.

30+ year old executable code over 3 architecture changes....

New Platform

[Dr. Evil]

Port it to minecraft. There seems to be some good 1970's CS work happening there.

It is broken (probably)

[sjbe]

Many people are shocked that computers/systems for 20 years still run,

Only people who don't know much. It's not shocking that such a thing would happen or that hardware can be made that robust. What IS shocking is that people put systems in place without any thought whatsoever to what people might want to do 20 years later. Seriously do you REALLY think it will be efficient or practical without problems for you to use the PC you are reading this on today in 20 years? Why would it be any different for a business or government?

If it ain't broke, why fix it?

Because it probably IS broken in a multitude of ways. Just because it can get a specific job done doesn't mean it does so efficiently or without problems. I've driven a lot of beater automobiles over the years and while they usually got me from point A to point B they were unquestionably broken if a number of ways. I have PCs that are 10-15 years old here in my company doing specific jobs and they definitely have problems. Yes we still get some productive work out of them but that doesn't mean I shouldn't think about replacing them when I can.

They sent man to the moon on less CPU horsepower than my Nexus 6.

Because that is all they had at the time. Nobody would even dream of doing that way today because we have better options now. Why limit yourself to yesterday's technology if you have a choice?

Voyager has been running for more than 35 years in the harshness of space.

Which is relevant how? You're comparing a spacecraft that human eyes will never see again with a earthbound computer system that we can modify or replace any time we want.

Extreme example here, but...

[ErichTheRed]

Not all legacy stuff is bad. Not all legacy stuff should be kept around to the point where you can't find people to run it, however,

I've had experience working in die-hard IBM mainframe shops as well as places that used the HP MPE operating system on the HP 3000 minicomputer system. In the 3000 case, the customer was relying on a service provider that was providing an application that was way way way out of date but still worked. All the IBM places I've ever worked have been slowly "modernizing" their application stack, but in most cases, the core transaction processing has remained on the mainframe because that was the best solution. It's extremely rare these days to see an end user facing green screen application, but they do exist as well. (Yes, I work in "boring" old school industry sectors, very few web-framework-du-jour hipsters here, but we're also not old farts.)

The problem I've seen is that vendors love the fact that customers are locked in and will do nothing to encourage them to get off. Most ancient mainframe code can run virtually unmodified on newer hardware, and that backwards compatibility is a big selling point. It allows IBM to go in, swap out your entire hardware platform at $x million, and keep billing you by the MIPS without changing any code.

But...the reverse problem is that "mainframe migration" projects often end up becoming case studies of how Big Consulting Company X was paid hundreds of millions to not deliver a working system. I believe I read about DWP's "Universal Credit" project that has Accenture, IBM or Oracle written all over it. These kinds of projects usually try to port all the business logic and transaction processing to some horrible-to-maintain J2EE monstrosity backed by an Oracle database. They usually fail because (a) no one correctly estimates the work required to pull all that business logic out of 30+ years of cruft, and (b) the consulting companies replace their star team (that travels with the sales force) with new grads in India (who do the actual work.) I've seen this cycle over and over again, and am still amazed that CIOs aren't wary of consultants.