Feeds

back to article Nuke plants to rely on PDP-11 code UNTIL 2050!

The venerable PDP-11 minicomputer is still spry to this day, powering GE nuclear power-plant robots - and will do so for another 37 years. That's right: PDP-11 assembler coders are hard to find, but the nuclear industry is planning on keeping the 16-bit machines ticking over until 2050 – long enough for a couple of generations …

COMMENTS

This topic is closed for new posts.

Page:

Silver badge

It just costs money

These types of deal aren't that unusual. If it ain't broken you don't fix it.

Come 2020 I'll bet there will still be sites getting VMS support.

Just reach for your cheque book.

Government contracts often have very long support plans built into them.

I remember being asked to provide HP-UX internals training in the run up to Y2K for a release than had been phased out in the 80s, because there was still a customer who was on contract.

He who pays the piper calls the tune.

7
0

Re: It just costs money

"If it ain't broken you don't fix it."

Bingo. It's the same reason why the computers and processors on deep space probes are frigging ancient by today's standards. The technology may be old and slow but it is *reliable*. Reliable counts for a lot when you're a million miles from Earth or when you're nuts deep in nuclear crap.

6
0
Silver badge

Re: It just costs money

I would imagine that if these things are bing used to control robotic systems in nuclear power plants, those are the systems which are on the 'hot' side of containment, such as the fuel loading robots and things like that. You could go about replacing the systems that control these, but you'd need to build a robot to do it, because radiation poisoning is apparently quite bad for even field service engineers.

So, you either have a choice of leaving the current system in place until the end of life of the plant, and trying to find some bods who know how to work it, or building a more complex and expensive system to perform a one-off maintenance job inside a sealed reactor at phenomenal expense, and not-insignificant risk.

3
1

Re: It just costs money

> inside a sealed reactor

No, they are surely not using a PDP-11 inside a "sealed" reactor. It will be in a human-friendly control room somewhere.

> fuel loading robots

While those are inside the containment vessel, they are not in any sense inside a "sealed" reactor; the radiation level is not unfriendly when everything is functioning correctly.

2
0
Silver badge

Re: It just costs money

Okay, my example may be a little contrived, I'm not a nuclear engineer after all. It isn't beyond the realms of possibility, however, that these systems are used to controls parts of the reactor that are not accessible once it becomes hot. If not the robots loading fuel, then maybe those transferring spent fuel to cooling ponds. A properly encased rod of U-235 might be safe to handle before it has spent some time inside the core, but certainly not afterwards!

Although the control systems are obviously not going to be inside the reactor, they have to be connected to these parts. Refitting, testing, and properly certifying any changes to these has got to be a hugely expensive logistical challenge, and I can see where this might cost significantly more than employing people who can maintain the existing working control system, even if that involves training them from scratch and paying them well over the going rate.

0
1
Bronze badge

Re: It just costs money

Actually it's because higher frequency cpu is more error prone when cosmic radiation hits it .

0
0
Go

Re: It just costs money

"the same reason why the computers and processors on deep space probes are frigging ancient by today's standards."

and they are difficult to get to for an upgrade...

2
0
Silver badge

Re: It just costs money

They're "ancient" even before launch. We routinely handle new kit with rad-hardened CPUs that would have been considered "state of the art" on desktops in 2001.

If it was possible to economically fly core memory, it would still be done. The environment outside earth's magnetosphere is mindbogglingly hostile to electronics (and biologicals)

2
0

Re: It just costs money

It's not the GHz clock cycle that is the problem. It's the smaller feature size of the transistors that increases the single event upset (SEU) rate. Yes, the two are inter-related, but one could conceivably build multi-core, chip symmetric multiprocessors based on the PDP-11 at today's feature sizes and not have GHz clock cycle times (and still end up with significant SEU rates.)

A couple of years ago, a NASA/JPL scientist pointed out that the alpha particles (helium nuclei) from lead solder were causing interesting issues with current x86_64 I/O pins -- radiation issues on commodity hardware.

0
0
Mushroom

Re: It just costs money

It's not just ordinary reliability as in "well tested", but also robustness against radiation - plenty of that in space. Those 30 year old CPUs running at 1MHz are quite resilient. Could it be that the nuclear industry is also having similar issues?

0
0
Anonymous Coward

Re: It just costs money

Radiation poisoning might just explain some of the field services engineers I've come across...

0
1

Re: It just costs money

I believe the shuttle flew on core memory. Very helpful with the Challenger investigation.

0
0
Bronze badge

Re: It just costs money

That is actually mostly because it takes quite a bit of time to certify equipment for space flight and the extreme conditions like radiation, extreme temperature gradients, and what not.

0
0
FAIL

Re: It just costs money

Well,

Another thought that might help would be to send some torte lawyers in to inspect and research whether it's safe or not for humans to chnage those systems. After all.. we want everything to be contracted and legal - correct?

0
0
Pint

DL0: 167300 g

I *think* that's the correct sequence, but it has been nearly 20 years...

Should restart a sleeping PDP-11/73 via the serial port - useful if it's been shut off and you're working via a modem.

Supporting a call logger on PDP-11 on RSTS 8.x was where I started in 1994, I'd have expected them to be all gone by now although I think the US air traffic guys had some until after 2000!

3
0

Re: DL0: 167300 g (Air traffic control)

UK Air traffic control retired their (last) PDP11 in 2006. It is on display at the National Museum of Computing at Bletchly Park.

0
0
Silver badge

Re: DL0: 167300 g

"DL0: 167300 g

I *think* that's the correct sequence, but it has been nearly 20 years..."

AIEEE! Who gave the command to pull out all the control rods?!!!

Run! Run for your lives!

AAAAAAAAAAAAaaaarrrrgghhurrrrglle...

3
1
Anonymous Coward

Bring out yer dead

Still using XP for our desktop. Got a couple of tandem non-stops floating around as well. Bleeding edge we are ...

2
0
Silver badge
Devil

The beauty of PDP-endian

PDP-endian... One "endian" nobody checks for any more and which will break nearly any network to host/host to network (including telemetry networks) conversion. I love the smell of meltdown early in the morning, it smells like radioactivity...

Yum...

3
0
Bronze badge

Re: The beauty of PDP-endian

You have unlocked a buried memory of getting ICL VME to talk to a PDP-11. That was one of the most satisfying achievements of my career.

3
0
Trollface

Re: The beauty of PDP-endian

Oh, crud. Now I remember ICL VME. I think there are still councils still running VME, although Series 39 has long given way to OpenVME on an x64 host.

To quote the Fujitsu website "2020 is no longer an issue for VME Applications." Not sure if they mean it will work past 2019, or nobody will be using it by then.

Not a zombie OS, exactly. More a sort of... isolated hill tribe of savages that hasn't been seen since the year dot and is barely civilised?

2
0
Bronze badge

Re: The beauty of PDP-endian

Not just byte sequencing but the bit pattern of floating point numbers. Digital had their own format before the IEEE standard became standard. Sun Microsystems produced a library that would enable VAXes to talk to SPARCs, but I forget its name. It was available on Linux X86 systems which I had talking to SPARCs.

2
0

Re: The beauty of PDP-endian

Ah, yes, the classic "nuxi" endian-ness.

0
0
Silver badge

2038 and all that

While not a PDP11 issue, the longevity of computers is a concern - and not just for people who want to flog us replacements. At this rate (maybe even exacerbated by the financial crisis) there could well be numbers of old, 32-bit Unix system running stuff when the 32-bit clock counter rolls over and we find ourselves back in the 1970's. With the popularity of small, cheap Linux embedded systems, some of these could be quite difficult to find - literally.

I just hope none of *them* are running nuclear plants

8
0
Anonymous Coward

Re: 2038 and all that

"there could well be numbers of old, 32-bit Unix system running stuff when the 32-bit clock counter rolls over and we find ourselves back in the 1970's."

Oooh - you mean when we went to the moon, and you could fly supersonic to New York.

65
0
Bronze badge

Re: 2038 and all that

Time to break out the kipper ties, polyester shirts and flared strides. Then kick back with a bottle of Johnny Walker and a fag to watch a mildly racist sit-com while the missus knocks up a Birds Eye Steakhouse Grill and chips.

Oh shit. I've turned into my Dad.

13
0
Thumb Up

Re: 2038 and all that

Hey, can you pick me up a Watney's Party 7 if you're out?

2
0
Bronze badge
Pint

Re: 2038 and all that

We might have been able to go to the moon and fly supersonically, but we didn't have access to twitter back then and couldn't "like" things so surely that evens things out?

(Perhaps not...)

3
1
Silver badge
Joke

<brummie>Kipper tie?</brummie>

Go on then.

Milk, no sugar. Ta.

3
0

Re: 2038 and all that

I remember a couple of years ago Audi advertised their "space-age" aluminium cars.

I already have a "space-age" aluminium car, a 1967 Land Rover.

6
0
Silver badge
Holmes

Re: 2038 and all that

when the 32-bit clock counter rolls over and we find ourselves back in the 1970's.

Depends. Real-time systems don't necessarily need to be aware of the current wall-clock time. "There's an item coming down the assembly line, it needs to be sprayed/welded/have a barcode stuck on/whatever", is not a task that requires the robot to know whether it's 19-jun-2013, 1-jan-1970 or 19-jan-2038.

In the run-up to Y2K, I was told several routers had to be replaced because their firmware was not Y2K-compliant and they couldn't be upgraded to handle the then-current firmware version that was. After a brief check I reported that the longest-running router had still 42 months to go before *it* would hit that particular date which was roughly half a year away on our calender, and that one which had just been rebooted happily lived in late 1993 with no ill effects. From which I deduced that if any of them would ever reach their Y2K-rollover, the worst that would happen was that they wouldn't be routing for a few minutes until they had finished rebooting, just as with any other interruption, the chance of which occurring would be way greater.

No routers were replaced that year.

7
0
Silver badge

Actually probably much less of a problem

I mean integers on most archtectures are cyclic, so you might not have any problem. Let me give you an example with a hypothetic 8 bit time.

Imagine it's 250, 10 seconds later it'll be 260-256=4.

Now imagine you want to have the difference between those times. That's 4-250=-246 which will overflow to 10.

So despite of having multiple integer overflows, time differences will still be OK. The only problem exists when you convert it into some other system.

0
1
Linux

Re: 2038 and all that

And it took an IBM 704 to sing Daisy.

2
0

Re: 2038 and all that

Motorola cable head-end equipment already suffers from this problem. By default, it picks an end date 25 years in the future. Now that we're less than 25 years from the 2038 problem, we're seeing end dates in the late 1960's when we forget to override the default end date.

0
0

if it aint broke....

A well understood, rock solid, stable industrial control system... Controlling a system that has no tolerance for failure..

What's not to like?

19
0
Silver badge

Re: if it aint broke....

A system that has no tolerance for failure has no place in the real world outside of a laboratory.

That's one of the fundamental failings of old-style nuke design.

BTW I believe DLR run a PDP11 in every single train.

2
0
Silver badge

Re: if it aint broke....

> A well understood, rock solid, stable industrial control system...

Indeed. Few people think of upgrading the control system for their washing machine, why would you risk an "upgrade" (are you going to get more features added to your nuclear reaction?) when there is so much to lose?

2
0
Silver badge

Re: if it aint broke....

Do they have spare parts?

I've seen circuit boards that are in use well beyond their supported life. Some of them do not survive power cycling as the capacitors have leaked.

0
0

Re: Do they have spare parts?

You could probably produce an equivalently behaving board for most parts. Most of the board would be empty apart from traces, and all the work would be done by a small FPGA. Bug for bug compatibility would be the biggest challenge.

0
0
Headmaster

Re: if it aint broke....

> BTW I believe DLR run a PDP11 in every single train.

I think it's an 8080 running Forth, from memory.

0
0
Silver badge

Re: Do they have spare parts?

There are a couple of companies making PDP11 addin boards for PCs that emulate a PDP11.

The green roadside boxes running traffic lights were PDP11s until recently

0
0
Anonymous Coward

Re: if it aint broke....

"I've seen circuit boards that are in use well beyond their supported life. Some of them do not survive power cycling as the capacitors have leaked."

SOP for reconditioning/repair services (at least in the contracts we have with a few nuke plants) is to replace *all* electrolytic caps on the board, regardless of condition, with new ones. No part substitutions without engineering approval, of course. From what I've heard, we just get the non-critical (pardon the expression!) stuff that's not involved with the safety of the reactor itself.

Dealt with some stuff that clearly went in when the plants were built. Deal with their spares as they rotate them through the stockroom. Just don't try and get an answer from anyone there during a refueling or maintenance shutdown, seems to be a rather ...active time at the plants.

1
0
Silver badge

Re: if it aint broke....

In some places I've worked I've had capacitors leak PCBs all over my hand when touched (valve kit older than my parents).

I still managed to keep the feckers going. :)

0
0
Silver badge
Facepalm

Re: Do they have spare parts?

There are a couple of companies making PDP11 addin boards for PCs that emulate a PDP11.

What's on the board is a PDP. The PC is just its fancy I/O system.

3
0
Silver badge
Boffin

Re: Do they have spare parts?

Most of the board would be empty apart from traces, and all the work would be done by a small FPGA

DEC already had single-chip PDP CPU's; you could put a large part of what was then still needed as external support circuitry on the die with it today without much effort.

0
0

Re: if it aint broke....

running it, sure it's not a problem

but as the fine article points out, finding people that are able to "well understand" the system to add new features and find obscure bugs (there are only obscure bugs left after 40 years in the field) is a problem

and there's also the problem of longevity of the hardware itself, I don't consider running nuclear power plants using hardware found on a garbage dump safe

0
0
Silver badge

Re: if it aint broke....

"I've seen circuit boards that are in use well beyond their supported life. Some of them do not survive power cycling as the capacitors have leaked.

Not a problem as the original Leyden Jars have a service life of 250 years if you rattle them every now and again with a broom handle.

I hear that broom handle replacement is budgeted for until 2025, after which they have plans to poke the jars with a recycled Dyson.

0
1
Bronze badge
Mushroom

Re: if it aint broke....

More features? How about tweets from the reactor core?

icon? "I'm feeling hot"

0
0
Linux

Re: if it aint broke....

>> I think it's an 8080 running Forth, from memory.

Out of interest, where else would it run from? (The 8080 that is, not the train.)

2
0
Happy

Re: if it aint broke....

sparde parts, 74 logic series is still made/

0
0

Page:

This topic is closed for new posts.