Booted up in 1993, this server still runs – but not for much longer (2017)
computerworld.com
computerworld.com
There are airliners that have been in service for over 50 years.
Yes, this drastically understates the change in computing but the actual technology hasn't changed that much, it just got denser and faster.
The biggest change from then to now is probably the amount of solid state electronics onboard. That's ridden roughly the same curve as other industrial and commercial applications.
IBM mainframes, famously, phone home if they detect a faulty component. Service personnel will be on site, same day, to replace it before you even knew it was there, let alone had the opportunity to ask.
It seems when people think of a server today, they think of a single computer, if not some software server somewhere in the cloud. These subject kinds of servers are far closer to complex systems of separate components, more like a network system, it’s why components of what are really separate networked computers can and need to be replaced.
An airliner is of course also made up of components that are also systems, however at least in my mind, the difference is the actual expected use case. An airliner is not a good comparison because it is never expected to remain in continuous operation, e.g., that at least one engine is always running even when it is being overhauled or repaired, to satisfy a requirement of continuous operation.
What is remarkable is that apparently that VOS thing has never crashed. I wonder if they also do software redundancy under the hood to keep uptime going during reboots. If the CPU is hotswappable, it must have something.
The main computer system in the Space Shuttle is an excellent example of this. It had 5 identical IBM System/4 Pi machines. Three of them ran identical code and handled the same work. The fourth ran a completely different codebase to handle the same work, preventing a bug in the main code from killing the whole system. A fifth computer handled other tasks but could be swapped over to the critical role if needed. You could lose 2/5 computers and still have insurance against a cosmic ray flipping a bit.
Many servers I've owned had multiple years of uptime, and the only reason they'd go down is because they will get decommissioned or because of power outage.
The article says:
> disk drives, power supplies and some other components have been replaced but Hogan estimates that close to 80% of the system is original.
so I am guessing there was no reboots, nor CPU replacements.
Are there any airliners that have flown for 24 years non-stop?
But really didn't the NSA have something that was roughly an airliner in perpetual flight?
it's more impressive to me that the electric fans within the Stratus server still operate than it is that the chips still work, and if we're going with the plane metaphor it's the silicon that puts the thing into the air, not the simple fans.
So really it just boils down to "well, the planes work is harder." as to why it fails, which is of course abstract and unsatisfying; comparing the comparative lifetime workloads of a chip that has shifted trillions of bits versus some jet engines that have moved millions of kilograms around the world.
and then another layer comes and makes it even weirder : the vast majority of hardware around the world that has been EoLd and replaced has been in that situation due to software, not the hardware itself; a paradigm that really doesn't exist in the physical engineering world in the same was as it does CS.
These two have very distinct failure modes. The failure mechanisms for solid state electronics are much more similar to those for static equipment (corrosion, thermal fatigue, creep, ...)
...with a very aggressive maintenance cycle (typical part life is 1000-10000 hrs of service). Most airliners of that vintage resemble the "Ship of Theseus" in real life.
On other side only moving element would be hard drives and only wear element really would be the electrolytic caps.
The were Awesome machines - don't let those filthy hobbitsis...err..tandem users cloud your judgement. Stratus was pure hardware FT, and tandem... wasn't :-P
Vos was pretty nice too - decent macro language - not all the utils you expect with a unix box, but in terms of the core OS and shell it was pretty good machine .. 40 years later I still love those machines and the career they helped get me into :-D
Both were really popular with banks and brokerage houses (back when I worked with stratus at brokerages, HFT was just emerging and down time was umm... frowned upon :-D
VOS was intuitive and did what I needed. Nothing flashy. Just useful. I miss that sometimes. Give me a tool that works; if it also happens to be great, lovely.
I didn't work much with the Stratus hardware, but reading about it was fascinating. I wish I remembered more of it.
A reliable beast - never failed once in the entire time I worked there. Based on PA-RISC if I recall correctly.
Speaking of reliable beasts, I own(ed) a HP PA-RISC workstation that never crashed in six years, and it was even purchased refurbished (second hand, and educational rebate, and still the cost of a car in 1996 - but worth every penny).
I'd add "R.I.P. HP 9000 715/75" but of course she is STILL working now, twenty years later, and she is sitting right next to me (also got a second 715/100XC for spare parts in case on day they are needed).
I think VOS being PL/1 based pretty cool - Stratus (and Intel fab PL/M stuff) was the only time I've ever gotten to work with PL/1...
That said, from what I gather from working with those guys, there are a couple of generations of unix-y things on Stratus:
1) The older FTX Unix was apparently a pretty bog standard SVR4 port from the user perspective, with the machdep stuff and scaffolding to deal with the fault tolerance layered on a VOS based abstraction layer in the kernel. I knew a lot about SVR4 at one time, long ago, so they chatted me up quite a bit about that.
2) Later they added a POSIX layer to VOS and ported a bunch of Unix tools (e.g. GCC). I dunno exactly what the business case is here or why people who want Unix wouldn't just live in the Unix environment. Neither did anyone else I talked to.
3) These days, I understand people are running Linux (RHEL) as well as windows as on top of VOS. All the worlds virtualized these days.
I obviously didn't get to code anything on the Stratus, but agree PL/1 is interesting as a systems language (the other big user being IBM of course). I understand 'Stratus PL/1' is a lot like 'Burroughs Large Systems Algol': they started out there and have kept adding vendor specific extensions to the point it's not really the base language any more. At least IBM gives it new names when what the programmers use drifts far enough from the base language definition (e.g. PL.8).
When the jackpot is fed by all the machine in the state, they can grow much faster.
That's not particularly surprising to be honest. If it's a server operating system, on hardware that doesn't change, there's not really a reason to update the OS unless there's a security problem found
Apparently the latest VOS release was in 2025, which means that somewhere there's a team of developers maintaining it. That's really the thing I find the most interesting, that there exists teams developing operating systems like VOS, strange DOS clones or UnixWare (perhaps less so these days). Must be a strange anonymous life.
This perfectly characterizes most embedded systems work. Competent software development using bespoke or off the shelf components that are not mainstream. Nobody ever realizes they’re using your computer software if you’ve done your job right. One day someone throws your appliance away because it’s obsolete.
https://www.neoteo.com/en/a-commodore-64-stands-the-test-of-...
I was in Fairbanks visiting family, I had some downtime, so I decided to do some “slumming” to break up the monotony and relive some of the good old days before I was more boardroom than server room.
Besides, a cool grand for what I figured was 2 hours work appealed to me on some “miss the struggles” level, and it was better than walking around the tourist traps or doing video calls at some depressingly fecund coffee shop.
So, I drive up in my borrowed SUV with peeling paint, grab my trusty tool bag I leave with family for adventures like this, and walk boldly through the doors of this apparently normal retail establishment, clipboard in hand. It was 2019, and the smell of denim and factory pressed shirts wafted through my sensorium as I walked through the awkwardly desynchronized automatic doors.
It was a perfectly normal, somewhat 90s nostalgic retail store.., but walking up the stairs (the elevator was out of service) to the second floor, I could just sense that something was a little off.
The office had a smell. I can’t say what it was, it was businesslike, and not unpleasant, but it spoke to me in an unknown language and I was at that moment curiously heartened that my contract was being paid by a 3rd party. It smelled of stagnant ideas, missed opportunities, and a fossilized board of directors.
I’m not just being colorful here in prose. All of these thoughts were projected into my forebrain by some lizard sense I had developed in 40 years of skulking the halls of enterprises large and small. I was involuntarily imagining a half used tin of bengay on the boardroom table, right by the shiny beige speakerphone with the long stroke buttons and the filthy woven grill.
Curious by now about the flood of intuitions, I treaded the slightly dingy burnt orange carpet tiles up to the POC office, checked in, and was shown the door tot the server room. Through there she said, gesturing at a plain flat golden pine door with the elbow thingy at the top. “Through there, straight through, the door at the back. You can’t miss it. I think it’s unlocked.”
Unconcerned with the casual lack of escort (pretty common really), I opened the door and was immediately transported to surreal alternative world. It was dark, there were tens of flickering screens across to the right and one half hanging, flashing fluorescent fixture dangling precariously from the uni-strut above. The suspended ceiling tiles were about 50 percent in attendance.
Fragments littered the floor, and the door closed behind me with a loud thump. for just a moment I was touched by a singular thought….. (run). I shook it off, by now, far too curious to turn back, come what may.
Briefly thinking of my heavy tool bag as a potential weapon, then realizing the ridiculousness of the notion, I let my eyes adjust as I took in the scene with what I confess could only be described as some kind of unhealthy delight.
This was a large room, easily 800 square feet, and a was clearly dedicated to loss prevention surveillance. There was an epic workstation with 2 giant flat screen plasma monitors front and center, and ten smaller monitors on either side. Most of them were dark, and most of the others were flickering with static.a singular monitor had a poor but recognizable image of a small nondescript room with a small desk and two metal chairs, one on each side of the desk in an overtly adversarial position. The image smelled of stale sweat and musty cardboard.
There was seating for two pilots at the helm of this derelict craft, replete with an impressive array of buttons, strange industrial joysticks, and an oddly out of place black dell workstation keyboard with the cord ripped in half. As I made my way up to admire the helm, I nearly tripped on an extension cord, only catching myself on a water filled trashcan with plastic sheeting aspiring to the heavens up through the dark abyss of the half-intact ceiling, which I only now noticed was comprised entirely (if sparsely) of black painted tiles.
Somehow this jolted me out of my entranced state, and I looked for the door I was supposed to find, quickly locating it towards the back of the room, just as I had been told.
Somehow it seemed like not an ordinary door now. It was with great anticipation and not a little trepidation that I examined the door. It was quite ordinary, really, and the knob was less dusty than it might have been. This was encouraging.
I tried the door, and it opened easily. Opened, that is, to a time 30 years in the past.
The buzzing fluorescent lamps were on and radiating furiously, instantly shattering the wreckage of the surveillance spaceship to smithereens. There was a gigantic, wheeled, beige machine in the middle of the room, about the size of an industrial floor scrubber. It hummed patiently, and featured a dim 10 inch green CRT, a beautifully sculpted beige keyboard, and a little less than a quarter inch of dust. I froze like a tourist gaping at a stegosaur’s nose, but quickly recovered my composure as I remembered I was here on purpose.
Conveniently, someone had swept a path through the center of the room, back to the newly installed floor rack with a couple of ciscos, a backup cellular telemetry box, and a pile of screws and a throwaway screwdriver scattered haphazardly on top of one of the routers.
The UPS I was to replace was actually pretty complex to configure, but I very carefully completed it in about 45 minutes including some firmware regressions to accommodate some nonstandard plugin the vendor installed. By very carefully I mean without stepping too hard on the floor, dropping any tools, and afraid to sneeze even though I was under intense respiratory assault from the unavoidable clouds of fine carpet dander that had accumulated on the linoleum tile over years of janitorial malfeasance.
I was terrified to make any sudden moves, because sitting on a stackable plastic chair next to the pliestocene PBX was a beige box. Its fans groaned with the wails of a thousand parched capacitors, and its hard drive made a crunchy rattling sound that seemed entirely improbable for a functioning device. Just visible under the dust on the front was a badge proudly displaying “pentium pro” and a giant cable was fed from the back of the artifact, like an implant from the borg, something I strangely remember as “IBM type 1” token ring wire, though I can’t be sure anymore that means anything at all.
At any rate, it was obvious that the entire on-prem server infrastructure was this ancient and faltering nightmare box, and I knew that a fly landing on it could bring it down. The easy money contract was starting to look like a setup.
With terror in my heart, I finished my work, carefully closed the door, and put as much distance between me and that roiling cauldron of liability as I could. I don’t even remember going back through the wrecked starship, just checking out, and thanking Voltaire that I made it out of there in time.
It now made abundant sense that the job had been left on the board for nearly a month and the contract had kept rising… god only knows how many sensible technicians had noped out of there before I was hypnotized into compliance by the sheer bizarre fascination of the thing.
Poking around afterward, I found that the box apparently was the server that ran the terminals in the tire service area. They also had that fat cable that looked more like waveguide than wire, and it must have been an absolute nightmare to install. It was definitely token ring, humming along in 2019.
Sears closed for good a few months later.
Timber that was very expensive when it was fitted, and carpet tiles that have repeatedly got damp and not dried out properly. Very old cigarette smoke from when people were allowed to smoke in the office that is now embedded in the blinds and suspended ceiling tiles.
I was recently in a large suite of government offices, fitted out in the 1970s and largely unchanged since, and that was the exact smell. We're imagining the same smell right now.
How do the maintenance, development, and operations costs of a system like this compare to - for sake of argument - a typical line-of-business Java system running on commodity servers you own in a DC somewhere?
Can anyone with experience in both comment?
A VAXCluster or Sysplex(S/390) would I guess be more analogous but those can be a bit Ship of Theseus since you can change the hardware out over time.
My current contract has a PDP-11/70 running RSX-11 that's been up since 1992. It's uptime is longer than mine! Poor thing is getting retired next year though.
Stratus ztC Endurance extends fault tolerance to modern enterprise and operational environments with intelligent, predictive technology that prevents failures before they impact operations. Delivering up to *99.99999%* [my emphasis] availability, the platform combines proactive health monitoring, automated failover, redundancy, and simplified serviceability to keep applications continuously available and ensures that data stays protected.
Stratus ztC Endurance is ideal for organizations running mission-critical workloads that cannot tolerate downtime or data loss."
Source: https://www.penguinsolutions.com/en-us/solutions/fault-toler...
Booted Up in 1993, This Server Has Been in Continuous Operation Ever Since - https://news.ycombinator.com/item?id=13507972 - Jan 2017 (3 comments)
Perhaps there have been others on the topic?
(You can write real fault tolerant software but a wad of tangled crap and hidden dependencies crammed into Kubernetes is not it.)
A horror about the unchecked madness of human technological development, and a movie about dinosaurs.