6

Inspired by a YouTube video I saw recently that claims a very high uptime for the computers on the Voyager probes, I was curious -- what computer currently has the longest continuous runtime?

For the purposes of this question, let's define "computer" as a stored program digital electronic computer.

closed as too broad by snips-n-snails, Tomas By, Joe, Curt J. Sampson, pipe Oct 5 at 10:12

Please edit the question to limit it to a specific problem with enough detail to identify an adequate answer. Avoid asking multiple distinct questions at once. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 5
    Given the possibility of the answer being a satellite, I worry that the answer is going to be something confusing to do with relativity. Helios-B seems to have made it to 0.02% the speed of light, if that helps anybody. – Tommy Oct 4 at 18:29
  • 5
    Space hardware often resets and goes into a safe mode. At least a soft reset if not a hard reset of almost all systems and a slow boot/recovery process to get everything working again. – Brian Oct 4 at 18:48
  • 2
    If “uptime” is defined as cpu above 90% utilization then we used to run our unix machines running cfd for 40 days continuous to solution... – Solar Mike Oct 4 at 19:51
  • 2
    Not really about "longest", but there are some fun stories in the comments: hardware.slashdot.org/story/13/03/14/2029205/… – fadden Oct 4 at 20:52
  • 1
    +1 Great youtube link ... as mentioned in other comments critical systems use Watchdog which is HW invoked interrupt (with some frequency) that resets the control system occasionally to prevent/recover from control system freeze so Up time got reseted too ... – Spektre Oct 4 at 21:11
11

The question is tricky in a fundamental way.

[...] very high uptime [...] longest continuous runtime

One part lies in the definition of these, partly contradicting, terms as well in what the computer is. Uptime is a term used for high level operating systems - such as a Unix system - but the Voyager systems are embedded computers running their application on bare metal. So the basic definition for uptime is already hard to tie to a core component.

Embedded systems are quite different from general purpose systems by being almost entirely interrupt-driven. One feature of embedded systems is that they routinely restart - restart being in fact the most important non-maskable interrupt. All to increase stability. So of course, some 'working since installed' time can be counted (*1), but it is rather meaningless in sense of computing value, as the systems aren't really running continuously.

Even by settling on some hypothetical overall working time, the result is still a rather meaningless value. After all, any such number would only be useful to compare comparable systems - and the Voyager computers are in no way comparable to any non-embedded systems in operation.

The other part is within the meaning of "uptime" itself. It can be given for various components with different meaning. It could define how long a system has been continuously powered. Or how long an OS has been running unstopped on a machine, or how long an application has run unstopped.

Think for example about an x86 machine running Linux running some large scale application like a database server - or better, a MUD. To a hardware person, the power up time is all that counts, while players would only look at the duration their game world persists. Fun part: neither cares about the OS-uptime. In fact, not even the admin does, as his job is to provide a running MUD, not just hardware or an OS.

The single case that the OS-uptime would matter is only true in a situation where the OS itself is the service provided - to many independent users. Not often nowadays, where everyone got has their own machine using services on remote servers - making one of our most favoured benchmark values rather useless to real life :))

Bottom line: The question sounds cool, but doesn't work out in this case.


Now, if we accept all of this and just settle for some 'working since installed' time, the best candidates might well be some traffic light systems of the early 1970s. :)


*1 - Well, and then it can't, as probe systems are usually don't work full-time, but go into sleep mode as often as possible to save on power.

  • 1
    Naa, they go down. Especially in the old days systems had daily downtimes for backup and disk reorganization. More important, these systems have been replaced with new hardware and extended software dozends of times since the 1970s. So while the core software may still have the same structure, it hasn't been running couninous in any definable way. – Raffzahn Oct 4 at 18:52
  • Embedded systems don't work that way, at least not most of them. – user Oct 5 at 10:00
  • 1
    Even with non embedded systems, uptime means different things to different people. There are mainframes that allegedly have uptimes measured in months or years but I think that usually refers to the operating system: in such cases they might still take the application off line for backups etc. My first paid work was on a system running on Unisys large computer architecture. Theoretically, it could run forever - even backups were done on line, but the application was so buggy, if it went a week without crashing we were doing well. – JeremyP Oct 7 at 9:14
  • @JeremyP I don't see any confusion... the question asks about computer up-time, so applications crashing (or even being started / stopped in a controlled manner) etc. have nothing to do with it. So long as the machine / operating system is still going (and hasn't been rebooted), that's one span of up-time. – TripeHound Oct 7 at 9:36
  • @Raffzahn I was intending to refer to only non-embedded systems (the comment I was responding to starts "Even with non embedded systems..."). Should have made that clearer. – TripeHound Oct 7 at 11:07

Not the answer you're looking for? Browse other questions tagged or ask your own question.