back to article Amazon cloud to BEND TIME, exist in own time zone for 24 hours

Amazon’s mighty cloud will exist in its own time zone for a 24-hour period next month. For 12 hours each side of midnight on June 30, Amazon will make each second on its AWS servers’ slightly longer than the accepted standard definition of one second. Amazon’s cloud will also run for up to half a second behind Universal …

  1. Alan J. Wylie

    Just like Google

    Sounds exactly the same as Google's "leap smear" approach

    The Register article from 2011

    Google Blog, 2011

    and on-going discussion in nntp:comp.protocols.time.ntp

  2. Stuart 16

    Google have been doing this for years. You keep using that word 'News' I don't think it means what you think it means!

    1. Richard 22
      WTF?

      So News only encompasses entirely novel things? By that logic all news bulletins would be extremely short - we had an election 5 years ago, so why report on the one this year? People are murdered all the time - why bother to report on a new one? Reporting of things which haven't happened yet, even if they are similar to things which have happened before seems valid, and may be useful to users of the service in this case - at the very least it appears that there is relevant information in the article to EC2 users.

      1. Steve Knox
        Headmaster

        Yes.

        So News only encompasses entirely novel things?

        What do you think the "new" in "news" means?

        Origin of news

        1425-75; late Middle English newis, plural of newe new thing, novelty (see new ); on the model of Middle French noveles (plural of novele), or Medieval Latin nova (plural of novum); see novel2

        (http://dictionary.reference.com/browse/news?s=t)

        1. Richard 22
          FAIL

          Re: Yes.

          Entirely novel and new are distinct concepts.

          The first wristwatch was an entirely novel thing. The Apple Watch is a new thing.

          The first democratic election was an entirely novel thing. The recent UK general election is a new thing.

          The first use of time smearing to compensate for leap seconds (who ever did it) was an entirely novel thing. Amazon using in this specific case is a new thing.

  3. ratfox
    Coat

    Sounds complicated

    Wouldn't it be more practical to have a gyroscope turn really really fast to accelerate the rotation of Earth?

    1. Alister

      Re: Sounds complicated

      Wouldn't it be more practical to have a gyroscope turn really really fast to accelerate the rotation of Earth?

      What we need is every able-bodied member of the population to stand facing away from the direction of the Earth's rotation, and start running really, really hard all at the same time...

      I'm sure this would be trivial to organise.

      Guys?...

      Oh come on, guys?

      Really?

    2. Justin Pasher

      Re: Sounds complicated

      Obligatory xkcd

      https://what-if.xkcd.com/26/

      1. VinceH

        Re: Sounds complicated

        That explains things. We don't need explanations - we need a solution!

  4. John H Woods Silver badge

    Why have you used ...

    ... a hoax BttF date, rather than the real one (Oct 15 this year?)

    1. John H Woods Silver badge

      Re: Why have you used ...

      >>> oops I meant 21 Oct 15 (this year) thank you for your tolerance

  5. This post has been deleted by its author

  6. Ken Moorhouse Silver badge

    What is so sacred about 9,192,631,770?

    Why not adjust the source of the measurement (to say 9,192,631,771) then filtering that down the hierarchy?

    1. frank ly

      Re: What is so sacred about 9,192,631,770?

      There would still be an eventual discrepancy because Earth's rotation speed is _continuously_ slowing down. There is nothing 'sacred' about the defined caesium standard, but it is very stable and accurate and easily reproduced by scientists all over the world, which is why it's used.

      1. Ken Moorhouse Silver badge

        Re: What is so sacred about 9,192,631,770?

        I understand that there is no fixed method that can ever be used, there's too many variables involved.

        The current thinking seems to be that it is ok to transiently kludge the interpretation of the standard in ways that are at the discretion of the organisation deriving the standard. My question is that surely the better method would be to tweak the ultimate standard itself, then there is no ambiguity in derivation, and no discrepancies between organisations using different forks of the primary standard.

        There are precedents that can be called upon to back up my suggestion. For example, the definition of the metre has been tweaked a few times over recent years.

        1. Eric Olson

          Re: What is so sacred about 9,192,631,770?

          The idea is to stop using physical items that can chance shape, size, density, etc. and have to be adjusted for the ambient conditions at the time of measurement. Seconds used to be defined as a 1/86,400th of a mean solar day. But of course, the Earth is very much a physical item that changes based on ambient. Notice the use of "mean" in the old definition of second. More importantly, with a slowing rotation, anything that uses second as a unit will change over time. That means that you would have to constantly update measurements of fundamental phenomena, like the speed of light. Today it might be 299,792,458 meters/second, but using an Earth-derived second means it might be 299,792,458.5 meters/second in a couple years. Defining seconds by the number of vibrations of a Cesium atom means it's tied to a fundamental property of nature, though one that can be still influenced by external forces, like local gravity.

          Even today the margin of error for an atomic clock is enough to add uncertainty to the true speed of light, and we have other. Adding a leap second for the sake of coordinating time is purely for those of us Earth-bound souls. And since most of us only measure time down to the second (with some people going to ms), that resolution is good enough. But perhaps we need to stop thinking about time from an earth-derived value.

  7. Anonymous Coward
    Anonymous Coward

    The whole idea of leap seconds is stupid and should be abolished forthwith. We already make do with local time which is set with time zones and varies within the time zone by plus or minus 30 minutes from true earth time at a given location (more or less, depending on the zone, since they use political maps).

    Let us wait until we are at least an hour off and create a leap hour several millennia from now. Then we can go back to worrying about the Unix clock rollover in 2038.

    1. Yet Another Anonymous coward Silver badge

      That was the proposal a couple of years ago - it was vetoed by the UK

      1. Anonymous Coward
        Anonymous Coward

        Actually it's still a proposal and the UK does not have veto power. An article in the Times (the real one, not the NYT, LAT, or other pretenders) science section suggested that the opposition stemmed from the belief that the abolition of leap seconds would mean the demise of Britain's role in timekeeping.

        The various proposal will be discussed at the World Radiocommunication Conference in November 2015. This is the group that has final say.

        1. Ken Moorhouse Silver badge
          Joke

          An article in the Times (the real one)

          You mean the one derived from the NPL standard.

  8. Anonymous Coward
    Anonymous Coward

    Why don't they just move the datacentres 1 second round the earth, problem solved.

  9. Crazy Operations Guy

    1 second per month in 3752

    By that time, I would assume that we would be off of earth and worrying about much more important things. Especially since keeping accurate time would be a right pain in the posterior to keep accurate anyway, what with general and special relativity rearing their ugly heads...

    1. Borg.King

      Re: 1 second per month in 3752

      "By that time, I would assume that we would be off of earth"

      I'll be gone by then.

    2. Brewster's Angle Grinder Silver badge

      Re: 1 second per month in 3752

      Post Newtonian effects are already an issue. We talk about time on the surface of the Earth (the geoid) which is different to that at the centre of the earth or the centre of the solar system. If you move an atomic clock gravitational effects have to be accounted for and I believe GPS takes account of them, too.

      1. Crazy Operations Guy

        Re: 1 second per month in 3752

        Yeah, I remember reading about how GPS satellites have to be engineered with both general AND special relativity taken into account, in addition to such things as the speed of propagation of Electromagnetic waves in atmosphere and, to a lesser extent, the gravitation lensing effect.

    3. VeganVegan
      Alien

      Re: 1 second per month in 3752

      If we're off the Earth, there is a decent chance that we would have learned how to travel really, really fast. In which case, time dilation might become significant and need to be corrected for; and we're back to leap seconds, and perhaps even leap years or decades, depending on how fast we are going!

  10. Crazy Operations Guy

    Astronomical time is pointless anyway

    Having illogically defined time zones already makes synchronizing with astronomical time pointless. I suppose it would make a bit of sense if our time zones where split correctly so that local 'noon' was at the point where the sun was directly 'overhead'.

    This whole leap seconds stuff, to me, is about as useful as moving the distance markers along the highway because the pavement expands and contracts with age. Sure moving it a few centimeters would make it more accurate, but you'd have to disrupt the flow of traffic to do it, and its such a small change that almost no one will ever notice the change itself, only the side effects.

  11. Ben Liddicott

    Also what Windows does, more or less

    I.e. ignores the leap second, and just treats it as time skew.

  12. Henry Wertz 1 Gold badge

    I don't see the problem

    Leap seconds are being used so, eventually, you don't end up with 12 midnight happening in the middle of the day. After all, leap years are there for a similar reason -- older calendars did not have leap years, and eventually became inaccurate enough that people began to notice. If you don't want to deal with leap seconds, don't deal with them, your clock will not be off by much over the likely lifetime of your clock or computer. But, personally, I have network time on my phone, and ntp on the computers.

    Note, ntp does have the "-x" option -- normally, (per the man page), ntp will jump the clock if it's off by more than 128ms, less than 128ms it does in fact slew the clock (speed it up or slow it down until it has correct time.) -x sets this cutoff to 10 minutes. The man page warns most UNIX systems allow a maximum slew of 0.5ms/second, so it'd take almost 14 days to slew 10 minutes. For 1 second, that'd come to about 33 minutes, so Amazon must be using a much slower slew rate. If you object that much to having your clock jump 1 second, use -x.

  13. John Savard

    Units of Measurement

    The length of the second, when combined with the mass of the kilogram and the length of the metre, gives us the size of other fundamental units like the ohm, the ampere, the watt, and the joule. So we can't change the length of the standard second.

    But it certainly is possible to have civil time run on non-standard seconds, as was the case before 1972. Atomic Time is based on Ephemeris Time, and we could continue to have that as well.

    1. Brewster's Angle Grinder Silver badge
      Boffin

      Re: Units of Measurement

      "Atomic time" is not based on "Ephemeris Time". "Atomic time" (TAI) is the average of a bunch of atomic clocks. "Ephemeris Time", well these day TT(TAI), is the independent variable used to calculate ephemeris. If your clock advances 86,400 second of atomic time then you add 86,400 to the coordinate used in your ephemeris and consult your model to determine where the sun, moon and stars will be.

      The leap second changes neither "Ephemeris Time" nor "atomic time", but adds 1 SI second to UTC (wall clock time) so the difference between the wall clocks and atomic clocks (UTC-TAI) increases from 35s to 36s which ensures the the (mean) sun is overhead at midday.

  14. BobRocket

    Do I get paid this time

    or is like the last time (and the one before that) where they make my working week a bit longer at my expense.

  15. TRT Silver badge

    I thought that...

    the Amazon was shrinking every minute? Or is that every week? Whichever of those, illegal logging, deforestation, species extinction, global warming etc.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like