The world uses two primary systems for measuring time: GMT (and its various alternative names such as Zulu Time and Western European Time); and Coordinated Universal Time (UTC). The former is based on solar time (the rotation of the earth) – and fluctuates slightly. The latter is based on the constancy of atomic clocks: it is used to synchronize time standards on the internet, banking, air traffic and other critical computer-based international systems – and does not fluctuate. If left to their own devices, these two systems would gradually drift apart. Since we have no control over the earth’s rotation, we keep them synchronized by the occasional adjustment of a ‘leap second’ into UTC.
The most recent leap second occurred over the weekend – on 30 June at 23:59:60 UTC to be precise. Leap seconds always occur either on the last day of June, or the last day of December. Not all computers handled this inclusion of an extra second into their timekeeping smoothly, and reports of problems are already emerging.
The Australian Herald Sun has reported that global airline reservation system, Amadeus, crashed. Qantas was unable to check in passengers. Fifty flights were delayed and passengers stranded for several hours.
Many older versions of popular Linux distributions have a problem with how the Network Time Protocol (NTP – used to synchronize computer clocks over variable latency networks) handles a leap second. If administrators don’t apply the available patches, or simply shut down for the duration of the change, systems can crash. “Sites such as Reddit, Gawker, LinkedIn, Foursquare and Yelp crashed after a ‘leap second’ was added to the universal clock,” reports the Herald Sun. A Pirate Bay crash has also been blamed on the leap second, although other reports suggest it was a simple server problem. “One of the server racks hosted by the Swedish Pirate Party was moved, and this took down a crucial TPB server,” notes TorrentFreak.
The leap second problem is not new. There have been 25 leap seconds since they started in 1972. Back in September 2011, Google’s official blog explained how the internet giant handled the 2008 leap second (following minor problems caused by the 2005 leap second) with what it termed ‘leap smear’. “We modified our internal NTP servers to gradually add a couple of milliseconds to every update, varying over a time window before the moment when the leap second actually happens,” it wrote. “This meant that when it became time to add an extra second at midnight, our clocks had already taken this into account, by skewing the time over the course of the day. All of our servers were then able to continue as normal.” With no reports of Google problems this year, it would appear the method still works.