How to Prevent Costly Leap Second Interruptions

At the end of June this year, several high profiles websites suffered disruption and went down due the inclusion of an additional second to the international time system. The websites, including the social news and networking sites Reddit, Foursquare and Linkedin, were disrupted for several hours thanks to the inclusion of a Leap Second to Coordinated Universal Time (UTC), the world’s global timescale.

While a single second may not sound like much, for websites that rely heavily on synchronised operations, the added time caused such disruption that several sites went down for nearly twelve hours and others suffered varying problems such as lost data. These disruptions were not unique, either. Leap seconds are added to the international timescale regularly, often twice a year, in a tinkering with time that is becoming increasingly controversial.

Synchronisation

Websites and other modern technologies all rely on the world’s global timescale for synchronisation. UTC enables computers across the globe to communicate in perfect synchronisation, and without it, many of the online transactions we take for granted, such as social networking, banking and shopping simply couldn’t happen.

Computers rely heavily on time in the form of timestamps as a means of telling when a transaction has occurred or needs to occur. If computers weren’t accurately synchronised, computers would have no way of knowing if process had or hadn’t happened and all sorts of errors would take place. Lost data, transaction failure and vulnerability to fraud would all be common without proper synchronisation. You only have to think back to the Millennium Bug to realise the sorts of problems that computers and other technologies could suffer from without precise and accurate synchronised time.

Coordinated Universal Time

UTC was devised in the 1970s because advances in technology meant there was a need for a single, universal time to ensure machines had a single timescale to which devices across the globe could synchronise without being affected by time zone difference. As. UTC is based on the time told by atomic clocks it is exceptionally accurate, enabling computer networks on opposite sides of the globe to be synchronised within a few milliseconds of each other. Without atomic clocks, networks could have differences in several seconds, which would result in many of the errors mentioned above.

In fact, atomic clocks are even more accurate than the rotation of the Earth. While we may all think the Earth takes 24 hours to make one complete rotation, this is not exactly the case. Each day can vary minutely. Sometimes it can take a little longer than 24 hours, sometimes a little less time. All these tiny differences mean that by the end of the year, the Earth’s rotation and Coordinated Universal Time are slightly out of sync with each other, and so require a slight adjustment.

Leap Seconds

Since UTC was introduced and atomic clocks began governing our time, the growing trend has been that the rate at which the Earth spins is gradually slowing down. We now know that this has been the case for millions of years. For instance, 100 million years ago, the length of a day on Earth was about 22 hours, and the day has been lengthening ever since. This gradual slowing of Earth’s rotation is caused by tidal forces that act as a brake and slow the rate at which Earth spins. These tidal forces are in turn caused by the Moon, which is gradually getting farther away as the Earth’s spin slows down.

While these cosmological changes are very slight, and equate to only a second or two a year, it does mean that UTC requires the addition of an extra second twice a year to ensure it is synchronised with the rate at which the Earth is revolving. If this didn’t happen, eventually the day would drift into night and noon would no longer be the middle of the day, although this would take millions of years, and the decision to add additional seconds to keep the Earth and UTC synchronised is a controversial one.

Controversy

The International Earth Rotation and Reference Systems Service (IERS) decide when to add additional leap seconds. IERS monitor the rate at which the Earth revolves, and since the 1970s, they have added 25 leap seconds to UTC. However, these changes have proved increasingly controversial.

Because of the problems that adding these extra seconds cause, and the fact that it would take thousands of years before any difference in UTC and mean solar time (when noon happens when the sun is at its highest) would be noticeable, critics argue leap seconds are not necessary. They argue other measures could be taken, such as a leap hour every thousand years or so, which would prove far less disruptive, or even a leap minute every thirty years. However, so far, no consensus has been reached and leap seconds look set to stay for the foreseeable future.

Network Time Protocol

Computer networks are synchronised to a form of UTC using Network Time Protocol (NTP). NTP is a software algorithm that ensures all devices on a network are synchronised together. It works by using a single source of UTC, which it then distributes around a network, adjusting system clocks to ensure each device is within a few milliseconds of the UTC source.

If systems clocks are left to their own devices without continual adjustment from NTP, they soon drift. This is because most computers have simple quartz oscillators on their motherboards to maintain time, but these can lose several seconds a week. For a computer, which can perform hundreds of processes each second, the more accurate the time, the more efficent the network will be, which is why networks use atomic clock based systems such as UTC.

Often, problems with leap seconds occur because sources of UTC change automatically at a certain time. For instance on 30 June of this year, at 11:59:59 an additional second was added, which meant the time stuttered for a second, thus causing problems for websites and networks that struggled to accommodate the change. However, a simple solution is available to prevent this disruption.

GPS Time Servers

The Global Positioning System (GPS) provides an accurate source of time generated by atomic clocks. While GPS time signals are not a source of UTC, it is a time generated by atomic clocks and is easy for NTP to adjust it to match UTC, as the signals contain all the necessary data for conversion, such as the number of leap seconds to include.

GPS time signals are what satellite navigation devices use to calculate positioning. By using the timestamps in the GPS signal, a sat nav device can triangulate its position to within a few metres. This can only be done using the precise timing produced by atomic clocks because the time signals travel from the satellites at the speed of light, and even a fraction of a second of discrepancy could see positioning information out by several miles.

By using GPS Time, computer network administrators can make their own provision to make alterations for leap seconds, rather than be taken unawares by automatic adjustments made to UTC time sources such as radio transmissions and internet based time servers, over which network administrators have no control.

GPS time servers

Receiving a source of GPS Time is simple and inexpensive too. Dedicated NTP GPS time servers are readily available that can receive GPS time signals and use it as a source of time synchronisation for a network. These time servers use a roof top antenna that receives the time, which it then distributes around a network using NTP. A single GPS time server is capable of servicing a network of hundreds of machines so it is an inexpensive solution for time synchronisation. Furthermore, as the GPS time signals contain all the relevant information to convert GPS Time to UTC, the network can automatically be converted enabling synchronisation with other networks, while leap second provisions can be planned into the network, avoiding sudden disruptions.

Some networks, such as the one used by search engine giants Google, introduce a leap second over a prolonged period, introducing it in tiny increments for several weeks, which in effect smears the leap second of a longer period, preventing the sudden stutter of an additional second.

Security and accuracy

GPS time servers also ensure a network is secure. Some sources of UTC time, such as those taken from internet time servers, provide a gateway in network firewalls through which viruses and other malicious software could gain access. GPS Time on the other hand, is external to a network and it is impossible for somebody to tamper with it. Furthermore, internet time servers may not be as precise as many users imagine them to be. Some time servers can be several seconds out of sync with UTC. In addition, if the online time server suffers an outage, the network will be without an accurate source of time until service is restored, which could result in system clocks drifting so the network loses its synchronisation. GPS Time by comparison is always precise, always accurate and never goes down.

This post was written by

Richard N Williams

Richard N Williams is a technical author and a specialist in the NTP Server and Time Synchronisation industry. Richard N Williams on Google+

Related Reading