Keeping track of time

Volume 9 Number 2 February 11 - March 10 2013

From the early days of modern seafaring to the synchronisation of time over the internet, the world relies on clocks, and depends on their accuracy. By Annie Rahilly.

The historical ties between the advancement of civilisation, and the precision of timekeeping, are very strong.

A classic example is the development of accurate seafaring clocks enabling the determination of longitude, which led to the discovery of the new world. Another is the fleet of orbiting atomic clocks underlying the GPS system which is now integral to modern economies. 

Darryl Veitch, University of Melbourne Professorial Fellow in the Melbourne School of Engineering, is now conquering a new frontier of equal importance; the accurate synchronisation of computer clocks over the Internet.

Every computer ever built incorporates a clock as part of its core infrastructure, essential to many of its operations, as well as to much of the software it houses. Like a quartz watch, these clocks are based on an underlying crystal oscillator, together with software that counts its `ticks’ and converts them to a time in seconds. 

Such clocks are imperfect. They drift quite rapidly away from the true time, and it is necessary to repeatedly address this through synchronising them to reference clocks of higher accuracy. 

“We live in a unique point in history where almost all of the organisation of societies’ knowledge, as well as its critical infrastructure such as the electricity grid, telecommunications, transport and banking, is being put into the hands of computers, Professor Veitch says.

“Increasingly, this management is being performed by sets of computers widely separated geographically, using smart co-ordination over the Internet to reduce costs, improve performance and redundancy, and more recently to reduce energy footprints. However the tighter the teamwork and the larger the team, the more critical synchronisation becomes – timing is everything.” 

Near perfect time can be achieved through atomic clocks but installing such a device in every single computer is totally impractical and expensive. 

Professor Veitch, his colleague Dr Julien Ridoux and their `SyncLab' team have devised an inexpensive and convenient alternative for including time-keeping in computers.

Computer clocks are synchronised using a system known as Network Time Protocol (NTP). The problem is that NTP, which is one of the internet’s oldest protocols, is showing its age.

A new synchronisation algorithm (the RADclock algorithm), promises to be the core of the next generation of internet timekeeping. 

Other applications are found in telecommunications, Smart Grids, banking, internet measurement, temporal security, and back-up systems for the critical yet vulnerable GPS service. Whereas the status quo NTP solution is limited to millisecond level precision and is easily disrupted, critical and future applications require microsecond level precision, as well as graceful degradation in response to the frequent and severe disruption that is business-as-usual in the internet.

A testbed, which includes atomic clocks, high-end GPS receivers and other specialised hardware, is routinely used to test clock performance in field experiments.

The performance of the RADclock synchronisation algorithm, and the authoritative nature of its evaluation, has attracted the interest of numerous players in the precision timing arena.

The project has received funding from Google in California, the Australian Research Council, the FreeBSD Foundation, as well as Symmetricom, the market leader in timing solutions, and an ARC Linkage Grant with Symmetricom as Partner Organisation has just begun. 

It has also enjoyed equipment donations from Endace, Australia’s National Measurement Institute (NMI), Meinberg, and Symmetricom. An expanded global testbed is now being put in place including a number of universities both in Australia and abroad. Substantial progress has been made toward its adoption as the new timing standard, including the inclusion of the clocks’ kernel or most central components into the next release of the influential FreeBSD computer operating system.

The successful technology transfer to FreeBSD is particularly significant. It represents not only an important step toward universal adoption, but constitutes an open pipeline enabling widespread dissemination, directly toward end users, of future elements of the research effort.

Furthermore, as it is the operating system of choice for data centres, RADclock will soon be underpinning capabilities in Cloud Computing. The MAC OSX is also largely built on FreeBSD.

Darryl Veitch is a Professorial Research Fellow in the Department of Electrical and Electronic Engineering at the University of Melbourne.  He has just celebrated a coveted milestone: a 1998 paper, published in the premier journal the 'IEEE Transactions on Information Theory' has now been cited over 1000 times. 

www.eng.unimelb.edu.au