Clock Synchronisation Over Networks
Collaborators: Owen Jones (UniMelb) Paul Barford (Uni WiscMad) Rocky Chan (PolyU)
Every computer on earth has a software clock on board which is essential for its own operation as well as that of applications running on it. The vast majority of these clocks are kept synchronised using a system called NTP that connects to clocks of greater accuracy over the Internet. The problem is, this system has many problems, leading to incorrect timestamps which in turn leads to software failures, and likely system collapse under the weight of the Internet of Things.
This research program will define and solve the research problems needed to design a network timing system which is highly accurate, robust and reliable, and whose performance can be trusted and verified, even at the scale of the Internet of Things. It will make use of a state of the art timing testbed at UTS which includes GPS receivers and specialised timing appliances donated from leading manufacturers. In the past this project has received funding from the ARC, Google, and Symmetricom, and its partners include the National Measurement Institute, the body responsible for Australia’s reference time. The (to be updated) website describing the work is at www.synclab.org, where our publications can be found under the Documentation tab. In the interim, recent publications can be found at http://crin.eng.uts.edu.au/~darryl .
The following (non-exhaustive) list of sub-projects are available.
Network Timing Health: mapping out the web of timing servers across the Internet, to find the good, the bad, and the very bad. See our recent publication to see how many servers got it wrong in 2015's end-June leap second event!
Server Anomaly Detection: developing rigorous methodologies to detect anomalous events in timing servers, even from the other side of the world.
Timing Network Design: re-engineering and hardening the timing system to cope with its huge load while maintaining high accuracy and reliability, and resisting attack. There will be opportunities to trial this alongside Australia’s reference time at the NMI.
Universal Clocks: designing a clock system and algorithms capable of synchronising to all available sources: NTP, PTP, and pulse-per-second from atomic clocks or GPS.
Spreading the Word: pushing the project’s clock solutions into Linux and BSD operating systems to improve clocks for millions. We are already partly in FreeBSD.
