ITER networks | The quiet heroics behind time synchronization

ITER will rely on hundreds of computers, ranging from diagnostics and actuators to servers running physics models or providing a human interface to operators in the control room. All of these computers will need to be connected—and they must all agree on the time. Clocks on the ITER network need to be synchronized to within 20 nanoseconds for at least three key reasons. First, measurements coming from different diagnostic and instrumentation sensors have to be correlated for analysis. While data is time stamped on acquisition, if the clocks do not agree the data from an entire experiment could be of no value. Second, control systems need to issue commands to be carried out by multiple actuators in lock step. For example, corrections to the plasma shape or position inside the vessel require several magnets to take actions in a synchronized manner. Finally, ITER needs a general platform for real time distributed processing. Several plant systems require multiple processes that start and finish at predicable times—and this can only work if all systems run off the same clock. Ensuring conformance of computing equipment from different suppliers "One of our challenges is that different suppliers from around the world provide pieces of computing equipment," says Bertrand Bauvir, leader of the Central Control Integration Section. "To make it easier for all systems to interoperate, we use an international standard for time synchronization—the IEEE 1588 standard, which is also called precision time protocol (PTP)." PTP operates along similar principles as White Rabbit, an extremely precise synchronization technology that was developed at CERN around the same time. But while the CERN technology requires specialized networking equipment, the 20-nanosecond accuracy needed by ITER can be achieved with commercial-off-the-shelf network components procured from industry leaders. ITER can rest assured that PTP does the job. The standard was first published in 2002 and has since undergone two major revisions. For over a decade, PTP has enabled a variety of applications where time is of the essence, including two that modern societies could never do without: financial trading systems and mobile phone towers. The protocol works by having a master clock send the time to all other computers. The other systems set their clocks using that time and add the transmission delay—the time it takes for a message to get from the master to the recipient. To compute the delay, messages are sent back and forth and the sending and receiving times compared. To overcome the fact that workloads fluctuate, network switches detect PTP traffic and provide information on transit latency. Even though PTP is a solid protocol, no time synchronization system works unless all suppliers adhere to the standard. To make sure of this, Bauvir and colleagues test every new piece of equipment before it is connected to the ITER network. "We validate not only that each component operates, but also that it reaches its specified performance," says Bauvir. "Our work doesn't stop there. Once the network is running, we will continuously monitor every piece of equipment—collecting performance data, health data and configuration changes. We will continuously update a map of our time synchronization system to be able to detect at any moment if something is abnormal."