Traffic matrices for more reliable digital networks

We’ve all cursed an ineffective digital network, whether it’s delays streaming the latest Game of Thrones or a dangerous mobile phone overload during bushfire season. But no-one wants to pay extra for an over-engineered network. The secret to designing and testing a digital network to find the happy medium is a mathematical tool called a traffic matrix: a model of all the digital traffic within the network. Network engineers use traffic matrices to test a new network’s response to every feasible set of digital conditions. How much will data-flow fluctuate? Where are the hotspots? Where are the bottlenecks? They need to know how their network will fare in response to various situations: a spike in traffic along one route; the failure of a key gateway; or any number of other faults. Designing a set of traffic matrices that will cover every possible digital scenario is particularly difficult when the network doesn’t yet exist. “The answer is to use a system called maximum uncertainty,” says Professor Matthew Roughan, a mathematician at the ARC Centre of Excellence for Mathematical and Statistical Frontiers who specialises in traffic matrices, who worked with Dr Paul Tune at the University of Adelaide. “We minimise inadvertent assumptions within the system—such as where bottlenecks will appear, or the size of the largest digital traffic spikes.” The result will be networks that can cope with the real world—without needing to resort to expensive, ‘gold-plated’ over-engineering. Read about more work from the Centre here.