IC manufacturers face some unique challenges while deploying devices at the network edge. For power and bandwidth reasons, large volumes of data must be analyzed near the edge in real time rather than send it to a central data center. This requires highly efficient algorithms and innovative hardware/software design to accomplish some of the functions of more powerful computers in the data center. A big chunk of edge devices are remote, distributed or battery operated with limited connectivity, making energy efficient computation even more critical. In such scenarios, a single edge compute device that aggregates and processes the data from multiple sources could be used. A prime example is a car collecting contextual information about traffic, pedestrians, weather, along with its own sensors, processes part of it for the driver and car’s functions and sends the rest to the cloud for further processing. This talk focuses on how clustered and distributed systems of IoT/edge devices with balanced compute performance and energy efficiency can pave the way for a sustainable future.
