When Clones Flock Near the Fog

Abstract

FogMQ is a message brokering and device cloning service in fog and edge computing. Excessive tail end-to-end latency occurs with conventional message brokers when a massive number of geographically distributed devices communicate through a message broker. Latency of broker-less messaging is highly dependent on computational resources of devices. Deviceto-device messaging does not necessarily ensure low messaging latency and cannot scale well for a large number of resourcelimited and geographically distributed devices. For each device, FogMQ provides a high capacity device cloning service that subscribes to device messages. The clones facilitate near-theedge data analytics in resourceful cloud compute nodes. Clones in FogMQ apply Flock; an algorithm mimicking flocking-like behavior and allows the clones to autonomously migrate between heterogeneous cloud platforms. Flock controls and minimizes the weighted tail end-to-end latency. We have implemented FogMQ and evaluated it in a geographically distributed testbed. In our functional evaluation, we show that FogMQ is stable and achieves a bounded tail end-to-end latency that is up to 34% less than existing brokering methods.