Demo: Scaling on the Edge – A Benchmarking Suite for Human-in-the-Loop Applicationss

摘要

Previous works on cloudlets, one of the earliest incarnation of edge computing, enable small data-centers at the edge of the Internet. Many futuristic applications become viable with these clusters that are only one wireless hop away. One of the most promising genres of these emerging applications is human-in-the-loop applications such as wearable cognitive assistance. In these applications, sensor data, for example video and audio, are continuously streamed to a cloudlet, where they are analyzed in realtime in order to assist users to complete a particular task. Benchmarking infrastructures for these human-in-the-loop applications is challenging - the main issue arises from the involvement of humans. Applications' execution path and resource utilization vary among users. In this demo we present a methodology and benchmarking suite capable of tackling this challenges through the use of prerecorded sensory input traces, which allows for efficient scaling of benchmark scenarios.