Here, there, anywhere: Profiling-driven services to tame the heterogeneity of edge applications

Edge computing alleviates network bottlenecks by engaging devices at the edge for data processing tasks. These devices possess limited computing resources, while edge execution environments are inherently heterogeneous. It is non-trivial to dynamically allocate the limited resources of heterogeneous devices to balance high performance and low resource utilization. To that end, this paper presents a profiling-based methodology that effectively matches edge computational tasks with the available devices to best satisfy programmer-defined non-functional requirements. Edge tasks are divided into microservices, which are then profiled for their resource utilization. To execute each task, the runtime consults the profiling results to determine the optimal device-to-microservice matching. We realize our methodology as µHTA, whose service-oriented architecture manages the heterogeneity of edge environments and optimally executes microservices on the available mobile devices. We evaluated µHTA by applying it to two realistic edge mobile applications. Our methodology can help developers bridge the gap between the statically specified non-functional requirements and the dynamic nature of edge environments, while reducing the developer burden of optimally utilizing edge-based computing resources.