Sensing flow execution engine for concurrent mobile sensing applications

Abstract

Emerging mobile sensing applications are changing the characteristics of smartphone workloads. Based on the embedded sensing capabilities of smartphones, mobile applications can identify users’ contexts such as location and activity anytime and anywhere, and provide autonomous, situation-aware services. Whereas typical mobile applications run alone in the foreground interacting with users, sensing applications concurrently run in the background, providing unobtrusive monitoring services. This new class of applications and their workloads bring new challenges on resource-limited smartphones. First, mobile sensing applications consume considerable amount of resources on smartphones for continuous sensing and processing. Such continuous, high resource consumption limits the number of services a user can run. Also, it can significantly reduce the smartphone’s battery lifetime. Second, greedy and contentious resource use by concurrent sensing applications could degrade the performance of other daily use of smartphones. Also, some sensing applications could starve due to the resource contention, resulting in serious degradation of their service qualities. Lastly, from the viewpoint of developers, it is very challenging to build sophisticated sensing logics from scratch without any programming support. It involves multi-lateral challenges including careful selection of machine learning algorithms, repetitive parameter tuning, and low-level resource optimization. To address these challenges, I have designed and developed SymPhoney, a sensing flow execution engine for emerging sensing applications. The engine provides an intuitive and flexible programming model that significantly lightens the burden of developers. More importantly, it effectively coordinates the execution of concurrent sensing flows, minimizing the degradation of their service qualities even under contention. Also, the engine executes the sensing flows in a highly efficient way. As its key approach, ...