Resilient partitioning of pervasive computing services

Abstract

The vision of pervasive computing is being realized with many elements of pervasive computing becoming available: hand-held, wall-sized and wearable computers, wireless technology and devices to sense and control appliances. While this vision is becoming more realistic from the hardware point of view, the matching software tools have not matured at the same rate. Pervasive computing environments are naturally distributed over multiple devices connected via wireless technology. This thesis explores the application partitioning techniques for development of pervasive computing applications. Using these techniques, software applications can be decomposed into parts that can be distributed amongst multiple devices. Application partitioning in a pervasive computing environment raises new requirements: run time application partitioning that is user controlled, retractable and resilient to network failures. We observe that current state of the art software tools miss out these important characteristics that apply to the pervasive computing environment. We propose a resilient actor model to structurally add this property to the pervasive applications. This model ensures that an application can be dynamically partitioned to run on any available device in the surroundings at run time. The partitioning realized using resilient actor model is user guided and the resulting partitioned application is resilient to network failures.