Testing Intermittent Battery-free Systems

Abstract

Existing tools for debugging battery-free applications are limited to specific architectures or require code changes of the Device Under Test (DUT) to function. These tools also cannot measure the efficiency of the application designed for battery-free systems. Currently, there is a lack of independent broad comparisons of intermittent systems. Our work, therefore, evaluates state-of-the-art frameworks and their artifacts and finds shortcomings in reproducibility and their performance. To overcome these shortcomings, we introduce DIPS+, a multi-platform debugger and measurement platform for intermittent systems with ARM and MSP430 support. DIPS+ introduces new methods to analyse applications for battery-free systems. One method finds the minimum energy budget required for forward progress, crucial for determining the minimal capacitor size for intermittent systems. Furthermore, DIPS+ offers functions to perform automatic profiling tests, like code start-up time, which gives valuable insights into the system's efficiency. DIPS+ achieves significant improvements in debugging performance, with 11 times faster connection time and reduced code execution by 157 times on the DUT before full reconnecting after intermittency occurs. The evaluation of selected frameworks reveals substantial overheads caused by the additional overhead of saving and restoring of system's state. This causes certain benchmarks to take up to 110 times longer to complete than their uninstrumented counterparts. These findings raise concerns about the viability of task-based approaches as an effective solution for managing intermittency in battery-free Internet of Things devices.