ICEr: An Intermittent Computing Environment Based on a Run-Time Module for Energy-Harvesting IoT Devices with NVRAM

Abstract

With the development of energy-harvesting technology, various applications have been developed that can be operated only with harvested energy, thereby making energy-harvesting technology suitable for edge devices in poor environments where battery replacement is difficult. However, devices with energy-harvesting technology have limitations: an application can operate intermittently in an energy-harvesting device, and the device’s energy is greatly affected by the environment and the state of the device. Intermittent computing causes abnormal progress or affords incorrect results. The factors affecting the energy of the device can change the operation of the device. To solve these problems, we propose the “Intermittent Computing Environment based on a run-time module” (ICEr), which dynamically controls and manages an application for normal operation in intermittent computing. ICEr comprises an energy checker and a controller. The energy checker measures the energy state of a device at run-time, and the controller controls and manages an application through Backup, Restore, Sleep, and Wakeup. The controller optimizes those operations by considering the energy state and memory state together to minimize time and energy overhead. In this study, two kinds of experiments were conducted. In the first experiment, Embench was selected as the target application to validate ICEr and measure its performance. This experiment validated that ICEr behaves dynamically in various environments. Moreover, it showed a reduction in relative execution time overhead of up to 50% and a reduction in energy overhead of up to 49.5% against Hibernus, depending on the environment. In the second experiment, ICEr was applied to the Temperature Measurement Application, and the improvement of the energy efficiency for the real Internet-of-Things (IoT) application was confirmed.