UC San Diego

Intermittent computing systems harvest limited energy from the environment to power their sensing and computing work. Due to low and irregular power harvesting, these systems contend with unpredictable interruptions. Research has focused on ways to cope with the intermittent energy, leveraging soft and hard intermittency, and driving advances in intermittent computing hardware architecture, time-keeping, and software runtimes. However, intermittent computing systems are still difficult to design, program, and deploy. This work explores ways to simplify intermittent computing state management by exploring low power energy sources and analyzing boot and sleep performance. I present a module for real wall-clock timekeeping, leveraging extremely low power sources to maintain time indefinitely. I then analyze the true costs of hard intermittency for intermittent computing hardware platforms, and develop a model and analysis demonstrating that in most cases, especially where computing is required, soft intermittency is actually the ideal mode of operation. Wall-clock time retention enables the use of more complex networking protocols, and opens the door to easier synchronization and collaboration across networks of intermittent computing devices. And lastly, I argue that batteries, coupled with soft intermittent systems, can last decades, and should be considered for intermittent computing designs moving forward.