CheckIn: Efficiently Checkpointing Intermittent Sensor-Based Internet of Things (IoT) Networks

One of the major limitations of current Internet of Things (IoT) sensor networks is the finite energy supply available for computation and communication. A potential solution that has been proposed is energy harvesting, which enables embedded devices to mitigate their dependency on traditional battery-driven power sources. However, energy supply due to energy harvesting varies both spatially and temporally, leading to nodes crashing due to energy exhaustion, with application(s) losing their operational state. Efficient state checkpointing in non-volatile memory (NVM) has been proposed to enable forward progress, albeit at the expense of significant overheads (viz., energy and time), thus a trade-off is necessary. It has previously been shown that, while checkpointing has a positive impact on the performance of some applications in transiently-powered nodes, checkpointing may adversely affect the efficiency of other applications. Our objective in this paper is to study checkpointing, to better understand how to boost its performance. We thus make the following contributions: (i) we introduce the CheckIn problem, which formalises the checkpointing problem in intermittent (transiently-powered) networks as a task scheduling problem, (ii) we prove that CheckIn is NP-complete, (iii) to circumvent the computational complexity, we propose an adaptive heuristic and compare its efficiency against known checkpointing techniques. Results show that our heuristic significantly outperforms other checkpointing techniques.