Towards Understanding Checkpointing in Transiently Powered IoT Networks

The finite energy supply available for computation and communication is one of the major shortcomings in IoT/sensor networks. To circumvent this issue, energy harvesting has been proposed to enable embedded devices to mitigate their dependency on traditional battery-driven power sources. However, energy supply due to energy harvesting often varies, leading to nodes crashing due to energy exhaustion, with the application(s) losing their 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). This work is based on the observation that, as the network evolves, IoT applications adapt/refresh to ensure they have up-to-date state of information, thereby reducing the need for checkpointing. However, little is known about (i) whether checkpointing is actually needed for any application, (ii) if needed, what to checkpoint for a given application and (iii) when checkpointing is needed in an application. In this paper, we address the first two problems and make the following novel contributions: (i) we formalise a variant of the checkpointing problem for IoT networks and show that it is NP-complete, (ii) we introduce a theory of state checkpointing and develop two types of checkpointing: (a) local checkpointing and (b) network checkpointing, (iii) we show that there is no general checkpointing strategy for the network-based application that will improve efficiency and (iv) we briefly survey related works and show that most if not all, works on checkpointing of IoT-based applications that lead to increased efficiency have focused on node-centric (i.e., local) programs. We run experiments on an actual testbed to confirm our expectations, and our results show that checkpointing can actually have a negative impact on applications when wrongly used. Thus, compared to existing works, we make major gains: No checkpointing is needed for network-based applications.