Dynamic Load Sharing in Memory Constrained Devices: A Survey

In recent years, we have witnessed unprecedented growth in the adoption of IoT devices. These devices provide immense benefits in various areas such as healthcare, industry, and home automation. However, most of these devices operate on a strict power requirement and are memory constrained. They can have limited CPU and I/O capabilities that may require applications to run at limited capacity and lead to network latency issues. One solution to ease the computational load is to have devices execute compute intensive and I/O intensive resources in the cloud. However, such an approach will incur additional network overhead. In addition, this approach might not be favorable to devices with no internet connectivity, such as wireless sensor networks. In an IoT ecosystem with heterogeneous devices of varied compute capabilities connected over a network, it is beneficial to ensure that unused compute resources (e.g. CPU and I/O) are efficiently distributed across all devices in the network. An approach that has the potential to reduce network latency, improve resource efficiency, and increase the Quality of Service (QoS) is to implementing load-sharing techniques locally across devices in an IoT network. The goal of this paper is to provide a review of recent work on load sharing techniques in an IoT ecosystem in order to create a robust compilation of the current state of the art in load-sharing in IoT. In addition, we also outline issues with integrating load sharing into the IoT ecosystem.