Key points are not available for this paper at this time.
With the increasing computing needs of the new systems and applications, cloud offloading has become a popular choice for constructors to keep the prices of their devices affordable. However, this solution only shifts the scaling problem from the end devices to the cloud, increasingly enhancing the capacities of cloud infrastructures. As a way to reinforce the cloud capabilities on the edge without needing to add extra computing resources, we propose PyCloudIoT, a collaborative energy-efficient Function-as-a-Service (FaaS) computing platform (pltf.) with low-to-medium availability targeting the execution of punctual stateless functions over the already deployed IoTs and gateways. As these resources are extremely dynamic, with intermittent availability, heterogeneity and faultiness, the addition of strong control mechanisms is key to efficient operation. In this paper, we discuss the PyCloudIoT Consensus Model (PCM), which enables the coordination and orchestration of resources dynamically and compensates for the faults of the IoT computing farm. Compared to SOTA, PCM shows promising results with a performance and energy consumption improvement of 20% and 66% and 37% and 65% respectively compared to the best configurations of Raft and Pirogue (4+1 quorum), achieving at the same time a slightly stronger fault tolerance level.
Blanco et al. (Mon,) studied this question.