From Cloud to Edge: Seamless Software Migration at the Era of the Web of Things

Abstract

The Web of Things (WoT) standard recently promoted by the W3C constitutes a promising approach to devise interoperable IoT systems able to cope with the heterogeneity of software platforms and devices. The WoT architecture envisages interconnected IoT scenarios characterized by a multitude of Web Things (WTs) that interact according to well-defined software interfaces; at the same time, it assumes static allocations of WTs to hosting devices, and it does not cope with the intrinsic dynamicity of IoT environments in terms of time-varying network and computational loads. In this paper, we extend the WoT paradigm for cloud-edge continuum deployments, hence supporting dynamic orchestration and mobility of WTs among the available computational resources. Differently from state-of-art Mobile Edge Computing (MEC) approaches, we heavily exploit the W3C WoT, and specifically its capability to standardize the software interfaces of the WTs, in order to propose the concept of a Migratable WoT (M-WoT), in which WTs are seamlessly allocated to hosts according to their dynamic interactions. Three main contributions are proposed in this paper. First, we describe the architecture of the M-WoT framework, by focusing on the stateful migration of WTs and on the management of the WT handoff process. Second, we rigorously formulate the WT allocation as a multi-objective optimization problem, and propose a graph-based heuristic. Third, we describe a container-based implementation of M-WoT and a twofold evaluation, through which we assess the performance of the proposed migration policy in a distributed edge computing setup and in a real-world IoT monitoring scenario.