LW-CoEdge: a lightweight virtualization model and collaboration process for edge computing

Abstract

Edge Computing is a novel paradigm that extends Cloud Computing by moving the computation closer to the end users and/or data sources. When considering Edge Computing, it is possible to design a three-tier architecture (comprising tiers for the cloud devices, edge devices, and end devices) which is useful to meet emerging IoT applications that demand low latency, geo-localization, and energy efficiency. Like the Cloud, the Edge Computing paradigm relies on virtualization. An Edge Computing virtualization model provides a set of virtual nodes (VNs) built on top of the physical devices that make up the three-tier architecture. It also provides the processes of provisioning and allocating VNs to IoT applications at the edge of the network. Performing these processes efficiently and cost-effectively raises a resource management challenge. Applying the traditional cloud virtualization models (typically centralized) to virtualize the edge tier, are unsuitable to handle emerging IoT application scenarios due to the specific features of the edge nodes, such as geographical distribution, heterogeneity and, resource constraints. Therefore, we propose a novel distributed and lightweight virtualization model targeting the edge tier, encompassing the specific requirements of IoT applications. We designed heuristic algorithms along with a P2P collaboration process to operate in our virtualization model. The algorithms perform (i) a distributed resource management process, and (ii) data sharing among neighboring VNs. The distributed resource management process provides each edge node with decision-making capability, engaging neighboring edge nodes to allocate or provision on-demand VNs. Thus, the distributed resource management improves system performance, serving more requests and handling edge node geographical distribution. Meanwhile, data sharing reduces the data transmissions between end devices and edge nodes, saving energy and reducing data traffic for IoT-edge infrastructures.