Resource-Aware Dynamic Service Deployment for Local IoT Edge Computing: Healthcare Use Case

Abstract

Edge Computing is a novel computing paradigm moving server resources closer to end-devices. In the context of IoT, Edge Computing is a centric technology for enabling reliable, context-aware and low-latency services for several application areas such as smart healthcare, smart industry and smart cities. In our previous work, we have proposed a three-tier IoT Edge architecture and a virtual decentralized service platform based on lightweight microservices, called nanoservices, running on it. Together, these proposals form a basis for virtualizing the available local computational capacity and utilizing it to provide localized resource-efficient IoT services based on the applications’ need. Furthermore, locally-deployed functions are resilient to access network problems and can limit the propagation of sensitive user data for improved privacy. In this paper, we propose an automatic service and resource discovery mechanism for efficient on-the-fly deployment of nanoservices on local IoT nodes. As use case, we have selected a healthcare remote monitoring scenario, which requires high service reliability and availability in a highly dynamic environment. Based on the selected use case, we propose a real-world prototype implementation of the proposed mechanism on Raspberry Pi platform. We evaluate the performance and resource-efficiency of the proposed resource matching function with two alternative deployment approaches: containerized and non-containerized deployment. The results show that the containerized deployment is more resource-efficient, while the resource discovery and matching process takes approximately 6–17 seconds, where containerization adds only 1–1.5 seconds. This can be considered a feasible price for streamlined service management, scalability, resource-efficiency and fault-tolerance.

Keywords

• IoT
• edge computing
• distributed computing
• virtualization
• resource discovery
• microservices