Towards end-to-end resource provisioning in Fog Computing over Low Power Wide Area Networks

Abstract

Recently, with the advent of the Internet of Things (IoT), Smart Cities have emerged as a potential business opportunity for most cloud service providers. However, centralized cloud architectures cannot sustain the requirements imposed by many IoT services. High mobility coverage and low latency constraints are among the strictest requirements, making centralized solutions impractical. In response, theoretical foundations of Fog Computing have been introduced to set up a distributed cloud infrastructure by placing computational resources close to end-users. However, the acceptance of its foundational concepts is still in its early stages. A key challenge still to answer is Service Function Chaining (SFC) in Fog Computing, in which services are connected in a specific order forming a service chain to fully leverage on network softwarization. Also, Low Power Wide Area Networks (LPWANs) have been getting significant attention. Opposed to traditional wireless technologies, LPWANs are focused on low bandwidth communications over long ranges. Despite their tremendous potential, many challenges still arise concerning the deployment and management of these technologies, making their wide adoption difficult for most service providers. In this article, a Mixed Integer Linear Programming (MILP) formulation for the IoT service allocation problem is proposed, which takes SFC concepts, different LPWAN technologies and multiple optimization objectives into account. To the best of our knowledge, our work goes beyond the current state-of-the-art by providing a complete end-to-end (E2E) resource provisioning in Fog-cloud environments while considering cloud and wireless network requirements. Evaluations have been performed to evaluate in detail the proposed MILP formulation for Smart City use cases. Results show clear trade-offs between the different provisioning strategies. Our work can serve as a benchmark for resource provisioning research in Fog-cloud environments since the model approach is generic and can be applied to a wide range of IoT use cases.