A Game-Theoretic Based QoS-Aware Capacity Management for Real-Time EdgeIoT Applications

Abstract

More and more real-time IoT applications such as smart cities or autonomous vehicles require big data analytics with reduced latencies. However, data streams produced from distributed sensing devices may not suffice to be processed traditionally in the remote cloud due to: (i) longer Wide Area Network (WAN) latencies and (ii) limited resources held by a single Cloud. To solve this problem, a novel Software-Defined Network (SDN) based InterCloud architecture is presented for mobile edge computing environments, known as EdgeIoT. An adaptive resource capacity management approach is proposed to employ a policy-based QoS control framework using principles in coalition games with externalities. To optimise resource capacity policy, the proposed QoS management technique solves, adaptively, a lexicographic ordering bi-criteria Coalition Structure Generation (CSG) problem. It is an onerous task to guarantee in a deterministic way that a real-time EdgeIoT application satisfies low latency requirement specified in Service Level Agreements (SLA). CloudSim 4.0 toolkit is used to simulate an SDN-based InterCloud scenario, and the empirical results suggest that the proposed approach can adapt, from an operational perspective, to ensure low latency QoS for real-time EdgeIoT application instances.