Markov Approximation Method for Optimal Service Orchestration in IoT Network

Abstract

With the evolution toward 5G, the Internet of Things (IoT) is expected to manage resource dynamically and provide customized services for different users in a cost-effective manner. Many technologies, such as edge computing and network function virtualization (NFV), have responded to these demands and been even more critical for system mobility, scalability, flexibility, and resource utilization. In this vein, the efficient provisioning method of the IoT services in distributed clouds based on the diverse quality of experience (QoE) requirements is highly needed. This paper focuses on service chain (SC) orchestration and studies the optimal placement of virtual network functions (VNFs) with multiple instances to minimize cost and delay, as well as guarantee network load balancing. We propose a multi-objective optimization problem model and then convert this combinatorial optimization problem to the one which can be solved by distributed methods based on the Markov approximation. At last, a VNF placement with multiple instances algorithm (VPMIA) based on Markov chain is designed to solve the above problem in a distributed manner. The simulation results show that the proposed algorithm can outperform the random placement algorithm and the single-path placement algorithm in cost saving by 22% and 31%, respectively, with a high SC acceptance rate. Besides, it can guarantee the QoE requirements and make the network load balanced with the different numbers of SCs.

Keywords

• Distributed algorithm
• Internet of Things
• load balancing
• Markov approximation
• multiple instances
• service chain