摘要

In this paper, we propose a mobility-aware adaptive flow-rule placement scheme, named as Mobi-Flow, with art aim to maximize the overall performance in a software-defined access network (SDAN). The proposed scheme consists of two components - path estimator and flow-manager. The path estimator predicts future locations of end-users present in the network, depending on their history location sets, and delivers the predicted locations to the flow-manager. We use the order-k Markov predictor to predict the next possible locations of the end-users. Based on the predicted locations, the flow-manager determines access points (APs) in the network, which can be associated with the users to fulfill the requirements of the latter. We formulate an integer linear programming (ILP) to determine optimal number of APs, so that overall cost associated with flow-rule placement is minimized. Further, we propose a greedy approach to determine optimal number of APs, as solving the ILP is NP-hard. Consequently, the flow-manager implements the flow-rules at APs, so that adequate actions for the incoming requests can be taken in an adaptive manner, without querying the controller. We consider a practical scenario of an loT environment, in which both static and mobile users are present. Therefore, the proposed scheme, Mobi-Flow, can be integrated atop the SDN controller to support the emerging concept of SDN-based loT networking. Through extensive simulations, we show that Mobi-Flow is beneficial for minimizing the delay, number of activated APs, control overhead, energy consumption, and cost in the network, compared to the existing schemes-open shortest path first (OSPF), minimum occupied rule capacity (MRC), distributed (non-SDN), and MoRule. Particularly, the proposed scheme is capable of reducing the cost by 39, 38, 65, and 11 percent, compared to OSPF, MRC, distributed, and MoRule, respectively.

  • 出版日期2019-8