QRTP:QoS-aware resource reallocation based on traffic prediction in software defined cloud networks

Abstract

The complexity of network routing infrastructures is increased dramatically. To tackle this issue, a promising state-of-the-art paradigm called software defined networking (SDN) is introduced. Due to the differences in the network requirements of applications, QoS-aware routing plays an important role in the networks. Recent proposed resource allocation algorithms focus on the current traffic matrix which is not applicable for dynamic networks (e.g., over-the-top services). In this paper, we exploit an estimation of flow matrix which gives our scheme the ability to sufficiently reduce the total packet loss and simultaneously rises the network throughput. In this way, we mathematically formulate the QoS-aware resource allocation in software defined cloud networks using traffic prediction. The corresponding optimization problem is to minimize the maximum link utilization subject to delay, bandwidth, and flow conservations. To solve this optimization problem, two schemes are proposed: 1) exact solution and 2) fast suboptimal one. In the simulation, we first compare the two proposed schemes from accuracy and optimization time perspective. Moreover, the impact of prediction on resource allocation is discussed. In this way, it is shown that in the existence of burst traffic, our proposed scheme which exploits traffic estimation, decreases the total packet loss, more than 20 percent in compare with conventional approaches. Moreover, the total throughput increases 30 percent. The experimental results show that relaxed solution is a good approximation of optimal solution in all test cases while the optimization time of approximate algorithm is superior to the original one.