An Autonomic Software Defined Network (SDN) Architecture With Performance Improvement Considering

Alireza Shirmarz,Ali Ghaffari

doi:10.29252/jist.8.30.121

Abstract

SDN makes the network programmable, agile, and flexible with data and control traffic separating. This architecture consists of three layers which are application, control and data. The aim of our research is concentrated on the control layer to improve the performance of the network in an autonomic manner. In the first step, we have categorized the performance improvement researches based on network performance improvement solutions proposed in the recent papers. This performance improvement solution clustering is one of our contributions to our paper. The significant contribution in this paper is a novel autonomic SDN-based architecture to ameliorate the performance metrics including blocking probability (BP), delay, jitter, packet loss rate (PLR), and path utilization. Our SDN-based autonomic system consists of three layers (data, autonomic control, and Route learning) to separate the traffics based on deep neural networks (DNN) and to route the flows with the greedy algorithm. The autonomic SDN-based architecture which has proposed in this paper makes better network performance metrics dynamically. Our proposed autonomic architecture will be developed in the POX controller which has developed by python. Mininet is used for simulation and the results are compared with the commonly used SDN named pure SDN in this article. The simulation results show that our structure works better in a full-mesh topology and improves the performance metrics simultaneously. The average performance is improved by about %2.5 in comparison with pure SDN architecture based on the Area Under Curve (AUC) of network performance.

Highlights

The number of internet users is growing rapidly, so the network resources will be restricted for Internet service providing in the future
The main question in this paper is how to improve software defined network’s performance in an autonomic manner? Which we solve this problem with network performance improvement approaches categorization in Software Defined Network (SDN) and a novel autonomic SDN-based architecture to cover our problem and improve the network performance metrics simultaneously
Three scenarios are defined with a different number of nodes and links which will be defined in the following subsection

Summary

1- Introduction

The number of internet users is growing rapidly, so the network resources will be restricted for Internet service providing in the future. According to this Internet growth, different architectures have been proposed in various scrutinies [1]. The application layer is based on network application and is connected with the controller with northbound APIs. Shirmarz & Ghaffari, Autonomic Software Defined Network (SDN) Architecture With Performance Improvement. We investigate the solutions and propose an SDN-based architecture to make dedicated performance metrics improved. This our proposed architecture's advantages and drawbacks will be discussed in comparison with other ones

2- Related Works

For each path in paths

Initialize the hybrid deep learning network with obtained

5- Experiments and Results

5-2- Evaluation

6- Experiments and Results