Proposed Scheme for Overhead Reduction and Qos Provision in SDN-Based Data Center Networks

Abstract

Software defined Data Center Network (SDDCN) architectures need the flexibility, agility, scalability, and improved analytics for the efficient working and reliable networks. However, as the traffic load of the network begins to grow the substantial collapse in the network performance has been observed. The huge amount of network broadcast traffic mainly ARP overhead, is one the dominant factors contributing towards this performance degradation. To this end, many solutions have been proposed that effectively reduces this broadcast traffic. However, to the best of our knowledge existing solutions do not focus on reducing the redundant ARP processed by controller especially for fattree and diamond loop topologies for large scale SDDCNs. In this research, a framework ARP-OR has been proposed to reduce ARP overhead traffic for the extensive and exhaustive software based data centers. Moreover, QoS has also been provided through ingress policing and queue management. The developed approach not only reduces the ARP broadcasts more effectively but also it suppresses all the redundant ARPs before the control plane processes them. Thereby, reducing the ARP request packets and correspondingly ARP replies and Flow modification messages. ARP-OR finds its scalability for tree topology, fattree topology as well as its variant diamond topology. It is prototyped on an open source RYU controller and experiments were conducted on Mininet emulator software for all the aforementioned topologies. Packet counter visualizer has been implemented for collecting the real time statistics of network traffic. ARP-OR has convincingly reduced the ARP traffic to almost 4.2% compared to existing approaches and 1% in contrast to simple l3 learning switch application. The proposed solution controls the network congestion, mitigates latency and provides efficiency, reliability, flexibility, and scalability. Consequently, it promises an exceptional deal for future deployment of software based scalable DCNs.