SDN-RA: An Optimized Reschedule Algorithm of SDN Load Balancer for Data Center Networks Based on QoS

Abstract

With the development of cloud computing during the latest years, data center networks have become a great topic in both industrial and academic societies. Nevertheless, traditional methods based on manual and hardware devices are burdensome, expensive, and cannot completely utilize the ability of physical network infrastructure. Thus, Software-Defined Networking (SDN) has been hyped as one of the best encouraging solutions for future Internet performance. SDN notable by two features; the separation of control plane from the data plane, and providing the network development by programmable capabilities instead of hardware solutions. Current paper introduces an SDN-based optimized Reschedule Algorithm (called SDN-RA) for cloud data center networks. The SDN-RA performance is validated and compared as results to other two corresponding SDN; ECMP and Hedera methods. The simulation environment of current work implemented using Fat-Tree topology over Mininet emulator which is connected to the Ryu-SDN controller. The performance evaluation of SDN-RA shows an increase in the network in terms of throughput and link utilization besides a reduction of RTT delay and loss rate.

Intelligent load balancing in data center software‐defined networks

“…Additionally, Tennakoon proposed a Q‐learning approach for load balancing in SDN, enhancing user experience and resource allocation in 5G networks. Kadim et al 10 contributed to SDN‐based load management algorithms, introducing dynamic load management and hybrid SDN‐based load balancing and scheduling mechanisms. Fu et al 11 proposed a deep Q‐learning‐based routing strategy for SDN‐based DCNs, demonstrating superior performance in reducing average delay and packet loss rates for mice‐flows while increasing average throughput for elephant‐flows compared to equal‐cost multipath (ECMP) routing and selective randomized load balancing (SRL) + FlowFit.…”