Handling Heterogeneous Traffic  for Software Defined Data-Center Network Using Spike Neural Network

Software Defined Networking (SDN) allows for more flexible network administration than traditional architectures. Software-defined networks (SDNs) efficiently manage data flows and optimize network resources. However, heterogeneity influences the quality of the services. (QoS) needs and network resource demands. They behave differently when traveling to their last point. Currently, numerous data center networks (DCNs) struggle with the unfair use of several network resources by big packets (Elephant flowing) arriving during any instant affecting specific flows (mice flow). Elephant Flows (EF) account for just a small percentage of the entire traffic. Nevertheless, they are considered Long-lasting (LLF) and often burn network resources. Their actions cause congestion and delays in most Mice Flows (MF). Forecasting and categorizing flow traffic is essential for optimal resource usage, QoS provisioning, and reducing network congestion and delays. This paper suggested a third-generation Single Spike Neural Network (SSNN) supervised learning approach using temporal coding to identify heterogeneous traffic. The classifier approach uses three features: flow time, byte rate, and packet rate. The SSNN is then taught to categorize the traffic into two classes. This training classifies two types of traffic: elephant and mice flow. The effectiveness of the algorithm was examined when classifying traffic using many metrics and its efficiency was proven as it was able to reduce the average error and its accuracy reached 99%. The suggested model's usefulness is demonstrated by its efficient training procedure, which provides rapid and accurate results.