Analysis of the performance of SDN cloud controllers using non-Markovian queueing models

Cloud computing is valuable for businesses, offering benefits like 'measured service'. Implementing software-defined networks (SDN) instead of physical switches reduces hardware costs. SDN provides better control, scalability, and flexibility. Efficient SDN management is crucial for cost savings in cloud computing. Concerns with SDN include issues with data plane performance, ternary content addressable memory (TCAM), and controllers. Effective controller handling maximises throughput. Cloud computing generates high traffic volumes, with data and control traffics operating in different planes. Optimising traffic can enhance performance, and utilising queueing mechanisms can improve controller performance. Non-Markovian queueing models are proposed to handle heterogeneous buffers with varying traffic intensities. Models like M/M/1, M/M/n, M/M/s/n are used to manage traffic and improve controller performance. Both centralised and distributed systems are considered, taking into account behavioural distributions. Queueing models enhance QoS metrics, reducing waiting time, average queue intensity, and probability of idle server.