The Framework Using Rerouting and LSTM to Keep an Eye on the 5G Backhaul Network

Scalability in 5G as well as Multi-access Edge Computing systems may be achieved with the help of SDN. It has been recommended to migrate the data plane elements to share the workload of SDN controllers more evenly. To enable network flows that demand low latency communications, we suggest a proactive load balancing strategy on the SDN control plane, as opposed to the reactive procedures used by most prior research. In order to avoid load imbalances and prepare ahead for data plane migrations, we first predict the demand of SDN controllers. Second, we improve load balancing while adhering to delay limitations by optimising the migration procedures. They are both initially routed through distinct techniques, and the need of activating the reroute mechanism is determined by calculating the movement value of the difference in link utilisation. To anticipate the workload on SDN controllers, we first build two prediction models using LSTM (Long Short-Term Memory) techniques. Then, we compute the accuracies and prediction errors of these two systems and compare them. The results show that compared to the ARIMA model, the LSTM framework is 55% more reliable when generating long-term forecasts. After recasting the issue as a non-linear binary algorithm and proving its NPcompleteness, we provide a reinforcement learning technique for deciding which data plane pieces need to be migrated and when, given constraints on acceptable latency. The simulation findings show that the suggested method is quite close to ideal.