Automating networks through the machine learning driven insight

Many elements of infrastructure have the capability to adapt with principles such as scale on demand. Networks have generally stayed static due to their complexity and fixed nature of the underlay provider and the ability to perform consistent rapid change. In recent times Artificial Intelligence has been integrated in almost all elements of information and communication technologies with the use of machine learning and deep learning becoming quite commonly applied in various day to day applications. The majority of network components produce large amount of raw data much of which currently is ignored or discarded in favour on focusing on availability and key performance metrics. Machine learning algorithms have been developed for communication networks to be applied to fundamental problems in networking, including traffic prediction, routing and classification, congestion control, resource, fault management, QoS and QoE management and network security. Software defined networks (SDN) provide the ability to program network configuration through the use of open standards and DevOps tools such as Terraform and Ansible. Central orchestration engines provides the ability to deploy configuration to the edge through a controlled and consistent method. So, we have the building blocks for an adaptable network, driven by metrics provided on a real or near real time basis. Key Takeaways will be: • Generation of real time metrics to provide insight in the performance of a network for end user experience • Consumption of metrics to be processed by the appropriate machine learning algorithm with validation of learning by network professionals • Dynamic configuration of networks through programmatic interface centrally orchestrated • An iterative programmatic approach to continually change with appropriate lag to continually adapt to the needs of the business applications and introduction of new services