Development of an Intent-Based Network Incorporating Machine Learning for Service Assurance of E-Commerce Online Stores

Managing the large number of network devices is a major challenge for organizations. The management of these devices requires meticulous care since manual configurations of traditional networks often lead to errors. Manual configurations result in inefficient troubleshooting. Instead of worrying about the development of new services, the organizations’ staff is exhausted by the arduous task of maintaining network devices. Specifically, one of the organizations’ core services is their e-commerce service available through their website, through which they present and sell products. Companies that have had a great growth require optimizing the use of CPU, memory, bandwidth resources, among others, in order to provide an available and scalable service for the large number of customers. Due to this, this paper proposes: 1) an Intent Based Network (IBN) based on Cisco’s IBN architecture applied to online stores in e-commerce, 2) a scalable and efficient method that uses Neural Networks (NN) for resource prediction in order to adapt to the large number of customers and the growing demand in this era of Big Data. The IBN architecture provides automation, flexibility and unlike Software Defined Networks (SDN) includes the integration of artificial intelligence and data mining techniques in order to analyze future trends and problems for network service assurance. The neural network predicts the appropriate bandwidth value based on the trend of the number of users that the online store server will consume and other resources such as CPU, memory and storage usage. This value is automatically configured in the network, so that the service always has a high availability. We have created a dataset by collecting data from the network we have developed. We have defined a base neural network model and subsequently optimized this model by minimizing the Mean Absolute Error (MAE). We have shown that the optimized neural network outperforms other machine learning methods such as Random Forest (RF), Support Vector Machine (SVM) and K Nearest Neighbor (KNN). This research opens the door to the development of IBNs incorporating artificial intelligence for e-commerce and other services.