Master of Science in Data Science (MSDS)
Network Slicing, Machine Learning, Blocking Estimation
With the growing popularity of 5G networks and their vast use cases, users must be provided with the services they request. Network slicing is a technique for establishing numerous distinct logical and virtualized networks over a shared multi-domain infrastructure. In particular, the emerging elastic slicing, which allows users to scale up or down the reserved amount of resources during the slice life cycle, is a promising technique to accommodate diverse network services more affordably. However, users may sometimes need more resources, which is impossible to provide due to the abundance of users and lack of additional resources. To our knowledge, no mathematical measure quantifies the level of resource isolation. Therefore, we propose an estimation technique that quantifies the degree of isolation to determine the possibility of blocking scale-up in the future. The goal is to determine request blockage, which indicates the blocking of scale-up requests. As slices keep moving in and out of the system, we improve the estimation of request blockage using the observed blocking incidents. We train multiple machine learning models to accurately estimate the request blockage, improving the estimation of blocking scale-up requests and comparing their performances. In this work, we tested the performances of Naive Bayes, Decision Trees, Random Forests, Gradient Boosting, and neural networks. This solution further enables us to solve the admission control problem to avoid accepting user requests that cannot be processed. Also, it can prevent the network provider from paying penalties by not accepting requests when future scale-up requests are likely to be blocked.
Movva, Nitin Datta, "On Measuring the Degree of Resource Isolation in Elastic Network Slicing" (2023). Master's Projects. 1341.
Available for download on Wednesday, January 01, 2025