An E2E Network Slicing Framework for Slice Creation and Deployment Using Machine Learning.

Area of Science:

• Telecommunications Engineering
• Computer Science
• Network Architecture

Background:

• 5G networks require flexible and dynamic features, achievable through network slicing.
• Network Function Virtualization (NFV) and Software-Defined Networking (SDN) are key enablers for network slicing.
• Network slicing allows for customized, end-to-end isolated services based on specific requirements.

Purpose of the Study:

• To develop a machine learning-based method for constructing network slices.
• To efficiently allocate resources to these newly created network slices using dynamic programming.
• To optimize key performance indicators (KPIs) such as user access rate and resource efficiency.

Main Methods:

• A substrate network was constructed, considering KPIs like CPU capacity, bandwidth, delay, link capacity, and security.
• Network slices were generated using a Multi-Layer Perceptron (MLP) with the ADAM optimization algorithm.
• Resource allocation was performed using Dijkstra's algorithm to find the shortest path, maximizing user access and resource efficiency.

Main Results:

• The proposed model effectively categorizes network slices for different services (mMTC, eMBB, uRLLC).
• Optimum network slices were allocated to requested services.
• The model demonstrated high resource efficiency and reduced total bandwidth utilization.

Conclusions:

• The developed approach successfully constructs and allocates resources for network slices in 5G networks.
• The integration of machine learning and dynamic programming leads to efficient resource management.
• This method offers a promising solution for enhancing 5G network performance and customization.