Abstract

Network virtualization technologies allow 5G and beyond mobile networks to support new vertical applications of various service types. Network slicing (NS) further provides a concept for multi-tenant access to common network infrastructure spreading from core and transport to radio access networks. Nevertheless, the management and orchestration of network slices across multiple infrastructure and administrative domains have only recently gained attention. In this work, we first present a reference architecture for NS management and its brokering layer that enables NS discovery and dynamic composition. We detail the designs of the main NS brokering components, i.e., NS information model and NS request handling, which address some unique challenges of multi-domain NS brokering. Our approach combines machine learning and logic reasoning for the selection of NSs based on their functional, non-functional attributes and other policy constraints. Finally, we validate the NS request matching approach and observe a high prediction accuracy. We conclude the work with a discussion on possible improvement to the NS matching approach and its extension for cross-domain NS management.

@INPROCEEDINGS{9059366,
  author={X. {Dang} and F. {Sivrikaya}},
  booktitle={2020 23rd Conference on Innovation in Clouds, Internet and Networks and Workshops (ICIN)}, 
  title={A Lightweight Policy-aware Broker for Multi-domain Network Slice Composition}, 
  year={2020},
  pages={123-130},}