A Nested Decomposition Model for Reliable NFV 5G Network Slicing
Brigitte Jaumard, Quang Duong · IEEE Transactions on Network and Service Management · 2023
With the 5th generation of mobile networking (5G) on our doorstep, optical network operators are reorganizing their network infrastructure to deploy different topologies (virtual networks/applications) on the same network infrastructure on demand. This new paradigm, called network slicing, can be enabled by segmenting the network resources based on the requirements of the application level. This paper investigates a nested decomposition mathematical modeling to design a reliable 5G network slicing problem, i.e., every virtual path is protected against any single link failures by a dedicated backup disjoint virtual path. This new modeling revises and improves some previously proposed decomposition models. Then, we propose a column generation algorithm to solve the new modeling exactly. Moreover, this paper provides the computation of dual bounds with Lagrangian relaxation to assess the solutions’ accuracy that many existing nested-decomposition applications have omitted. Extensive computational results show that we can get$\varepsilon $-optimal reliable 5G slicing solutions with small$\varepsilon $(about 2% on average) in fairly reasonable computational times. In addition, we also propose several acceleration schemes using parallel programming to reduce computational time. The experimental results show that the slice-based scheme outperforms the path-based one in terms of parallelism.