I'd like to introduce my research work done in this summer, "Metis: Online Multi-Resource Allocation for Network Slicing in 5G"
With a radical transformation towards an expanding mobile cellular network, the infrastructure of mobile networks in 5G will provide diverse services in a form of network slices which can be deployed and implemented in a customizable way. Prior work fails to consider efficient resource allocation in a dynamic real network with time-varying network utility. In this paper, we initialize the study of the multi-resource allocation problem for network slicing in an online manner, where the utility functions vary with the time. We present Metis, an online network slice resources allocation framework that combines the time-varying property of network utility function. Given the bandwidth and processing capacity constraints and the virtual network functions (VNF) isolation requirements, we aims to maximizing the cumulative network utility in a long term. We formulate the multi-resources allocation problem as a concave optimization program. Further, we develop a set of distributed algorithms based on online ADMM to solve it. Regret analysis shows that Metis can provably converge to the optimal solution. Large-scale simulations show this convergence property in dynamic network settings.
Presenter: Dongyue Li
Time: Begins at 18:00, Wednesday, 5/12/2018
Welcome to attend!