Towards Efficient Large-Scale Network Slicing: An LP Dynamic Rounding-and-Refinement Approach

In this paper, we propose an efficient algorithm for the network slicing problem which attempts to map multiple customized virtual requests (also called services) a common shared infrastructure and allocate resources meet diverse service requirements. The has been formulated as mixed integer linear programming (MILP) formulation in literature. We first novel (LP) relaxation of MILP formulation. show that compared with natural LP formulation, is much more compact terms smaller numbers variables constraints, stronger providing better bound, makes it particularly suitable be embedded based algorithm. Then design two-stage dynamic rounding-and-refinement on relaxation. stage, proposed uses rounding procedure place functions all services into cloud nodes while taking traffic routing consideration; second iterative refinement obtain solution their end-to-end delay constraints being satisfied. Compared existing algorithms either have exponential complexity or return low-quality solution, our achieves trade-off between quality computational complexity. particular, worst-case polynomial, solving large-scale problems. Numerical results demonstrate effectiveness efficiency