The Static Virtual Machine Placement and Routing Spectrum Assignment for Multi-tenant in Elastic Optical Networks

Abstract

The static routing spectrum assignment is a fundamental problem in typical elastic optical networks, where the traffic demands are given and stationary. However, an elastic optical data center network is shared by multi-tenant, and hence the traffic demands are varying and determined by the virtual machine (VM) placement of multi-tenant under the given VM demands. In this paper, the problem of virtual machine placement and routing spectrum assignment (VMPRSA) is introduced and the static problem is proved to be NP-Hard. Furthermore, an optimal integer linear programming (ILP) model is formulated, with the target of minimizing the spectrum used to serve all the traffic demands driven by multi-tenant VM placement. Since the ILP model cannot scale to the big networks, a two-tier heuristic algorithm framework is proposed, i.e., the first tier is VM placement and the second is routing spectrum assignment (VMP + RSA). The VM placement of tenants can produce traffic demands, it thus is focused and two VM placement algorithms are proposed, namely random placement (RP) algorithm, and residual node capacity priority (RNCP) algorithm. The simulation results indicate that the ILP model provides the optimal solution, and the RNCP algorithm yields the better sub-optimal solution.