Energy-Aware Virtual Machine Management in Inter-Datacenter Networks Over Elastic Optical Infrastructure

Abstract

Datacenters (DCs), deployed in a large scale to support the ever increasing\ndemand for data processing applications, consume tremendous energy. Powering\nDCs with renewable energy can effectively reduce the brown energy consumption.\nOwing to geographically distributed deployment of DCs, the renewable energy\ngeneration and the data processing demands usually vary in different DCs.\nMigrating virtual machines (VMs) among DCs according to the availability of\nrenewable energy helps match the energy demands and the renewable energy\ngeneration in DCs, and thus maximizes the utilization of renewable energy. We\nfirst elicit the renewable energy-aware inter-datacenter (inter-DC) VM\nmigration problem in an inter-DC network over the elastic optical\ninfrastructure, present it as a many-manycast communications problem, and then\nformulate it as an integer linear programming problem. The objective is to\nminimize the total cost of the brown energy consumption of DCs in such inter-DC\nnetwork via VM migration. We use CVX and Gurobi to solve this problem for small\nnetwork configurations, and we propose a few heuristic algorithms that\napproximate the optimal solution for large network configurations. Through\nextensive simulations, we show that the proposed algorithms, by migrating VM\namong DCs, can reduce up to 19.7% cost of the brown energy consumption.\n