VRMap: A Cost and Time Aware Remapping of Virtual Data Centres Over a Geo-distributed Infrastructure

A virtual data center (VDC) comprises multiple correlated virtual machines (VMs) with communication dependencies. Provisioning VDC requests across Geo-distributed data centres (DCs) offer benefits to service providers (SPs) in the form of improved revenue, enhanced performance and increased survivability. On the way to achieve such salient objectives, SPs often face the concomitant challenge of dynamic reconfiguration/remapping of multiple VDCs. Remapping VDCs introduces two arduous challenges: (A.) finding a minimum cost remap for VMs' and virtual links (VLs), (b.) migrating VMs image and memory contents across DC networks with minimum overheads and service disruption. In this paper, we propose a model called VRMap that finds an optimal remapping plan that aims to balance the remapping cost and migration overheads without compromising the service quality. VRMap makes use of genetic-metaheuristic to achieve the optimal solution. The performance of VRMap is compared with greedy and random heuristics. Simulation results show that VRMap achieves comparable performance with regards to remapping cost but attains significant gain in terms of miaration time and downtime.