Traffic-aware virtual machine placement in geographically distributed Clouds

Abstract

In this work, we focus on the problem of virtual machines (VMs) placement in geographically distributed data centers, where tenants may require a set of networking VMs. The aim of the present work is to plan and optimize the placement of tenant's VMs in a distributed Cloud environment while considering location and system performance constraints. Thus, we propose ILP formulations which have as objective the minimization of traffic generated by networking VMs and circulating on the backbone network. The different experiments conducted on the proposed formulations show the effectiveness of our model for large-scale Cloud systems in terms of running time and computational resources.

Network-aware virtual machine placement in cloud data centers with multiple traffic-intensive components

“…Teyeb et al [7] study VM placement problem in geographically distributed data centers with tenants requiring a set of networking VMs. In their work, an ILP formulation of the placement problem is provided that takes location and system performance constraints into account.…”

“…The mentioned work employs the simplified form of the formulation for Hub Location problem discussed in [8] to find an optimal placement. Teyeb et al [7] make some assumptions about the distributed data centers that seem unrealistic. For example, by assuming that the distributed data center parts do not send traffic to each other they simplify the problem.…”

“…Similarly, [14] is another work (very closely related to [7]) that studies VM placement in geographically distributed data centers. The mentioned work aims to minimize IP-traffic within a given backbone network by placing VMs in data centers that are connected over an IP-over-WDM network.…”