Affinity and Conflict-Aware Placement of Virtual Machines in Heterogeneous Data Centers

Virtual machine placement (VMP) problem has been a key issue in IaaS/PaaS cloud infrastructures. Many recent works on VMP prove that inter-VM relations such as memory share, traffic dependency and resource competition should be seriously considered to save energy, increase the performance of infrastructure, reduce service level agreement violation rates and provide better administrative capabilities to the cloud provider. However, most existing works consider the inter-VM relations without taking the heterogeneity of cloud data centers into account. In practice, heterogeneous physical machines (PM) in a heterogeneous data center are often partitioned into logical groups for load balancing and specific services, cloud users always assigned their VMs with specific PM requirements, which make the inter-VM relations far more complex. In this paper, we propose an efficient solution for VMP with inter-VM relation constraints in a heterogeneous data center. The experimental results prove that our solution can efficiently solve the complex problem with an acceptable runtime.