State of Practice of Non-self-aware Virtual Machine Management in Cloud Data Centers

Abstract

Hardware virtualization is the prevalent way to share data centers among different tenants. In this chapter, we present a large-scale characterization study that aims to better understand the state of the practice, i.e., how data centers in the private cloud are used by their customers, how physical resources are shared among different tenants using virtualization, and how virtualization technologies are actually employed. Our study focuses on IBM corporate data centers as a major infrastructure provider and reports on their observed usage across a 19-day period. We aim at answering two key questions in virtual machine management: (i) whether the VM are dynamically adjusted according to the system state given the high flexibility in resizing virtual resources, and (ii) what type of self-learning policy governing the VM migration can be found in real data centers given a plethora of studies on VM migration. Our study illustrates that there is a huge tendency in over-provisioning resources while being conservative to several possibilities opened up by virtualization (e.g., migration and co-location), indicating the lack of autonomic VM management and a great potential for developing self-learning systems.