Dynamic Scheduling of Virtual Machines
Flavien Quesnel · 2014
This chapter shows several approaches to schedule Virtual Machine (VMs) dynamically. It focuses on fault-tolerance and scalability, in other words, the ability to efficiently manage infrastructures composed of many nodes and VMs. The chapter recalls some design choices that have an impact on scheduler architectures, especially regarding monitoring and decision-making. It presents the limits of a centralized approach, which cannot quickly solve quality of service violations, so that a reconfiguration plan can be outdated before being fully applied. The chapter addresses multiagent approaches, which distribute the scheduling work over all nodes composing the infrastructure. An ideal decentralized approach to schedule VMs is to be able to manage overbooked infrastructures and extremely reactive to solve overload problems.