An Adaptive Utilization Accelerator for Virtualized Environments

Abstract

One of the key enablers of a cloud provider competitiveness is ability to over-commit shared infrastructure at ratios that are higher than those of other competitors, without compromising non-functional requirements, such as performance. A widely recognized impediment to achieving this goal is so called "Virtual Machines sprawl", a phenomenon referring to the situation when customers order Virtual Machines (VM) on the cloud, use them extensively and then leave them inactive for prolonged periods of time. Since a typical cloud provisioning system treats new VM provision requests according to the nominal virtual hardware specification, an often occurring situation is that the nominal resources of a cloud/pool become exhausted fast while the physical hosts utilization remains low.We present a novel cloud resources scheduler called Pulsar that extends OpenStack Nova Filter Scheduler. The key design principle of Pulsar is adaptivity. It recognises that effective safely attainable over-commit ratio varies with time due to workloads' variability and dynamically adapts the effective over-commit ratio to these changes. We evaluate Pulsar via extensive simulations and demonstrate its performance on the actual OpenStack based testbed running popular workloads.