A statistical approach to virtual server resource management

Abstract

Resource management is of key importance in a wide array of computer and network environments. Failure to effectively allocate resources necessary for smooth infrastructure operation may result in impediments in successful service provisioning. This is especially prominent in environments whose workloads present fluctuating resource demand, such as in Cloud‐based infrastructures. The mathematical science of statistics offers an extensive theoretical and practical toolbox that can be of potential use in tackling such problems in the information technology domain. We propose a novel approach to resource management of virtualized data center components. We have worked toward creating a mechanism that allows for dynamic resource allocation, adapting to the changing demand patterns. Our system incorporates a resource controller, based on Statistical Process Control, which permits the online management of a virtual machine's processor and memory capacity through real‐time analysis of its observed performance. We successfully demonstrate our approach, with real data, on an architecture running the core banking environment of a financial institution. The controller successfully manages the resource allocation of the virtual machine and stabilizes its performance while business workloads are being processed. Our approach can be extended to realize different workload models and to manage other types of hypervisor provisioned resources.