Predicting application performance for multi-vendor clouds using dwarf benchmarks

Abstract

Future Internet applications are becoming increasingly dynamic and can be composed of a wide range of services controlled and hosted by different stakeholders. This paper addresses the challenge of resource provisioning for applications that have specific Quality of Service (QoS) requirements and where consumers of Cloud resources want to avoid lock-in to any specific Infrastructure-as-a-Service (IaaS) provider. Application modelling can be used to predict performance of applications given certain resources, workload and configuration. However, application modelling is a significant challenge for Cloud consumers due to the limited and varying information IaaS providers disclose about infrastructure resources. We demonstrate in this paper how Dwarf benchmarks can be used as a uniform and informative way of characterising compute resources, which is successful for application modelling, achieving high prediction accuracy on a range of applications. © 2012 Springer-Verlag.