Workload analysis and demand prediction for the HP ePrint Service

摘要

The proliferation of service oriented computing boosted by advances in cloud computing has led to design of new services that combine the power of both trends. The HP ePrint Service allows customers to print from anywhere to an HP ePrint-enabled printer accessible via the Internet. As ePrint is a hosted service, it must provide customers with a high quality of service while keeping the costs of supporting the load as low as possible. Understanding the load and the nature of this new service is crucial for properly designing the service's support infrastructure under rapidly growing user demand. With the complexity of services increasing and application requirements for QoS growing, the research challenge is to design an integrated framework of workload analysis combined with system measurement and modeling techniques that support performance analysis for the service. In this work, we present a detailed workload analysis of ePrint and a new performance tool for the automatic evaluation of required capacity for processing its diverse workload in a production environment while satisfying QoS requirements.