IBM's research in "Predictive Elasticity in Clouds" is aimed at improving the performance, while lowering the cost, of cloud-hosted applications by deriving a better initial allocation of resources and placement of the application.
Typically before an application is deployed on a cloud, the Application Operator needs to make a decision regarding the number of virtual resources to allocate to the application as well as decide upon a billing model with the Cloud provider, such as pay-on-demand or reserved VMs. This work extends CloudWave by first using CloudWave monitoring data to analyze the historical behavior of the application (or of another application which is similar), and then helping determine the initial deployment conditions for CloudWave's Living State Manager. Additionally, this work can be used to predict long term costs for proposed adaptation actions, and can be used as sub-Adaptation Engine (i.e. a type of Recommendation Engine).