Profiling the execution phases of an application can lead to optimizing the utilization of the underlying resources. This is the thrust of this paper, which presents a novel system-level application resource demand phase analysis and prediction prototype to support on-demand resource provisioning. The phase profile learned from historical runs is used to classify and predict phase behavior using a set of algorithms based on clustering. The process takes into consid- eration application?s resource consumption patterns, pricing schedules defined by the resource provider, and penalties associated with Service-Level Agreement (SLA) violations.