Prediction & Adaptation: System performance is heavily dependent on the nature of the workload it encounters. Instead of optimizing for a particular workload, we focus on building access predictors that exploit repeatable data access behavior. By implementing efficient access predictors, and combining simple predictions to construct more elaborate sets of likely future events, we improve storage system performance and automate data management algorithms ... more

Energy: Energy is fast becoming a critical resource and with the explosive growth in digital data volumes storage and storage servers are increasing their energy demands to critical levels. Through workload prediction and data replication and grouping we aim to improve data access performance while simultaneously reducing the energy consumed in data storage and retrieval ... more

Reliability: Our most recent efforts have focused on novel schemes to improve data storage reliability, which is an increasingly important goal with the explosion in data volumes. Our approach focuses on the construction of computationally efficient schemes that offer customizable levels of fault tolerance while making the most practical use of available resources ... more


Storage systems, file systems, distributed systems, operating systems - systems: the automated management of resources with the goal of increasing benefit and reducing cost. My main research area lies in storage systems management and mechanisms that enable storage systems to perform well in the face of varying workloads and technology trends. To this end I have storage projects falling under the following three main themes: Prediction, Energy, and Reliability.