Making intelligent data staging and storage placement (where and how to place the application's data) choices is vital for doing automated resource management for on demand computing. When the storage resources available for a distributed application span a wide geographical distance, under some circumstances, staging data near the application improves application performance. In such a scenario, the storage placement strategy affects an application's performance. However, the benefits in performance of a particular staging or placement strategy is dependent on the computational, network, and storage resources available to the application and more importantly, the application itself.
In this project, we plan to explore a model based approach for addressing
the problem of intelligent data staging and storage placement. Our
approach is to capture the key parameters in terms of a model, and find an
informed mapping of the workloads to the storage infrastructure available
to the application. In addition, finding the right framework which
incorporates a model based approach is another research challenge which we
plan to address through this work. Our framework will initially focus on
performance goals and at a later stage we will extend it for capturing
performability, which characterizes availability along with the
performance.