The cloud market should make computing accessible to more users and promises unprecedented flexibility for their applications. These promises fall short, because cloud computing remains too complex for most users: biologists, physicists, astronomers, financial-experts, or data-analysts have to choose (before execution) the cloud computing resources (CPU types, CPU numbers, memory, disk space, network bandwidth) suitable for their applications.
My goal with this project is to break this usability barrier, and provide a system to guide users when “shopping" in the cloud computing market. Essentially, I will provide cloud computing a-la-carte: users specify desired application behaviour, and CloudIA provides them with a customized cloud deployment configuration.
Surprisingly, cloud providers have not already solved this problem, although they should tailor to their customer needs to survive. Actually, providers rely on the naivety of users, and offer pre-packaged, typically under/over-provisioned services. In turn, users have to shrink/expand their needs to fit these packages, an "adaptation" leading to losses in performance and/or money.
With different requirements from application to application and from execution to execution, the pre-packaging approach does not scale. Thus, CloudIA will embed methods for systematic translations of requirements into feasible cloud services offerings. I will (1)develop new methods to automatically extract resource requirements as a function of the user’s performance expectation, (2)design mechanisms to construct (near-)optimal sets of deployment solutions (i.e. solving the trade-off between the resources required to maintain the expected level of performance and their costs), (3)show how dynamic cloud computing service offerings, needed to address the dynamic requests from the users, will change the cloud providers market (for the better).
My results will have a strong impact on both cloud users, by simplifying their interaction with the providers, and cloud providers, by creating a dynamic view of cloud computing commodities, leading to a more competitive cloud computing market.
Faculty Electrical Engineering, Mathematics and Computer Science, TU Delft - Dr. Alexandru Iosup: here
Unfortunately, the proposal was not awarded a VENI grant. However, you may contact me (a.m.oprescu@uva.nl) if you would like to learn more about its technical content.