CAREER: Heterogeneous Elastic Computing over the Cloud - from Theory to Practice

Modern cloud-services rely on virtual machines due to their high efficiency and flexibility. Alongside regular services, current new offerings from cloud-service providers allow exploiting under-utilized virtual machines at a fraction of the original cost. Such computing resources, however, are highly heterogeneous and elastic. Heterogeneity means that virtual machines can have different computational speeds and storage constraints. Elasticity means that these virtual machines can be preempted under short notice (on the order of minutes) if a high-priority job appears; on the other hand, new virtual machines may be available over time to compensate for any shortage of computing resources. Such behavior can result in computational failure or significantly increase computing time. In response to the challenges of both heterogeneity and elasticity in cloud systems, this project will formulate new heterogeneous elastic computing frameworks, aiming for optimal and implementable solutions. The results of this project can lead to sizable economic benefits. The project's educational activities are designed to integrate research into teaching, and include mentoring both graduate and undergraduate students, alongside outreach programs to undergraduate and K-12 students with the goal of fostering interest in transformational computing technologies. Another highlight of this project is the newly developed YouTube channel by the investigator. Motivated by practical measurements and constraints, this project formulates new heterogeneous elastic computing frameworks with both coded and uncoded storage placements. Then, it develops novel methodologies using combinatorial and information-theoretic tools to establish fundamental tradeoffs for such systems. Using these theoretical tools, the project designs low-complexity algorithms for real applications and evaluates them on Amazon Elastic Compute Cloud (Amazon EC2) in order to show significant gains of the proposed approaches compared to the state-of-the-art solutions. The project is structured around research topics: (1) heterogeneous coded storage elastic computing; (2) heterogeneous uncoded storage elastic computing; (3) secure uncoded storage elastic computing from user’s perspective and (4) heterogeneous elastic computing: convert theory to practice. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.