Delimitrou receives Facebook Hardware & Software Systems Research Award

Cornell ECE Assistant Professor Christina Delimitrou has recently been awarded a Facebook Hardware & Software Systems research award for her work on using online learning to improve Resource Efficiency in interactive cloud microservices. 

According to an announcement made by Facebook, "Continued research into hardware and systems is essential to Facebook as we develop algorithms to maximize impact and every day experiences. By sponsoring research, we extend our knowledge and share findings. We are especially interested in collaborating and sponsoring research at the intersection of computer systems and machine learning..."

"The award recognizes our work on using scalable machine learning techniques to make the management of complex, emerging cloud applications like microservices practical and effective," said Delimitrou.

Microservices have grown in popularity over the past few years as a way to manage the complexity of cloud services, as well as a way to facilitate and accelerate deployment, responsiveness, isolation and elasticity. Under this model the functionality of a large cloud service is decomposed to hundreds or thousands of tiny, loosely-coupled services, each with a very specific operation. Despite their benefits, microservices complicate aspects of cloud management like scheduling and resource allocation due to the dependencies that connect different microservices with each other. Because of these dependencies, making wrong scheduling decisions, even for a small subset of active microservices can easily lead to poor quality of service for the overall application, as well as resource and energy inefficiency for the system. Unfortunately managing such complex applications manually or even empirically becomes intractable as they increase in size and number. 

Delimitrou's work will instead leverage the massive amount of data cloud systems already collect to allow the system to automatically learn how many resources each microservice needs and adjust these decisions as applications, user demand, and hardware systems change. 

"Given the increasing complexity of cloud infrastructures and applications, such data-driven approaches offer practical solutions in the way we design and manage future systems whose scale renders previous approaches impractical," she said.

Delimitrou is an assistant professor in the School of Electrical and Computer Engineering at Cornell University, a member of the Computer Systems Laboratory (CSL), and the John and Norma Balen Sesquicentennial Faculty Fellow. Her main interests are in computer architecture and computer systems. Specifically, she works on improving the resource efficiency of large-scale datacenters through QoS-aware scheduling and resource management techniques. She is also interested in efficient server architectures design, serverless compute frameworks, resource disaggregation, and cloud security. Before joining Cornell, she earned a Ph.D. in Electrical Engineering at Stanford University, an M.S. in Electrical Engineering also from Stanford (2011), and a Diploma in Electrical and Computer Engineering from the National Technical University of Athens (2009). Find out more about her work at http://www.csl.cornell.edu/~delimitrou/.