Raja Sambasivan is the Ankur and Mari Sahu Assistant Professor in the Department of Computer Science at Tufts University. Previously, he was the Red Hat Visiting Scientist at Boston University’s Mass Open Cloud, where he led a research group that focused on creating problem diagnosis tools. He completed his dissertation (on diagnosis tools) and postdoctoral research (on incentive-compatible mechanisms for deploying new inter-domain routing protocols) at Carnegie Mellon University. During his career, Sambasivan has worked on a wide range of technologies related to the cloud ecosystem, including object-based storage, inter-domain routing, future Internet architectures, and big-data frameworks.
Ankur and Mari Sahu Assistant Professor, Department of Computer Science, Tufts School of Engineering
Red Hat Visiting Scientist, Mass Open Cloud
Postdoctoral Researcher, Carnegie Mellon University
Ankur and Mari Sahu Assistant Professor Raja Sambasivan’s research supports innovation in the cloud ecosystem — i.e., cloud data centers, distributed systems within data centers, and wide-area networks that connect data centers.
His research involves creating sophisticated debugging tools for distributed systems, as the immense difficulty of diagnosing problems dis-incentivizes innovation. It also involves building distributed systems that are easily evolvable to support innovation. Sambasivan's current research efforts focus on: 1) Creating an instrumentation framework for distributed systems that automatically searches the space of possible instrumentation choices to enable that needed to provide visibility into a new problem and 2) Creating powerful abstractions for diagnosing problems in distributed systems that are at least as powerful as the API-based abstractions developers use to build them.
In the past, Sambasivan and his collaborators created D-BGP, a version of BGP augmented with evolvability features to support the deployment of new inter-domain routing protocols. He also developed one of the earliest distributed-tracing infrastructures strongly suited for performance diagnosis tasks and a diagnosis toolkit called Spectroscope that builds on tracing to diagnose performance degradations in distributed systems.