I am a postdoctoral researcher in the Computer Science Department at Carnegie Mellon University. I am advised by Professor Peter Steenkiste and I work with the XIA team on creating and evaluating future internet architectures. My research focuses on how to how to improve inter-domain traffic management.
My research interests span the field of computer systems, including cloud computing, distributed systems, file systems, networks, and operating systems. I am also interested in applying statistics, machine learning, and visualization techniques to solve systems problems.
I completed my Ph.D. in the Electrical & Computer Engineering department at CMU in May 2013. I worked at the Parallel Data Lab (PDL) and was advised by Professor Greg Ganger. My research focused on problem diagnosis tools for cloud environments.
In 2007, I appeared in a PhDComics strip telling CS grad students to wear lab coats to work. After all, how else will others know that we are researchers too? In my spare time, I enjoy photography, tennis, running, and taking naps.
My dissertation focused on a novel technique, called request-flow comparison, for automatically localizing the sources of performance degradations in distributed systems. Such changes are common and are extremely difficult to diagnose manually because the problem could be contained in any one of the distributed system's many components or, worse, may be a result of interactions among them. My dissertation combines systems research, statistics, machine learning, and visualization.
Request-flow comparison's key insight is that performance changes often manifest as changes or mutations in the workflow of how individual request's are serviced---i.e., in the components and functions they execute or in their detailed timing. Exposing these mutations and showing how they differ from previous behaviour localizes the source of the problem and gives developers a starting point for their diagnosis efforts. Request workflows are obtained via recently developed end-to-end tracing techniques (e.g., Google's Dapper).
To evaluate request-flow comparison's effectiveness, my dissertation describes how an implementation of it (called Spectroscope) was used to diagnose real, previously unsolved problems in a distributed storage system called Ursa Minor and in certain Google services. My dissertation also explores which visualizations are most useful for helping developers understand request-flow comparison's results. My dissertation also includes a design study that aims to identify the key design axes of end-to-end tracing and specific design points for various tracing use cases (e.g., anomaly detection vs. resource attribution).
- Bootstrapping evolvability for inter-domain routing. Raja R. Sambasivan, David Tran-Lam,
Aditya Akella, Peter Steenkiste. HotNets'15.
- Diagnosing performance changes by comparing request flows. Raja R. Sambasivan, Alice X. Zheng, Michael De Rosa, Elie Krevat, Spencer Whitman, Michael Stroucken, William Wang, Lianghong Xu, Gregory R. Ganger. NSDI'11.
[Abstract] [Paper] [Slides] [Talk video] [Code]
- So, you want to trace your distributed system? Key design insights from years of practical experience. Raja R. Sambasivan, Rodrigo Fonseca, Ilari Shafer, Gregory R. Ganger. CMU-PDL-14-102.
- Visualizing request-flow comparison to aid performance diagnosis in distributed systems. Raja R. Sambasivan, Ilari Shafer, Michelle L. Mazurek, Gregory R. Ganger. IEEE Transactions on Visualization and Computer Graphics (Proc. Information Visualization 2013), Vol. 19, no. 12, Dec. 2013.
[Abstract] [Paper] [Slides] [Video figure] [Code]
- Automated diagnosis without predictability is a recipe for failure. Raja R. Sambasivan and Gregory R. Ganger. HotCloud'12.
[Abstract] [Paper] [Slides] [Talk video]
- Ursa Minor: Versatile cluster-based storage. Michael Abd-El-Malek, William V. Courtright II, Chuck Cranor, Gregory R. Ganger, James Hendricks, Andrew J. Klosterman, Michael Mesnier, Manish Prasad, Brandon Salmon, Raja R. Sambasivan, Shafeeq Sinnamohideen, John D. Strunk, Eno Thereska, Matthew Wachs, Jay J. Wylie. FAST'05.