CS Talk - Raja R. Sambasivan, Boston University
Host: Minlan Yu
Abstract
Cloud computing has rapidly emerged as the dominant computing platform of the 21st century. Its rise has also brought significant issues to the fore.Two key ones are the following. 1) The immense difficulty of diagnosing problems, both within distributed applications running in the cloud, and within the cloud infrastructure itself. 2) The Internet’s unsuitability for the needs of clouds. In this talk, I will discuss my PhD and postdoc research, which take strides toward addressing these challenges.The former focuses on a novel technique, called request-flow comparison, for automatically localizing the source of performance degradations within distributed systems and cloud infrastructure services. Request-Flow comparison has proven useful in helping solve new, previously unsolved problems both within a prototype distributed storage system at CMU and within certain services at Google.The latter focuses on creating mechanisms that allow BGP, today’s de-facto Internet routing protocol, to be easily replaced with more advanced protocols.
Bio
Raja is a visiting scientist at Boston University, working on problems related to cloud computing. Previously, he completed his postdoc and PhD at Carnegie Mellon University, where he worked with Peter Steenkiste and Greg Ganger respectively. His postdoc focused on evolvability for routing and his PhD focused on automating diagnosis of performance problems in cloud environments. In his spare time, he enjoys playing tennis.