Yale Computer Engineering Seminar Series
Dr. Fabrizio Petrini, Technical Lead, Data-Centric Systems, IBM TJ Watson Research Center
Title: Algorithms and Architectures for Large Scale Data Analytics
Abstract: The nature of big data, often consisting of connections, relations, and interactions among entities, frequently lends itself well to graph-based models. The fields in which graphs are proving themselves as a natural modeling abstraction cover a large number of application areas, such as biology, transportation, Internet and power grids, communication data, social networks and, more generally, various forms of relational data. With huge data sets, many of the locality assumptions current architectures are optimized for are not satisfied anymore. This has profound consequences, both in terms of algorithmic and of architectural design.
In this talk we will present some of our latest results on large scale graph algorithms, executed on machine configurations with up to 1.5 million cores and 6 million threads, and new architectural solutions that are rapidly appearing on the horizon to efficiently solve big data problems.
Bio: Fabrizio Petrini is a technical leader of the Data Centric Systems department of IBM TJ Watson Research Laboratory. His research interests include various aspects of multi-core processors and supercomputers, including high-performance interconnection networks, network interfaces, fault tolerance, and data-intensive computing algorithms for mining large data sets. Before joining IBM TJ Watson, he was a laboratory fellow at the Pacific Northwest National Laboratory, a Member of the technical staff at Los Alamos National Laboratory, a research fellow at Oxford University in the UK and a postdoc at the University of California at Berkeley. Petrini received a Laurea Degree and a PhD in computer science from the University of Pisa, Italy. He served as associate editor of IEEE Transactions on Parallel and Distributed Processing. He is the recipient of numerous awards from DOE, IEEE and ACM, including four best paper awards from the main conferences in parallel computing. He coordinated the seven winning IBM Graph500 submissions from November 2010 to November 2014.