Florin Rusu
Assistant Professor
School of Engineering
University of California, Merced
"GLADE: A Scalable Big Data Analytics System "

Tuesday, March 8, 4:00 PM
Packard Lab room 466

Abstract:   In this talk, we present GLADE, a scalable and efficient Big Data analytics system. GLADE is a multi-node multi-thread parallel system built around the user-defined aggregate (UDA) abstraction. It has a push-based relational columnar storage engine and provides runtime code generation for efficient query processing. Although necessary, these features alone are not sufficient for scalable Big Data analytics which requires more advanced methods, such as approximation and multi-query processing. Parallel online aggregation implemented in GLADE allows for approximate results with confidence bounds to be generated throughout the entire query processing, without increasing the overall execution time. Multi-query processing in GLADE goes beyond standard shared scans and intermediate result caching. Essentially, data access is shared throughout the entire memory hierarchy, from disk to CPU registers, across multiple queries. We provide a concrete example showing how these techniques are integrated into gradient descent optimization, the most popular method for training generalized linear models at terascale.

Bio:   Florin Rusu is an assistant professor in the School of Engineering at University of California Merced and a faculty scientist in the Scientific Data Management Group at Lawrence Berkeley National Lab. Florin's research interests lie in the area of databases and large scale data management in general, with a particular focus on designing and building infrastructure for Big Data analytics. Specific topics include approximate query processing, scientific data processing, and scalable analytics. Florin has designed and implemented several data processing systems over the past ten years, including DBO/TurboDBO, DataPath, GLADE, and EXTASCID. He has been an active member of the database research community, publishing articles and serving in the program committees of prestigious conferences and journals such as SIGMOD, VLDB, TODS, and TKDE. Florin is the recipient of a Hellman Faculty Fellowship in 2013 and a DOE Early Career Award in 2014.

© 2014-2016 Computer Science and Engineering, P.C. Rossin College of Engineering & Applied Science, Lehigh University, Bethlehem PA 18015.