Valentina Salapura

Keynote speaker for Fri, Oct 30 (tentatively from 9:30 AM -- 11:00 AM)

Title of the talk: To be decided

Abstract: Coming soon

Dr. Valentina Salapura is an IBM Master Inventor and System Architect at the IBM T.J. Watson Research Center. 

Valentina is with the IBM Research in the Services Innovation Lab where she is helping IBM realize the value of cloud computing.  In 2010, Dr. Salapura served as a lead for the Global Technical Outlook with the IBM Research Strategy and Worldwide Operations team to define IBM’s future research agenda and strategy working with the worldwide IBM research organizations. Previously, Valentina served as architect for Power Systems building workload-optimized systems for a Smarter Planet with a focus the processing unstructured data and business analytics. Valentina has been a technical leader for the Blue Gene program since its inception where she has contributed to the architecture and implementation of the BlueGene/Q, BlueGene/P, BlueGene/L and Cyclops systems. Valentina made seminal contributions to multiprocessor-based network architectures, power/performance characterization of a computer system, and emulation of microprocessors using FPGAs.  Before joining IBM Research in 2000, Dr. Salapura was a faculty member with Technische Universität Wien, where she also received her Ph.D. degree.

Valentina Salapura is recipient of the 2006 ACM Gordon Bell Prize for Special Achievements for the Blue Gene/L supercomputer and Quantum Chromodynamics. Dr. Salapura is the author of over 60 papers and several book chapters on processor and network architecture, and holds over 80 patents in this area. Dr. Salapura is a Fellow of the IEEE, and a Member of IBM Academy of Technology.

Jayant Haritsa

Keynote speaker for Sat, Oct 31 (tentatively from 9:30 AM -- 11:00 AM)

Title of the talk:   Plan Bouquets: A Fragrant Approach to Robust Query Processing

Abstract:  Declarative query processing with performance guarantees has been a highly desirable but equally elusive goal for the database community over the last five decades. The difficulty stems from two primary sources: errors in the cost models of the execution operators, and errors in the selectivity estimates that serve as inputs to these models. While the former error, which depends on the underlying computing environment, can be curbed to a fair degree, the latter is much harder to control since it is based on data distributions and correlations, which can be arbitrarily complex in nature. The net result is poor query execution plan choices, leading to grossly inflated and extremely unpredictable response times.  

In this talk, we present a conceptually new approach to address the selectivity estimation problem, wherein this process is completely eschewed for error-prone selectivities. Instead, a small "bouquet"' of plans is identified from the set of optimal plans in the query's selectivity error space, such that at least one among this subset is near-optimal at each location in the space. Then, at run time, the actual selectivities of the query are incrementally "discovered" through a sequence of partial executions of bouquet plans, eventually identifying the appropriate bouquet plan to execute. The duration and switching of the partial executions is controlled by a graded progression of isocost surfaces projected onto the optimal performance profile.  We prove that this construction results in bounded overheads for the selectivity discovery process and consequently, guaranteed worst-case performance. In addition, it provides repeatable execution strategies across different invocations of a query.

The plan bouquet approach has been empirically evaluated on both PostgreSQL and a commercial DBMS, over the TPC-H and TPC-DS benchmark environments. Our experimental results indicate that, even with conservative assumptions, it delivers substantial improvements in the worst-case behavior, without impairing the average-case performance, as compared to the native optimizers of these systems. Moreover, it can be largely implemented using existing optimizer infrastructure, facilitating easy incorporation in current database engines. Overall, the bouquet approach provides novel guarantees that open up new possibilities for robust query processing. 

[This is joint work with my PhD student, Anshuman Dutt.]

Jayant R. Haritsa is a senior professor of database systems in the Supercomputer Education & Research Centre and the Department of Computer Science & Automation at the Indian Institute of Science, Bangalore. He received a BTech degree from the Indian Institute of Technology (Madras), and the MS and PhD degrees from the University of Wisconsin (Madison). He is a Distinguished Scientist of ACM and a Fellow of IEEE. He has been elected a Fellow of Indian National Academy of Engineering, Indian Academy of Sciences, and National Academy of Sciences, India.  

Jayant Haritsa is the recipient of the ACM Infosys award, 2014.  Previously, his scientific contributions have been recognized through Shanti Swarup Bhatnagar Award, Swarnajayanti Fellowship, ACCS-CDAC Foundation Award, Dr.  Vikram Sarabhai Research Award, Sir C V Raman Young Scientist Award, Distinguished Alumnus of IIT-Madras Award, and IISc Alumni Research Excellence Award. He has served on the editorial boards of IEEE Trans. on Knowledge & Data Engineering, Very Large Data Base Journal, Proceedings of VLDB Endowment, Journal of Distributed & Parallel Databases, and Journal of Real-time Systems. He has been a Program Chair for the VLDB, ICDE, DASFAA and COMAD international database conferences.