Database Seminar Series (2006-2007)

Real-world data, especially that generated by distributed measurement infrastructures such as wireless sensor networks, tends to be incomplete, imprecise, and erroneous, and hence rarely usable in its raw form. The traditional approach to dealing with this problem is to first synthesize (filter) such data using a statistical or a probabilistic model, thus resulting in a more robust interpretation of the data. However current database systems do not provide adequate support for statistical modeling of data, especially when those models need to be frequently updated as new data arrives in the system. Hence most scientists and engineers, who depend on models for managing their data, do not use database systems for archival or querying at all; at best, databases serve as a persistent raw data store.

In this talk, I will present our approach to integrating statistical and probabilistic models into database systems, in the context of data management in wireless sensor networks. I will first present a data acquisition approach for wireless sensor networks that demonstrates how models can be used both to provide more meaningful answers to user queries, and to significantly reduce the energy cost of acquiring data from the underlying sensing devices. I will then present our recent work on the "MauveDB" system, which uses an abstraction called "model-based views" to seamlessly integrate models into traditional relational database systems.

The problem of data semantics is establishing and maintaining a correspondence between a data source (e.g., a database, an XML document) and its intended subject matter. We review the (relatively minor) role data semantics has played in Databases under the term "semantic data models", its more prominent place in ontology-based information integration, and then outline two new views: (i) Semantics as a composition of mappings between models, and (ii) Attaching intensional aspects (stakeholder goals) to Information Systems. In each case we consider the benefits of this view for the important problem of data integration/loading.

Joint work with John Mylopoulos and students (Univ. of Toronto)

Alex Borgida holds a PhD degree from University of Toronto, and is a Professor of Computer Science at Rutgers University, New Brunswick, NJ. His research is mainly concerned with knowledge representation and its applications. He has published in a variety of areas including Artificial Intelligence (description logics, explanation), Databases (exceptions, semantic data models, data mapping), Software Engineering (requirements modeling, software specification). The main unifying thread of this work is a belief in the importance of languages, which shape the way we think of the problem (an unabashed Whorfian!), and the need to be precise and logical about the semantics of such languages.

Alex is co-recipient of the most influential paper award of the 1994 International Conference on Software Engineering, and is proud to have contributed to the design and implementation of the Classic language/logic, which was used by AT&T as part of a system that configured "billions of dollars' worth of equipment sold".

Publish/subscribe (pub/sub) is a powerful paradigm enabling asynchronous interaction in large distributed settings ranging from Enterprise Application Integration, Internet-scale notification services, to processing events from RFIDs and monitoring the blogosphere. By limiting subscriptions to simple filters on message topics or content, pub/sub systems achieve great scalability in the number of publishers and subscribers. However, today's pub/sub is unfortunately rather limited; in particular, users cannot express conditions that involve more than a single message.

In this talk, I will overview the Cornell Cayuga System, a stateful pub/sub system for complex event processing. Cayuga supports powerful subscription features that includes maintenance of state across messages, parameterization and aggregation. Cayuga compiles subscriptions down to simple finite state automata that can be implemented very efficiently. I will conclude with experimental results and first experiences from several prototype deployments.

Johannes Gehrke is an Associate Professor in the Department of Computer Science at Cornell University and an Associate Director of the Cornell Theory Center. Johannes' research interests are in the areas of data mining, search, data privacy, complex event processing, and applications of database and data mining technology to marketing and the sciences. Johannes has received a National Science Foundation Career Award, an Arthur P. Sloan Fellowship, an IBM Faculty Award, the Cornell College of Engineering James and Mary Tien Excellence in Teaching Award, and the Cornell University Provost's Award for Distinguished Scholarship. He is the author of numerous publications on data mining and database systems, and he co-authored the undergraduate textbook Database Management Systems (McGrawHill (2002), currently in its third edition), used at universities all over the world.

Johannes was co-Chair of the 2003 ACM SIGKDD Cup, Program co-Chair of the 2004 ACM International Conference on Knowledge Discovery and Data Mining (KDD 2004), and he is Program Chair of the 33rd International Conference on Very Large Data Bases (VLDB 2007).

At Cornell, Johannes teaches in the Department of Computer Science, the Information Science Program, and in the Johnson Graduate School of Management. He has given courses and tutorials on data mining, data stream processing, and data privacy on Wall Street and all over the world, and he has extensive industry experience as technical advisor.

Joint work with Cristian Duda.

Work done in collaboration with Rene Müller and Donald Kossmann.