Database Seminar Series (2007-2008)

CarTel is a software and hardware platform for collecting and processing data from moving vehicles.  Data is captured from GPS, Wi-Fi, and OBD-II interfaces and stored in a local database on each car.  This stored data is transferred opportunistically, via available Wi-Fi networks, cellular modems, or by "muling" data on a user's cell phone or USB key, to a central "portal", where users can browse and visualize it.  To allow non-expert users to specify what data they would like to collect from remote vehicles, CarTel includes a simple database-like interface for programming and configuration.

In this talk, I will discuss the architecture of CarTel data management system and mention some applications of the technology to traffic and fleet management, driving safety, and fuel conservation.   I will also show a brief demo of the portal software, which provides a map-based interface for browsing about 1100 hours of driving data from collected by our system from six cars over the past year.

In this talk, I will present a summary of our recent research in understanding how disks fail and how file and storage systems handle such failures. Our findings reveal numerous design and implementation problems in a wide range of both open-source and commercial systems; put simply, file systems are broken (at least when it comes to reliability).

I will then present a number of current research directions that we are pursuing in order to build a new generation of robust and reliable storage systems. With more formal analysis and construction techniques, we hope to transform the "art" of file system design and implementation into a more rigorous and careful science.

Data warehousing and analytic processing have been active areas of database research and development for nearly two decades. Many experts now consider these topics to be "solved", especially with regard to performance. However, an argument can be made that users and databases have simply reached an uneasy truce with regard to analytic processing. If users avoid ad-hoc, exploratory queries that might take days to execute, then of course the database performs just fine.

In this talk, I will describe query processing in a database system called DBO that is designed from the ground up to support interactive analytic processing. DBO can run database queries from start to finish and produce exact answers in a scalable fashion. However, unlike any existing research or production system, DBO is able to produce statistically meaningful approximate answers at all times throughout query execution. These answers are continuously updated from start to finish, even for "huge" queries requiring arbitrary quantities of temporary secondary storage. Thus, a user can stop execution whenever satisfied with the query accuracy, which may translate to dramatic time savings during exploratory processing.

Professor Elmagarmid was a chief scientist in the Office of Strategy and Technology at Hewlett-Packard (HP). While at HP, he was responsible for software strategy coming out of the corporate CTO office and contributed to cross company roadmap initiatives. He also served on the technology council for HP. Professor Elmagarmid serves as a consultant in the areas of computing technology, the role of technology in societal development and strategic planning. He has several international clients in Qatar and Italy. Within North America, he has worked for Telcordia Technology, Bellcore, IBM, CSC, Harris, D. H. Brown and Associates, MCC, Bell Northern Research, Molecular Design Labs, King & Spalding LLP, and UniSql to name a few.

A relational ranking query uses a scoring function to limit the results of a conventional query to a small number of the most relevant answers. The increasing popularity of this query paradigm has led to the introduction of specialized rank join operators that integrate the selection of top tuples with join processing. The number of input tuples that an operator accesses is called the input depth of the operator, and this is the driving cost factor in rank join processing. This introduces the problem of depth estimation, which is crucial for the costing of rank join operators during query compilation and thus for their integration in optimized physical plans.

This talk will provide an overview of our recent work on DEEP, an estimation methodology for approximating the depths of rank join operators in a physical plan. DEEP relies on a general, principled framework in order to formalize depth computation in terms of the joint distribution of scores in the base tables. This results in a systematic estimation methodology that takes the characteristics of the data directly into account and thus enables more accurate estimates. The talk reviews the main technical points of our work and presents experimental results that validate the benefits of the proposed technique over existing methods.

Many kinds of real-life data exhibit logical ordering among their data items and are thus sequential in nature. However, traditional online analytical processing (OLAP) systems and techniques were not designed for sequence data and they are incapable of supporting sequence data analysis.

In this talk, I will present the concept of Sequence OLAP, or S-OLAP for short. The biggest distinction of S-OLAP from traditional OLAP is that a sequence can be characterized not only by the attributes’ values of its constituting items, but also by the subsequence/substring patterns it possesses. This talk will cover the concepts of sequence cuboid and sequence data cube as well as the implementation issues.