During this talk, I will give an overview of my thesis research and mainly talk about my recent work, Rx, which is a novel technique to survive software failures and improve system availability. In addition, I will briefly introduce another piece of my work, SafeMem, which leverages ECC memory technology to provide fine-grained memory monitoring with very low overhead. SafeMem has been used for detecting memory leak and memory corruption bugs. Finally, I will mention my current on-going project, LIFT, which use dynamic information flow tracking to secure your software.

As software permeates our daily life and becomes increasingly complex, software dependability becomes critically important. The main obstacle to building dependable software is software bugs, which often account for 40% of computer system failures and over 50% of security vulnerabilities. Existing dynamic bug detection techniques suffer from high overhead and can greatly benefit from the recent impressive advances in computer architecture.

In this talk, I will present my thesis work on using general, novel hardware support for detecting software bugs in production runs. First, I will introduce a simple and general hardware framework called iWatcher that enables efficient memory monitoring and can be used to detect a wide variety of bugs with very low overhead, orders of magnitudes smaller than software-only approaches. Next, I will present an innovative bug detection technique called PC-based invariants and an efficient novel hardware implementation of this technique to detect general memory corruptions during production runs. Finally, I will briefly describe my recent work on using the iWatcher hardware support for efficient data structure consistency checks and my future research plan.

Given two database schemata, the task of schema matching involves the identification of a mapping between attributes of the two schemata. New challenges such as that of Web service discovery and composition call for a fully-automatic schema matching process. In this talk we shall present (both intuitively and formally) several challenges in devising algorithms for fully-automatic schema matching, primarily the uncertainty in the outcome of such a process. We shall then outline the monotonicity principle as a tool in schema matching classification and show its usefulness in devising matching heuristics, based on top-K best schema mappings.

This work is a joint work with Ateret Anaby-Tavor, Alberto Trombetta, and Danilo Montesi

Zhenmin will continue his unfinished talk on Feb 8. See that slot for its abstract.

Thomson is a leading provider of integrated information solutions for knowledge professionals around the world (www.thomson.com). It employs over 40,000 individuals in 45 countries. Thomson consists of four primary market sectors: Financial, Legal, Scientific & Health Care, and Educational. Thomson Legal & Regulatory (TLR) focuses on providing productivity and work-flow solutions to practitioners in the legal domain (www.tlrg.com).

TLR R&D performs applied research and development to help drive the sustainable operation and growth of TLR companies by providing technological innovations that meet its customer needs. It specializes in in formation retrieval, information extraction, text categorization, text clustering, and data mining, with an emphasis on finding novel ways to organize and present legal and business information. This presentation will first provide an overview of TLR's business and its customers. It will then describe the work of TLR R&D in the fields of research mentioned above, in particular, detailing its efforts in areas of highly granular text classification [using taxonomies possessing O(10^5) nodes] and entity extraction & linking [harnessing authority resources consisting of O(10^6) entities].

Graphs are ubiquitous but divergent, with critical applications in domains ranging from software engineering to computational biology. However, it is challenging to analyze any reasonably large collection of graphs due to its high computational complexity. Development of scalable methods for the analysis of massive graph databases thus becomes one major thrust in data mining and database research.

In the core of many graph-related applications, there are two fundamental problems: how to mine graph patterns and how to process graph queries. My initial study grasped a tight connection between these two seemingly parallel areas. In this talk, I will first present a novel graph canonical labeling system that is able to speedup the discovery of frequent subgraph patterns. Next, discriminative pattern analysis is introduced for constructing compact yet high-quality graph indices, which are then applied to exact and approximate graph search in large graph databases. Such index mechanism is shown to be very effective in processing graph queries. The finding of graph pattern-based indexing is profound and yet to be fully explored since the same concept can also be applied to graph classification and clustering. In the end of my talk, I am going to examine broader applications of graph patterns, such as biological network analysis for functional annotation and program flow classification for software bug isolation.

Database query languages can be intimidating to the non-expert, leading to the immense recent popularity for keyword based search in spite of its significant limitations. The holy grail has been the development of a natural language query interface.

Whereas there has been a great deal of work in natural language understanding and question answering, our focus is on what happens once a natural query has been "understood". To be able to pose this against a database, one still requires knowledge of the database schema, and a mapping of the query to this schema.

In this talk, I will present our work on Schema-Free XQuery, permitting the expression of complex queries against XML databases with minimal schema knowledge. I will then show how these schema free querying facilities are being used in our ongoing work on NaLIX, a natural language query interface for an XML database.

Data integration systems provide a one-stop interface to multiple disparate data sources. Despite a lot of interest from both academia and industry over the last decade, these systems are still built and maintained in a manual, labor-intensive, and error-prone process. As a result, deployment has been limited. In this talk I will describe my work in the AIDA project at Illinois. Namely, I present three approaches to reduce integration costs: collaboration, automation, and relaxation. First I discuss the application of mass collaboration techniques to perform critical integration tasks at little cost to the system builder (reminiscent of open-source software efforts). I then discuss the use of automatic techniques (e.g. anomaly detection, machine learning) to reduce the high cost of system maintenance. Lastly, I describe initial work on relaxing one fundamental cause of integration costs - rigid structural requirements - resulting in IR-style best-effort integration scenarios. These techniques can help significantly reduce integration costs, promoting system deployment and offering solutions to build integration systems where not previously possible.

Alan Stacklin has more than 20 years of experience in developing and commercializing software technologies. Stacklin's has worked in the field of business intelligence, data mining, and product management. His proven ability to translate market needs into commercial software applications compliments Akoya's mission. Stacklin's successful business ventures include: ZDU.com, an online technology education platform, MetFAbCity.com, a supply chain automation exchange for the metal fabrication industry, and CPGNetwork.com, a business intelligence platform for the consumer packaged goods industry. His leading-edge software applications assisted in the growth of these companies.

Stacklin earned an MBA from the University of Rochester's William E. Simon Graduate School of Business, an MS from Union College and a BS from the University of Pittsburgh.

Discovering contrasts between collections of data is an important task in data mining. In this talk, we consider a new type of contrast pattern, called a Minimal Distinguishing Subsequence (MDS). An MDS is a minimal subsequence that occurs frequently in one class of sequences and infrequently in sequences of another class. It is a natural way of representing strong and succinct contrast information between two sequential datasets and can be useful in applications such as protein comparison, document comparison and building sequential classification models. Mining MDS patterns is a challenging task and is significantly different from mining contrasts between relational/transactional data. One particularly important type of constraint that can be integrated into the mining process is the gap constraint. We present an efficient algorithm called ConSGapMiner (Contrast Sequences with Gap Miner), to mine all MDSs satisfying a minimum and maximum gap constraint, plus a maximum length constraint. It employs highly efficient bitset and boolean operations, for powerful gap based pruning within a prefix growth framework. A performance evaluation with both sparse and dense datasets, demonstrates the scalability of ConSGapMiner and shows its ability to mine patterns from high dimensional datasets at low supports.

This talk is based on a paper which received the 2005 IEEE ICDM Best Paper Award and co-authored with Xiaonan Ji and James Bailey of University of Melbourne.

In contrast to traditional database queries that run once against a database snapshot, continuous queries continuously generate new results (or changes to results) as new data and updates arrive in streams. Many applications, e.g., publish/subscribe systems, need to handle a large number of long-standing continuous queries whose results are needed across a wide-area network. The naive approach, which checks each incoming data item against every query and sends a separate notification for each affected query, is not scalable. While there has been a considerable amount of work on continuous filters, more complex queries, such as joins and aggregates, are more challenging.

In this talk, I will first describe our recent results on processing a large number of continuous joins. Our techniques are input-sensitive---they exploit patterns in queries and data for efficient processing. We demonstrate the advantage of this approach over previously known techniques both theoretically and experimentally.

The second problem I will address is how to disseminate the results of many continuous queries efficiently to users over a network. Traditional solutions are either database- or network-centric, but we argue that there is a previously unexplored design space between these two extremes, and we show how to achieve better scalability by incorporating both database- and network-side considerations.

This paper presents a principled framework for efficient processing of ad-hoc top-k (ranking) aggregate queries, which provide the k groups with the highest aggregates as results. Essential support of such queries is lacking in current systems, which process the queries in a naive materialize-group-sort scheme that can be prohibitively inefficient. Our framework is based on three fundamental principles. The Upper-Bound Principle dictates the requirements of early pruning, and the Group-Ranking and Tuple-Ranking Principles dictate group-ordering and tuple-ordering requirements. They together guide the query processor toward a provably optimal tuple schedule for aggregate query processing. We propose a new execution framework to apply the principles and requirements. We address the challenges in realizing the framework and implementing new query operators, enabling efficient group-aware and rankaware query plans. The experimental study validates our framework by demonstrating orders of magnitude performance improvement in the new query plans, compared with the traditional plans.