A data stream is a potentially unbounded sequence of data tuples that arrive in real-time to an application. Examples of data streams include transactional data streams, such as a sequence of credit card transactions, which could be monitored for anomalies that might indicate fraud, and measurement data streams, such as sensor data or network performance data. This recent addition to the “Advances in Database Systems” series surveys the latest research in the handling of data streams.

The management of data streams presents many challenges not found in traditional database management systems. The notions of time and sequence play essential roles in streaming data. It is important to recognize temporal trends and patterns in the data via continuous queries. Such queries are continuously run on the data. For example, a traffic management system might want to continuously know the average speed of cars on a section of road over a moving window of the past five minutes. Based on the results of these continuous queries, triggers and alert thresholds can produce an automatic system response, or will signal the need for human intervention.

Another query issue is that the data stream can be unbounded. Thus, queries cannot use blocking operators (for example, Group-By or Sort) that require a finite data set for execution. Query performance and optimization strategies are significantly different for data streams.

Data quality in a stream may be inconsistent. Failures in sensors and monitors will produce erroneous data or no data at all. Delays in transmissions may produce data streams that are out of temporal order. Also, the amount of data being streamed may overwhelm the processing and memory capacities of the receiving system. When this happens, intelligent data dropping (or load shedding) policies are needed.

An introductory chapter sets the stage for subsequent chapters that address state-of-the-art research topics in streaming data management. The chapters are written by well-known researchers in the database field, and cover recent advances, with literature references through 2004. Topics covered include query execution and optimization on data streams, the use of Extensible Markup Language (XML) to format streaming data, geographical data streams, and streaming data applications.

This book is recommended for researchers who want to get up to speed quickly on the latest ideas in data stream management. Application developers who deal with real-time data streams will also benefit by gaining an understanding of the challenges and the research insights that will lead the way to effective system solutions.