There is a host of existing and newly-emerging applications that require sophisticated processing of fast, high-volume data streams. These stream-based applications typically track data from numerous continuous data streams (coming from such sources as sensor networks and stock feeds), processing them in a timely fashion for purposes of reduction, aggregation, and correlation, and mining them for signs of abnormal activity. Example applications include environmental monitoring, asset tracking, portfolio management, and network monitoring.

This tutorial will present an overview of the fundamental and practical problems and pertinent solutions in supporting stream-based applications, covering the recent body of research in this domain. The tutorial will first cover in detail a new class of database systems that are designed and optimized for real-time processing of data streams. The tutorial will then address the challenges and opportunities that arise when distributing stream processing functionality across multiple machines, as well as the efficient dissemination of data streams. Finally, the tutorial will outline the limitations of existing approaches and disclose several directions of research in the area.

1. Introduction
   1. Motivation: Stream-based Applications
   2. Challenges and Opportunities
2. Centralized Processing
   1. Inverted Architectures: Pull vs. Push
   2. Languages and Operators
   3. Resource Allocation and Quality-of-Service
   4. Approximation and Optimization
3. Distributed Processing and Dissemination
   1. Content-based Stream Routing and Dissemination
   2. Query Optimization
   3. Load Management
   4. Availability and Reliability
4. Open Problems

Ugur Cetintemel received the PhD degree in Computer Science from the University of Maryland, College Park in 2001. Ugur is currently an assistant professor at the department of Computer Science, Brown University. His primary research interests are in the architecture and performance of advanced networked databases and information systems. More information about his research is available at http://www.cs.brown.edu/~ugur/.

1. Introduction
   1. Context Definitions
   2. Context Types
   3. Examples of Context-Aware Applications
2. Context Models
   1. Organization
      1. Addressing Context Information
      2. Primary and Secondary Context
      3. Spatial Models
   2. Properties
      1. Scope
      2. Dynamism
      3. Complexity
3. Context Management
   1. Application-specific
   2. Shared
   3. Federated
4. Classes of Context-Aware Applications
   1. Selection
   2. Action
   3. Presentation
5. Context-Aware Communication
   1. Spatial Events
   2. Hoarding
   3. Geocast
6. Case Studies: A number of different projects will be discussed and classified along the presented criteria

This is a half-day tutorial discussing the design and implementation of software systems as well as open research problems related to data processing and collection in wireless sensor networks. During the first hour-and-a-half, we focus on the design of the TinyDB data collection system for networks of Berkeley motes running the TinyOS operating system. Then, during the remainder of the tutorial, we survey relevant literature from the database, networking, and OS communities and identify a number of unsolved and inadequately addressed research problems.

This tutorial is intended for anyone interested in wireless sensor networks with a general background in computer science, be they users of sensor networks looking for an easy way to collect data, developers interested in the design of TinyOS and TinyDB, or researchers in search of challenging new problems.

1. Mote and TinyOS Background
   1. Mote hardware and commonly available sensors and sensor boards.
   2. TinyOS design principles and basic concepts.
   3. The nesC programming language for TinyOS.
   4. Major components in TinyOS including the radio stack, power management features, and clock, timing, and time synchronization.
   5. Radio communication and multi-hop routing, including both tree-based and geographic variants.
   6. The Matchbox file system for non-volatile storage in the on-mote Flash.
2. TinyDB Usage
   1. The data model and query language of TinyDB and how it differs from traditional database systems.
   2. Building applications through TinyDB GUI tools.
   3. Building applications using TinyDB's Java API.
   4. Extending TinyDB with new attributes, aggregates, and types of sensors.
   5. Hands-on exercise with motes.
3. TinyDB Implementation
   1. The architecture of TinyDB.
   2. Its power-efficient execution framework.
   3. The interaction between routing and query processing, and the need for "abstraction Breaking" as it relates to performance in sensornets.
   4. The implementation of in-network query processing operators, including selections, aggregates, and storage primitives.
   5. Optimizing the ordering of query operators and data acquisitions.
   6. The design of the PC-side server, and deciding how to partition queries between the server and the sensor network.
4. Open Research Challenges
   1. Literature survey
   2. Query processing in heterogeneous sensor networks, e.g. adding nodes with faster networks, bigger processors, or larger batteries to a homogeneous mote network.
   3. Query processing in occasionally connected sensor networks.
   4. Distributed in-network storage and access methods.
   5. Statistical inferences through queries, including basic techniques for learning correlations between sensors and using those correlations to detect outliers and improve query performance.
   6. Declarative interfaces for collaborative signal processing.
   7. Energy-efficient efficient, in-network join algorithms.
   8. The role of adaptivity in sensor-networks.
   9. Multi-query issues and opportunities and the particular challenges of work sharing in sensornets.

Wei Hong is a senior researcher at Intel Research, Berkeley. His current research focuses on data management in sensor networks. He leads the Tiny Application Sensor Kit (TASK) project at Intel Research and co-designed/developed TinyDB, an open-source, in-network sensor database system with Samuel Madden. Prior to joining Intel Research, Wei co-founded and architected the products of two startup companies: Illustra Information Technology Inc. and Cohera Corp. Illustra developed the first successful commercial Object-Relational database system. It was acquired by Informix, now part of IBM. Cohera provided electronic catalog management solutions based on a novel federated database system that it developed. Its technology was acquired by PeopleSoft. Wei earned a Ph.D. in computer science from UC Berkeley and holds a master and two bachelor degrees from Tsinghua University in Beijing, China.

Samuel Madden is an Assistant Professor in the Department of Electrical Engineering and Computer Sciences and a member of the Computer Sciences and Artificial Intelligence Laboratory at MIT. He received his Ph.D. in Computer Science from the University of California at Berkeley in 2003. His research interests are in the area of distributed and adaptive data management and related networking and systems issues, particularly as they pertain to sensor networks and streaming data.

1. Location acquisition methods
2. Representation of location data
3. Storage, indexing, and querying
4. Aggregation and provisioning
5. Benchmarking