Research Project :: QStream

QStream - Quality Constraints for Processing Data Streams

Effectively analyzing data streams coming from multiple sensors is one of the most critical issues in dealing with sensor data. In general, streaming systems apply filters, aggregators, and possibly joins to multiple data streams. Unfortunately all currently known streaming systems are based on the best-effort principle, i.e. they try to adapt the given Quality-of-Service parameters such as result precision and latency to the actual load of the system.
QStream provides a framework for considering Quality-of-Service requirements referring to the relationship of data stream quality, delay of the result, and the necessary resource consumption. The proposed data stream processing model comprises of stream operators annotated by non-functional properties and operator graphs with buffers to compensate jitter from the original data stream or irregularity of stream operators' execution time. For a short introduction to the QStream project follow this slides.

Project Focus

  • Model of a QoS-capable DSMS
    • Composition of elementary operators to operator graphs
    • Estimation of the resources required to answer a continuous / standing query
    • Definition of Quality-of-Service (QoS) parameters
    • QoS negotiation
  • Architecture
    • Layered architecture to achieve data independency
    • Mapping of user's input (Query, QoS requirements) down to scheduled operators
  • Design
    • Duality of data stream schema and queries (phase model)
    • Modelling techniques for queries
    • Query optimization (intra query / inter query)
  • Adaptivity / Reconfiguration
    • Definition and gathering of data stream characteristics
    • Adaption of the DSMS to new data stream characteristics
    • Adding and Removing of data sources / queries
  • Processing of complex data streams (e.g. XML)

Papers

  • Technical Report: Trading Precision for Resources Consumption in a Sensor Data Streaming System
  • XML Stream Processing Quality (DBLP)

VLDB 2004 Demo

  • Demonstration Proposal (VLDB04.org)
  • Poster (pdf, ps)
  • Demo Slides (pdf)
Last Update: August 29, 2011 03:09 UTC
Author Database Technology Group, Technische Universität Dresden

Contact
Head
Prof. Dr.-Ing.
Wolfgang Lehner


Tel.: +49 351 463-38383
Fax: +49 351 463-38259


Visiting Address (View)
Nöthnitzer Str. 46
Room 3109
01187 Dresden

Postal Address
Technische Universität Dresden
Dep. of Computer Science
Institute for System Architecture
Database Technology Group
01062 Dresden
Germany