III: Small: Query Mesh - A Novel Paradigm for Query Processing

点击翻译按钮获取中文摘要

Technological advances in positioning, sensor and monitoringtechnology drive data acquisition devices to generate massive streamsof data. The goal of this research is to develop a new class ofhigh-performance stream data management systems capable of coping withscenarios with infinite data arriving in large volumes, and withnear-real time response requirements. The proposed query processing paradigm, termed the multi-route query mesh model (QM), overcomes amajor limitation in current query optimizers, both static and streamones alike, namely the assignment of a single `best' query executionplan for all input data. This approach, being based on the strongassumption of data uniformity, results in substandard performance forpossibly all data items. Instead, query mesh adopts a processingstructure composed of a data classifier and a multiple route planinfrastructure. Different learning models can be plugged asclassifier logic into the QM model. Given the complexity of the QMsolution space, cost-based search heuristics are designed toefficiently find high-quality query meshes. QM is adaptive supportingthe detection and incremental modification of the QM classifier andits routes. The intellectual merit of this project lies in the design, development andevaluation of a novel multi-route paradigm for stream queryprocessing, -- a perfect middle ground between the two currentextremes of single-plan versus route-less solutions. Experimentalstudies compare query mesh to state-of-the-art solutions. QM impactssociety by facilitating a wide range of stream-centric applications,including medical out-patient monitoring, emergency management, andbusiness intelligence processing, and by integrating projectactivities with education.Further information on this project can be found at the projectweb page: http://dsrgweb.cs.wpi.edu:8180/DSRG/