Article in Proceedings INPROC-2006-54

BibliographySchwarz, Thomas; Grossmann, Matthias; Nicklas, Daniela; Mitschang, Bernhard: Exploiting Type and Space in a Main Memory Query Engine.
In: Davis Junior, Clodoveu Augusto (ed.); Vieira Monteiro, Antonio Miguel (ed.): Proceedings of the VIII Brazilian Symposium on GeoInformatics : GeoInfo2006 ; Campos do Jordão, Brazil, November 19-22, 2006.
University of Stuttgart : Collaborative Research Center SFB 627 (Nexus: World Models for Mobile Context-Based Systems).
pp. 35-52, english.
INPE, November 19, 2006.
ISBN: 9 788517 000270.
Article in Proceedings (Conference Paper).
CorporationBrazilian Symposium on GeoInformatics
CR-SchemaH.2.4 (Database Management Systems)
H.3.1 (Content Analysis and Indexing)
H.2.8 (Database Applications)
KeywordsMain Memory Query Engine, Indexing, Spatial Index, Type Hierarchies, Deployable Query Engine
Abstract

More and more spatial data is accessible over the web or through portals of wireless service providers. In this context the main selection criteria for the data are the type of the requested data objects and their position in the real world. Integration and performance issues are challenged by the need to process ad hoc queries in an interactive fashion. In this paper we investigate how a main memory query engine can be used to meet these requirements. It has the added benefit of being easily deployable to many components in a large-scale data integration system. Hence, we analyze how such a query engine can best exploit the query characteristics by employing an index structure that leverages spatial and type dimensions.

In order to support query processing in the best possible way we investigate a specific multi-dimensional main memory index structure. Compared to the straightforward approach using separate indexes on type and position we can increase the performance up to almost an order of magnitude in several important usage scenarios. This requires to tweak the mapping of type IDs to values in the type dimension, which we discuss extensively. This enables the overall system to be used interactively, even with large data sets.

Full text and
other links
PDF (475557 Bytes)
ContactThomas Schwarz schwarts@ipvs.uni-stuttgart.de
Department(s)University of Stuttgart, Institute of Parallel and Distributed Systems, Applications of Parallel and Distributed Systems
Project(s)SFB-627, B1 (University of Stuttgart, Institute of Parallel and Distributed Systems, Applications of Parallel and Distributed Systems)
Entry dateDecember 18, 2006
   Publ. Department   Publ. Institute   Publ. Computer Science