Preference query evaluation over expensive attributes

  • Authors:

  • Justin J. Levandoski;Mohamed F. Mokbel;Mohamed E. Khalefa

  • Affiliations:

  • University of Minnesota, Minneapolis, MN, USA;University of Minnesota, Minneapolis, MN, USA;University of Minnesota, Minneapolis, MN, USA

  • Venue:
  • CIKM '10 Proceedings of the 19th ACM international conference on Information and knowledge management
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

Most database systems allow query processing over attributes that are derived at query runtime (e.g., user-defined functions and remote data calls to web services), making them expensive to compute relative to relational data stored in a heap or index. In addition, core support for efficient preference query processing has become an important objective in database systems. This paper addresses an important problem at the intersection of these two query processing objectives: efficient preference query evaluation involving expensive attributes. We explore an efficient framework for processing skyline and multi-objective queries in a database when the data involves a mix of "cheap" and "expensive" attributes. Our solution involves a three-phase approach that evaluates a correct final preference answer while aiming to minimizing the number of expensive attributes computations. Unlike previous works for distributed preference algorithms that assume sorted access over each attribute, our framework assumes expensive attribute requests are stateless, i.e., know nothing previous requests. Thus, the proposed approach is more in line with realistic system architectures. Our framework is implemented inside the query processor of PostgreSQL, and evaluated over both synthetic and real data sets involving computation of expensive attributes over real web-service data (e.g., Microsoft MapPoint).