Ranking queries optimization over external data sources

摘要

Ranking queries has rapidly become an essential support in many new data systems. Contrary to Boolean queries that return all of the search query matches, the top-k query ranks the pertinent objects according to a given scoring function, and returns only the top-k answers that best match the user specifications. However, some data systems need to rank data that are exposed through external, autonomous data sources that generally come in various proprietary formats and limited access schema. These external data sources is exposed through Web services which provide a standard way to interact with heterogeneous data. In this context, users queries are answered by composing multiple data Web services. In this paper, we propose an approach that optimizes the top-k queries processing over data services. Our approach is based on two strategies: Pipeline Parallel Strategy and Bounding Strategy which aim to reduce the composition execution cost and the number of unnecessary service invocations, respectively.