K2/Kleisli and GUS: Experiments in integrated access to genomic data sources

TLDR

Integrated access to heterogeneous biomedical data sources is a major challenge, with link‑driven federation, view integration, and warehousing explored as solutions, and users must weigh data characteristics, performance guarantees, and programming resources when choosing a strategy. The paper reports on experiences with two University of Pennsylvania systems, K2 (view integration) and GUS (data warehouse). K2 implements view integration while GUS operates as a data warehouse. Neither view integration nor warehousing clearly outperforms the other, but the authors provide practical tips for publishing database updates and using XML to streamline update processing in a warehousing context.

Abstract

The integrated access to heterogeneous data sources is a major challenge for the biomedical community. Several solution strategies have been explored: link-driven federation of databases, view integration, and warehousing. In this paper we report on our experiences with two systems that were developed at the University of Pennsylvania: K2, a view integration implementation, and GUS, a data warehouse. Although the view integration and the warehouse approaches each have advantages, there is no clear "winner." Therefore, in selecting the best strategy for a particular application, users must consider the data characteristics, the performance guarantees required, and the programming resources available. Our experiences also point to some practical tips on how database updates should be published, and how XML can be used to facilitate the processing of updates in a warehousing environment.