Scaling Up Query Allocation in the Presence of Autonomous Participants

Abstract

In large-scale, heterogeneous information systems, mediators are widely used for query processing and the good operation of a system strongly depends on the way the mediator allocates queries. On the other hand, it is well known that a single mediator is a potential scalability and performance bottleneck as well as a single point of failure. Thus, multiple mediators should perform the query allocation process. This task is challenging in large-scale systems because participants typically have special interests that are not performance-related. Mediators should satisfy participants interests as if there was a single mediator in the system — i.e., with no, or almost no, additional network traffic. In this paper, we propose a virtual money-based query allocation method, called VM _b QA, to perform query allocation in the presence of multiple mediators and autonomous participants. A key feature of VM _b QA is that it allows a system to scale up to several mediators with no additional network cost. The results show that VM _b QA significantly outperforms baseline methods from both satisfaction and performance points of view.

Work partially funded by the Dataring project of the Agence Nationale de la Recherche and the Epilog project of the PREDIT program of the french ministry of transportation.