NISU: A Novel Index Structure on Uncertain Data in Large-Scale Publish/Subscribe Systems

Abstract

Publish/Subscribe Systems are widely used in messaging systems due to loose coupling and asynchronous. As the number of objects participating in the message system increases, there has also been a spurt in the number of messages, which brings great challenges to the traditional event matching methods. After studying the status quo of event matching in the case of large-scale Uncertain Data, we proposed a simple solution named NISU(New Index Structure on Uncertain Data) for efficient event matching. In the proposed scheme, we use P-Skyline to filter unrelated events and subscriptions, then divide the attribute space into several attribute subspace in order to filter unrelated subscriptions, and finally make use of constraint satisfaction criterion for event matching. In addition, in the event matching process, we use confidence to relax the event matching criteria, avoiding the problem of inaccurate matching caused by Uncertain Data set. The experimental results show that the NISU is rapid, low consumed and efficient on large-scale Uncertain Data set.