Maintaining Aggregate Views over Streams

Sliding-window aggregate queries over streams are an important class of continuous queries for online monitoring applications. In this paper, the evaluation results of sliding-window aggregates are seen as views, and we propose an aggregate framework with a baseline algorithm and an extended self-adaptive algorithm for maintaining sliding-window aggregate views over streams. We propose a taxonomy of aggregates according to dimensions that are particularly important in evaluating different aggregates, and present optimizations in evaluating high overhead aggregates for reducing space and time complexity. Performance study and experimental results show that we have good choices for evaluating different classes of aggregates.