HYRAQ: optimizing large-scale analytical queries through dynamic hypergraphs.

In critical situations, making quick and precise decisions requires a rapid execution of a large amount of concurrent navigational and exploratory queries over collected data stored in repositories such as data warehouses. To satisfy the decision-maker's requirement, a deep understanding of the properties of these queries is necessary. In addition to their large-scale , they are ad-hoc , dynamic and highly interacted . By a quick analysis of these properties, we figure out that the first three are factual whereas the last one is behavioral. The literature has widely reported that the interaction of analytical queries has a crucial impact on selecting optimization structures (e.g., materialized views) in data storage systems. By keeping these four properties in mind, it becomes a necessity to find scalable and efficient data structures to simultaneously model them for better optimization of large-scale queries. In this paper, we first show the crucial role of the interaction phenomenon in optimizing concurrent data and mining queries by identifying its limited capacity in considering all factual properties. Secondly, we propose a dynamic hypergraph as a data structure to manage the four above properties and we show its great contribution in selecting materialized views. Finally, intensive experiments are conducted to evaluate the efficiency of our proposal and its connectivity with a commercial DBMS.