CAQE: A Contract Driven Approach to Processing Concurrent Decision Support Queries

Real-time analytical systems need to handle workloads comprised of expensive decision support queries with diverse quality of ser-vice requirements known as contracts. Contract driven multi-query processing, being an NP-hard problem, remains unaddressed to date. The traditional approach of blindly pipelining the entire input through a shared execution plan is not viable due to the diversity in query contracts. To tackle this challenge, we now develop a flexi-ble model to express contracts and accompany it with an effective means to measure the run-time contract satisfaction. We propose our Contract-Aware Query Execution framework CAQE. In this work, we exploit the principle that “different portions of the in-put contribute to disparate subsets of queries with varying degrees of progressiveness. ” Therefore, CAQE’s processing of the input chunks is driven by how the different query contracts are being met at run-time. To maximize the contract satisfaction of the work-load, CAQE leverages the dependencies of input chunks across the queries. This enables us to determine the impact of processing par-ticular input chunks on improving the run-time contract satisfac-tion. Our experiments demonstrate the effectiveness of CAQE in increasing the overall contract satisfaction of the workload, specifi-cally 2 fold better than existing multi-query processing techniques. 1