The Golden Mean Operator Scheduling Strategy in Data Stream Systems

A scheduling strategy, called Golden Mean (GM), is designed to achieve the dynamic balance between the memory usage minimization and latency minimization by taking the status of the future workload, the current memory consumption and QoS requirement of the query into account. In the GM scheduling strategy, the executing order of the operators is decided uniformly according to the Scoring Function. The static parameters in the Scoring Function can be customized to meet the need of various application scenarios, while the dynamical parameters can be adjusted automatically according to the current system status. In addition, the priority levels of queries can be guaranteed by GM. The experimental results indicate that the GM scheduling strategy is very effective in practice.