与以往尽最大努力的查询服务提供方式不同,讨论了数据流处理中确定性QoS的保证问题.以网络演算为理论基础,提出了一种数据流处理中的QoS建模和QoS保证方法,系统运行前验证所有查询在满足各自的QoS前提下的可调度性.在运行时为通过QoS可调度性验证的每个查询分配代表其QoS需求的服务曲线,从而保证各查询期望的QoS,为了提高查询处理效率,还讨论了保证QoS的批调度和查询共享,实验结果表明,该QoS保证方法能够有效地为数据流上的连续查询提供确定性的QoS保证.
Different from the traditional best-effort query service providing, the issue of deterministic QoS guarantee for data stream processing is discussed. Based on the theory of network calculus, an approach of QoS modeling and QoS guaranteeing for data stream processing is proposed. Before system running, the schedulability of all the queries with their QoS requirements satisfied simultaneously is verified. During run-time, service curves representing respective QoS requirements are allocated to each query admitted by QoS schedulability verification, in order to guarantee the expected QoS requirements. Moreover, QoS-guaranteeing batch scheduling and query sharing are extended to improve the query processing efficiency. Finally, experimental results show that the proposed approach offers deterministic QoS guarantee to continuous queries over data streams efficiently.