Modelling and guaranteeing quality of service over data streams

In a growing number of web-based applications, massive data takes the form of continuous data streams. Most of these applications have deterministic and sophisticated performance requirements that need to be met under unbounded, high-volume and time-varying data streams. To be different from best-effort service providing, deterministic quality of service is generalized and modelled based on the theory of network calculus in this paper. A simple and efficient admission control mechanism is provided to determine the schedulability of several continuous queries with their QoS requirements satisfied simultaneously. Once the queries are admitted into the DSMS, our proposed QoS-Guaranteed scheduling algorithm assigns priorities to ready tuples that ensure all the QoS requirements. Furthermore, a QoS-Guaranteed batch-processing scheme is extended to reduce the query-scheduling overhead. Experiments are presented to illustrate and validate our strategies.