Adaptive run-time overhead adjustments for optimizing multiple continuous query processing

The time-varying characteristics of infinite data streams require continuous queries to be adaptively processed. The order in which multiple join operations are evaluated has serious consequences for the algorithm performance because the selectivity of each join operation can differ significantly from the selectivity of the other operations. The evaluation order may be effectively determined using the k-EGA and A-SEGO schemes, as proposed in previous studies. These methods optimize target continuous queries by monitoring a set of their promising subplans simultaneously. Each scheme also employs a user-defined cost-bound parameter for controlling the number of monitored subplans. A more optimized global plan may be generated by using a more highly configured cost-bound parameter. However, this approach can increase the overhead associated with monitoring the subplans. This paper proposes a new scheme, Adaptive Run-time Overhead Adjustment (AROA), which provides a novel method for adaptively determining the value of a cost-bound parameter based on the system environment. Unlike the previously described A-SEGO scheme, the scheme proposed here automatically selects the cost-bound parameter to reflect the system workloads (e.g., the input tuple rate, and other parameters). This method not only augments the probability of generating an optimized execution plan, it reduces the run-time delay caused by the optimization process. Experimental verification of the proposed scheme AROA demonstrated that AROA outperforms the previous schemes.