Gloria: Graph-based Sharing optimizer for Event Trend Aggregation Contributions

We present a novel approach for optimizing the sharing plan for a workload of event trend aggregation queries with complex Kleene patterns. To the best of our knowledge, Gloria is the first sharing optimizer that offers flexible sharing decisions on nested Kleene patterns using effective online aggregation.

We introduce Gloria graph to model a variety of sharing opportunities in a diverse event trend aggregation workload. This allows us to transform the workload sharing problem into a path search problem. A cost model and effective pruning rules are also introduced to reduce the size of the Gloria graph.

We design a path search algorithm to find an event trend aggregation sharing plan from the Gloria graph. For a given complex query workload, our optimizer has a linear-time complexity in the number of nodes and edges in the Gloria graph, while still delivering optimality guarantees in most cases.

The experimental evaluation on three public data sets demonstrates the effectiveness of our pruning principles and the quality of the produced workload sharing plan. Our sharing plan achieves up to 10x performance improvement over the state-of-the-art approaches.

---