Predição de desempenho para junções por similaridade baseadas em conjuntos

Query performance prediction is essential for many important tasks related to cloud-based database management including resource provisioning, admission control, and pricing. Recently, there has been great interest in building prediction models to estimate execution time of traditional SQL queries. While suitable for typical OLTP/OLAP workloads, these existing approaches are insufficient to model performance of complex data processing activities for deep analytics such as cleaning and integration of data. These activities are largely based on similarity operations, which are radically different from regular relational operators. In this dissertation, we consider prediction models for set similarity joins. We exploit knowledge of optimization techniques and design details popularly found in set similarity join algorithms to identify relevant features, which are then used to construct prediction models based on statistical machine learning. We present an extensive experimental evaluation to confirm the accuracy of our approach.