Grant: Alliance Grants

Rising data volumes and the advent of complex data-driven applications have ushered in the need to support sophisticated analytics. Database management system (DBMS) remains the mainstay of enterprise data management and analytics. However, traditional DBMS technology is unable to keep up with the demands of modern data processing, which is characterized by factors including, (1) the emergence of new applications that may introduce new kinds of workloads, (2) the dynamic nature of existing applications leading to new workload patterns, and (3) the ever-changing landscape of modern hardware. Hence, there are new challenges that need to be addressed to accelerate DBMS performance on modern workloads. Although, due to the advent of in-memory databases, researchers have explored query compilation as an alternative to traditional tuple-at-a-time or "iterator model" of query processing, the topic of how to architect a query compiler is still perceived to be a challenging task. To support interactive response time, the database query optimizer needs to generate compiled plans that can dynamically adapt to changing runtime conditions. Adaptive Query Processing (AQP) is aimed at leveraging runtime techniques to improve query performance. Rather than solely relying on statically collected statistics, AQP can take advantage of runtime information. The objective of this proposed research is to enable the execution of SQL queries using Just-In-Time (JIT) compilation in a manner that utilizes adaptive query processing techniques. The research will be conducted in collaboration with researchers at IBM, and with the aim of developing adaptive query processing techniques within IBM's Db2 database engine. The research is expected to benefit Canadian economy by helping IBM to improve their Db2 product offering, and supporting the training of HQPs in data systems and analytics technologies.