Optimization of Analytical Queries in Heterogeneous Systems

Abstract

According to International Data Corporation (IDC) reports, humankind produced more than 64 zettabytes of data in 2020. Global data creation and replication are expected to grow 23% annually between 2020 and 2025. Such a pace exceeds the rate of hardware performance improvement. Technologies such as Hadoop and Spark can not keep up, and new ways of increasing analytic queries performance and efficiency are required. One such technique is expanding the memory hierarchy and using specialized hardware for query processing. The article discusses query optimization problems for in-memory DBMSs using hardware accelerators. We give an overview of classical approaches to query optimization and state-of-the-art in heterogeneous query execution research. The advantages and disadvantages of existing solutions are analyzed, and gaps are identified. A reference architecture for heterogeneous query processing is proposed.

Keywords

• dbms
• gpu
• analytical query optimization
• olap
• analytical query processing
• in-memory databases
• heterogeneous query processing