Hybrid Scan Query Optimization in HTAP

Develop flexible and effective query optimization methods for hybrid scans in hybrid transactional and analytical processing (HTAP) databases that enable data exchange between row stores and column stores and generate high-quality hybrid execution plans for analytical queries.

Background

The paper reviews hybrid row/column execution where parts of a query may be executed on the row store and others on the column store, and highlights limitations in current interfaces that hinder data exchange between row and column data. In the open problems section, the authors explicitly identify the need for improved methods to generate hybrid plans for analytical queries in HTAP settings.

This problem arises because existing hybrid execution interfaces are limited and do not fully support seamless data exchange across storage formats, making it difficult to produce optimal plans for complex analytical workloads.

References

There are several open problems for HTAP query optimization. The first one is about hybrid scans for analytical queries. As the existing interface has limited functionality for hybrid scans (e.g., data cannot be exchanged between the row and column data), it calls for more flexible and effective methods to generate hybrid plans.

— HTAP Databases: A Survey  (2404.15670 - Zhang et al., 2024) in Section 6 (Open Problems and Challenges)