Reproduce “CERT: Finding Performance Issues in Database Systems Through the Lens of Cardinality Estimation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jinsheng Ba, Manuel Rigger
CERT, a black-box testing methodology, uncovers performance issues in database systems by identifying violations of cardinality restriction monotonicity within the query optimizer's estimations. This technique, which avoids actual query execution, achieves a 386x higher throughput and an 85% confirmation rate for identified issues compared to existing methods.
Key insights
- The cardinality restriction monotonicity property (a more restrictive query should not have a higher estimated cardinality) serves as an effective and objective oracle for detecting flaws in query optimizers.
- Analyzing estimated cardinalities instead of actual execution times significantly boosts testing efficiency and robustness by eliminating environmental noise and measurement costs.
- Ensuring structural similarity between the query plans of original and restricted queries is crucial for preventing false positives and validating the relevance of cardinality comparisons.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.