Predicting whether a query merits CMP tuning

Develop prediction techniques that forecast whether a SQL query is worth subjecting to cost model parameter tuning, thereby avoiding unnecessary tuning overhead.

Background

The experiments show that the benefit of deterministic optimal tuning varies across workloads and queries; queries with more joins do not necessarily benefit more from cost model parameter tuning. Because tuning incurs nontrivial overhead, a predictor that identifies queries likely to benefit could prevent unnecessary tuning.

References

Since tuning CMPs comes with nontrivial overheads (even with the optimization techniques presented in Section~\ref{sec:optimization-tuning-time}), it is an interesting direction to develop prediction techniques to forecast whether a query is worth CMP tuning, for the sake of avoiding unnecessary tuning. We leave this for future work.

— Query Performance Tuning with Optimal Exploration of Optimizer Cost Model Parameter Space  (2610.02607 - Wu et al., 2 Oct 2026) in Section 4.1.1, “Query Groups with Different Number of Joins”