-
Type:
Bug
-
Resolution: Unresolved
-
Priority:
Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
None
-
Query Optimization
-
None
-
None
-
None
-
None
-
None
-
None
-
None
For TPCE locust queries q08/q10/q14/q15/q17 it seems we prioritize a plan with a HJ/NLJ on a COLLSCAN instead of an INLJ probing on the "smaller" table.
When comparing to classic, the classic plan is (closer) to an INLJ (probing on the small table). While the absolute difference is small (on the order of ~10s of docs) this cost is visible in terms of query latency because each query is very fast.
- is duplicated by
-
SERVER-139211 Join cost model picks COLLSCAN + hash join over index nested loop join for in-memory data once cardinality estimates are accurate
-
- Closed
-