You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@impala.apache.org by "Riza Suminto (Jira)" <ji...@apache.org> on 2025/01/06 17:18:00 UTC

[jira] [Created] (IMPALA-13644) Generalize and move getPerInstanceNdvForCpuCosting into AggregationNode.

Riza Suminto created IMPALA-13644:
-------------------------------------

             Summary: Generalize and move getPerInstanceNdvForCpuCosting into AggregationNode.
                 Key: IMPALA-13644
                 URL: https://issues.apache.org/jira/browse/IMPALA-13644
             Project: IMPALA
          Issue Type: Bug
          Components: Frontend
    Affects Versions: Impala 4.4.0
            Reporter: Riza Suminto
            Assignee: Riza Suminto


getPerInstanceNdvForCpuCosting is a method to estimate the number of distinct values of exprs per fragment instance when accounting for the likelihood of duplicate keys across fragment instances. It borrows the probabilistic model from formula described in IMPALA-2945. This method is exclusively used by AggregationNode only.

[https://github.com/apache/impala/blob/99529db6ad62ddc34cbfd924d7e41b1fce5b60a2/fe/src/main/java/org/apache/impala/planner/PlanFragment.java#L634-L642] 

We should move this method to AggregationNode and generalize it to accept NDV estimate calculated at AggregationNode.computeStats() as input. The number from computeStats should be more precise now after improvement from IMPALA-13405, IMPALA-13526, and IMPALA-13622.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)