Zhihua Deng created HIVE-29165:
----------------------------------
Summary: PartColNameInfo could introduce high hash collision due
to the wide table
Key: HIVE-29165
URL: https://issues.apache.org/jira/browse/HIVE-29165
Project: Hive
Issue Type: Improvement
Components: Standalone Metastore
Reporter: Zhihua Deng
Assignee: Zhihua Deng
In {{{}DirectSqlUpdatePart{}}}, the {{PartColNameInfo}} acts as a map key,
referring to those statistics to be updated or inserted. If the current table
has lots of columns, say 1000, then for each {{{}PartColNameInfo{}}}, it’s
assumed to hash into the same map buckets more than 1000 times. If the table
has thousands of partitions, locating or inserting the {{PartColNameInfo}}
could be very slow.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
