Details
-
Task
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
None
-
None
Description
benchmarks for option 'useReadAheadInput' follows :
Do_1.run_readahead mariadb thrpt 50 60664.345 ± 2084.866 ops/s
|
Do_1.run mariadb thrpt 50 71755.705 ± 2701.621 ops/s
|
Do_1000_params.text_readahead mariadb thrpt 50 3813.812 ± 63.488 ops/s
|
Do_1000_params.text mariadb thrpt 50 3779.422 ± 93.966 ops/s
|
Select_1.run_readahead mariadb thrpt 50 47460.177 ± 1725.788 ops/s
|
Select_1.run mariadb thrpt 50 61910.698 ± 1739.325 ops/s
|
Select_1000_Rows.text_readahead mariadb thrpt 50 5463.268 ± 112.607 ops/s
|
Select_1000_Rows.text mariadb thrpt 50 5449.366 ± 122.949 ops/s
|
Select_100_cols.text_readahead mariadb thrpt 50 10984.614 ± 171.044 ops/s
|
Select_100_cols.text mariadb thrpt 50 11185.643 ± 151.837 ops/s
|
useReadAheadInput permit fast retrieving big resultset, but for small queries, it slow down retrieval. This is due to socket inputstream call to available()
setting useReadAheadInput as false as default will permit better default.