Konstantin Orlov created IGNITE-28437:
-----------------------------------------
Summary: Sql. Reduce number of rows fetched during ordered index
scan
Key: IGNITE-28437
URL: https://issues.apache.org/jira/browse/IGNITE-28437
Project: Ignite
Issue Type: Improvement
Components: sql ai3
Reporter: Konstantin Orlov
Assignee: Konstantin Orlov
Currently, we use constant {{Commons#SORTED_IDX_PART_PREFETCH_SIZE}} to define
number of rows to prefetch from each partition when scanning the sorted index.
Such an approach doesn't scale well when number of local partitions growth
because we read more rows from storage than required, every read creates
additional contention on storage-related structures, excess rows occupy memory,
etc.
It's proposed to switch to value calculated based on the number of local
partitions involved into query processing.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
