Ivan Bessonov created IGNITE-27198:
--------------------------------------
Summary: Optimize double PK index lookup
Key: IGNITE-27198
URL: https://issues.apache.org/jira/browse/IGNITE-27198
Project: Ignite
Issue Type: Bug
Reporter: Ivan Bessonov
Assignee: Ivan Bessonov
Currently we use a {{PeekCursor}} for calling "get" from hash index storage.
This means that in best case scenario we do two lookups to the index tree:
* Read {{{}RowId{}}}.
* Make sure that the next element is {{{}null{}}}.
This is problematic, because a simple implicit get only requires two actions:
resolve {{PK}} and resolve {{{}RowId{}}}. Currently it is 3 lookups to the
B+Tree that could be reduced to 2, which is substantial.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
