ha_innobase::records_in_range() accesses the index records actually

Bug #724734 reported by Yasufumi Kinoshita
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Percona Server moved to https://jira.percona.com/projects/PS
Triaged
Wishlist
Unassigned

Bug Description

ha_innobase::records_in_range() accesses the bounds (start and end) of the records at the range actually.
Though it is more accurate than using something stored statistics about the index,
it is very expensive for complicated queries which have very many candidates of query plans.

If ha_innobase::records_in_range() uses also something stored statistics before,
we can choose behavior by option according to the system characteristics or the session characteristics.

The benchmark result about complicated queries (like TPC-H) should become better.

Changed in percona-server:
importance: Undecided → Wishlist
summary: - ha_innobase::records_in_range() accesses the table records actually
+ ha_innobase::records_in_range() accesses the index records actually
Revision history for this message
Baron Schwartz (baron-xaprb) wrote : Re: [Bug 724734] [NEW] ha_innobase::records_in_range() accesses the index records actually

I think this would be a great feature. I see many cases where queries
spend a lot of time in the "statistics" state, and are waiting in the
InnoDB queue. If the stored statistics can be accessed without
entering the InnoDB kernel, then this could be significantly relieved.

Revision history for this message
Yasufumi Kinoshita (yasufumi-kinoshita) wrote :

Yes, if already the statistics were stored, only dict_index_stat_mutex is needed to obtain through the "statistics" state.

Changed in percona-server:
assignee: nobody → Yasufumi Kinoshita (yasufumi-kinoshita)
Revision history for this message
Stewart Smith (stewart) wrote :

also one of the many candidates in server for RCU.

Changed in percona-server:
status: New → Triaged
Revision history for this message
Stewart Smith (stewart) wrote :

Yasufumi, is this on your immediate radar?

If not, let's leave it unassigned.

(trying to have your assigned bugs and tasks be what immediately looking at rather than catch-all)

Changed in percona-server:
assignee: Yasufumi Kinoshita (yasufumi-kinoshita) → nobody
Revision history for this message
Shahriyar Rzayev (rzayev-sehriyar) wrote :

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PS-2350

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.