Details
-
New Feature
-
Status: Open (View Workflow)
-
Critical
-
Resolution: Unresolved
-
None
-
None
Description
Currently when a table uses partitioning and multiple partitions only the needed partitions are scanned in queries that have WHERE clauses on the partitioning key. However for other queries all partitions need to be scanned and furthermore the scans are done sequentially.
This task is to launch scans on different partitions in parallel instead of sequentially whenever more than one partition needs to be scanned (similarly to what the SPIDER engine does). This would allow the execution to use multiple cores (when available) and provide much faster overall execution times.
Attachments
Issue Links
- blocks
-
MDEV-28596 Replace Spider's parallel search feature by partitioning engine's one
- Open
- is blocked by
-
MDEV-6897 Making Items reentrant
- Open
- is part of
-
MDEV-22162 Make partitions more flexible to use
- Open
- relates to
-
MDEV-16610 Torn reads from spider, with XA and isolation level REPEATABLE READ
- Open
-
MDEV-6096 Ideas about parallel query execution
- Open