Details
-
Task
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
None
-
2025-4, 2025-5
Description
As of now the major bottleneck for QA is data retrieval from MDB, namely CES retrieves innodb data using only one client connection. Given that MCS query execution is a pipeline, its performance is limited by the slowest part.
This issue describes the way to resolve this single-threaded data retrieval. There are at least two feasible ways:
- Innodb table partitioning coupled with optimization query re-write that turns single data retrieval CES into UNION of partition-specific CESes
- engine-independent statistics, namely mysql.column_stats table contents
Attachments
Issue Links
- includes
-
MCOL-4935 A prototype planner/optimizer using term rewriting
-
- Open
-
-
MCOL-6036 Rewrite optimizer engine + CES parallel read rewrite rule
-
- Closed
-
- is blocked by
-
MDEV-36922 Adding histogram data retrieval method for Histogram_json_hb
-
- Closed
-
- relates to
-
MCOL-6013 To be Scheduled !!! Query Accelerator | over shared memory data retrieval
-
- Open
-
- mentioned in
-
Page Loading...