You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We have 2 optimizations which improve the performance of distsql read, one is the batch coprocessor and another is the load-based replica read. But they are currently not compatible, because sending batched tasks to the replica store will produce region misses.
As an enhancement, the batched task should be re-split before being sent to the replica nodes. It might choose the best replica node in the coprocessor client.
The text was updated successfully, but these errors were encountered:
Enhancement
We have 2 optimizations which improve the performance of distsql read, one is the batch coprocessor and another is the load-based replica read. But they are currently not compatible, because sending batched tasks to the replica store will produce region misses.
tidb/store/copr/coprocessor.go
Lines 1126 to 1132 in fa28db1
As an enhancement, the batched task should be re-split before being sent to the replica nodes. It might choose the best replica node in the coprocessor client.
The text was updated successfully, but these errors were encountered: