-
Type: Improvement
-
Resolution: Done
-
Priority: Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
Labels:None
-
Execution Team 2021-02-22, Execution Team 2021-03-08
In many-writer, large-batch timeseries workloads, a great deal of time is spent waiting on the mutex while we do read-only operations on shared data structures. We can introduced a striped (i.e. sharded, partitioned) mutex to support this in a way that allows multiple concurrent readers, but still allows an operation to take an exclusive lock (by locking all stripes) to perform a write.
- is depended on by
-
SERVER-54335 investigate insert performance for TSBS
- Closed