-
Type: Improvement
-
Resolution: Unresolved
-
Priority: Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
5
-
2023-05-16 Chook-n-Nuts Farm, 2023-05-02 StorEng Bug Bash, 2023-05-30 - 7.0 Readiness, StorEng - 2023-06-13, 2023-06-27 Lord of the Sprints, 2023-07-11 WiredTractor
We recently encountered a case where a table had a lot of small on-disk pages (containing about 600bytes of data), with an allocation size of 4kb.
That means that the on-disk space usage is inefficient, using 4kb for every 600 bytes of data. It also leads to the cache tracking in WiredTiger being inaccurate. Since the cache tracking is based on the data size, but block sized memory is allocated.
This ticket should change how we manage and/or track the memory so that cache tracking is more accurate.
- causes
-
WT-11320 Perf degradation in multi-btree-zipfian read throughput
- Closed
-
WT-11276 Internal pages not being queued for eviction
- Closed
- has to be done before
-
WT-10832 Investigate reconciliation split logic not creating reasonably sized pages
- Open
- related to
-
WT-11322 Copy blocks smaller than the chunk size to properly sized buffer
- Open