-
Type:
Task
-
Resolution: Unresolved
-
Priority:
Major - P3
-
None
-
Affects Version/s: None
-
Component/s: Page deltas
-
Storage Engines - Transactions
-
110.161
-
SE Transactions - 2026-06-05
-
None
Problem
When a page is read into memory for disaggregated storage, tombstones are currently initiated to help build a delta that removes the corresponding key. However, this approach does not actually reduce the size of the stored data — it merely adds another key to the delta rather than eliminating one.
Proposed Change
Skip tombstone initiation when reading a page into memory in the disaggregated storage path. Since the tombstone does not contribute to delta size reduction and instead increases delta size, there is no benefit to creating it during page read.
Expected Outcome
Smaller page deltas for disaggregated storage, since unnecessary tombstone entries will no longer be written into the delta on page read.
- is related to
-
WT-17598 Batch per-update cache atomics in __wti_page_inmem_updates
-
- Closed
-