- Calls the next() method for the row store cursor, as opposed to continuously seeking, in order to reduce time taken during COLUMN_SCAN
- Processes records in batches. For each batch, we consider a sample size, and if that sample size reaches a certain amount of row store calls, we automatically defer to row store for the rest of the batch.
- is depended on by
-
SERVER-69105 Achieve column index performance acceptable for release
- Closed
- is duplicated by
-
SERVER-67142 Add heuristic for avoiding column store when the row store fallback is used often
- Closed