Best practices will improve performance with batch loads and large
updates.
For batch loads:
- Perform small- or medium-sized concurrent loads through the RLV store.
- Do not perform excessively large loads through the RLV store, because
loads exceeding maximum RLV memory may fail.
Even if a very large load does not exceed the maximum RLV memory (and
does not fail), merges of other committed RLV data may be triggered during the load.
This may impact load performance.
For large updates:
- Perform large updates on RLV-enabled tables with a table-level versioned
connection.
- Do not perform very large updates through the RLV store because
updates copy on-disk rows to the RLV in-memory store for modification.