On startup, LogScale discovers CPU topology and creates one mapper pool per L3 cache zone within each NUMA node. This supports zone-aware query routing, which routes query work for a segment to a specific zone based on segment ID hashing.

Set this variable to -1 to disable CPU topology discovery. LogScale then uses a single thread pool instead of per-zone pools. Set it to a positive integer to further split each discovered zone into that many pools. For example, setting this to 2 splits each discovered zone into two pools.

Zone-aware routing is controlled by the ZoneAwareMapperPool feature flag, which is disabled by default. When disabled, LogScale submits work to the least-occupied pool. You can toggle the feature flag at runtime without restarting LogScale.

The number of threads dedicated to query work is set at startup and cannot change at runtime.