We run aggregations fully distributed and when iterating over the values we heavily rely on the field-caches. They hold the values of the latest used fields in

We run aggregations fully distributed and when iterating over the values we heavily rely on the field-caches. They hold the values of the latest used fields in memory and therefor allow in-memory performance on them. for example they don't grow linearly with the amount of rows stored, but depend on the cardinality of the fields. Running aggregations over non-indexed data is not supported.