This document discusses aggregations in Druid. It explains that aggregations involve intermediate aggregator functions (IAFs) that aggregate input rows during indexing and combining aggregator functions (CAFs) that merge the outputs of IAFs during querying. It notes that IAFs operate on rows while CAFs operate on the outputs of IAFs. It also discusses the use of serializers and deserializers when reading from and writing to disk during indexing and querying, and mentions that intermediate and indexing types can be different for complex metrics.