I am interested in letting users perform aggregations on an arbitrary time range of past data. These aggregations will include averages, percentiles, and terms aggregations with multiple sub-aggregation-levels. The data is structured in a parent-child one-to-many relationship.
Performance-wise for aggregation queries, are there any downsides of using nested documents within a related parent doc compared to duplicating the parent doc fields and having a separate Elasticsearch document per child?