We are running a multi-tenant elastic search cluster, with tenant per index. To provide a better client experience and to avoid migration pains later, I am trying to compare Integer Vs. Long field storage. I am trying to get a better understanding of the advantages and disadvantages of using Long in place of Integer values.
I experimented by creating an index holding 10 Long fields and index with 10 Integer fields, with strict mapping. Indexed 3M+ documents with both indexes holding same documents, i.e., even the long fields contain int values but internally it's mapped to Long.
Storage size wise there was no difference as noted in this documentation https://www.elastic.co/guide/en/elasticsearch/reference/current/number.html#_which_type_should_i_use
I could not find any evidence that there is a degraded performance on search/query and, memory footprint was the same as well. I have tried the bucketed search and also searches by disabling cache.
Question: What are the scenarios in which, search performance degrades or memory footprint increases when using Long instead of Integer? How can I simulate this scenarios?