[Perf] Suggested data format for batch Ingestion

I see at : http://druid.io/docs/latest/ingestion/data-formats.html that Druid supports:

JSON, CSV, or a delimited form such as TSV, or any custom format
which is great.

Given these options, is there a recommendation/comparison matrix on which format to use for the fastest ingestion?



I’m not sure if anyone’s benchmarked it, but I’d guess they’re roughly similar speeds. TSV should be fastest since it’s simplest, although our JSON parser is more optimized than our TSV parser, so the result there may be “surprising”. But again, I haven’t tested it, so I’m not sure.