I have 100M item-timestamp pairs for 10M items in index and I would like to find percentiles of item lifetimes.
With aggregation I get 10M buckets and using percentiles_bucket I can aggregate buckets into few percentiles.
The problem is that ES outputs few hundred Mbytes of buckets and finally the wanted percentiles at the end of the output. How can I reduce the output just to precentiles? I don't need the 10M mid-buckets but only the final result of precentiles.
Please see http://pastebin.com/hCJ9xLir
Regards, Markus Paaso