I have a bit of a dilemma going here...
We are indexing 2 types of payloads in our system -- one contains xml request file (data) coming in from ERP connectors containing tax information that gets fed into our back end tax calculation engine which then outputs the tax calculations in the form of an xml response file that gets sent back to the originator. Both files have a couple of pieces of data in common one being a payload UID and another being a unique identifier for the client so what we usually do is use the UID to marry the Request & Response together by going in through either Kibana or writing a Lucene query to search the individual indices.
Not horrible but does take extra time so we were considering merging the two into a singe index with the UID as the unique identifier but I was just reading that the mapping type is being phased out and on top of that Elastic recommends as a best practice not to do this type of approach because it effects Lucene's ability to compress efficiently.
All of this being said what is the best approach using Release 6.0 changes and improvements to accomplish what we want to?