It just so happened that I did one not long ago, using hadoop+hive+elasticsearch to do BI for the company.

The process is like this:

Data from the generating end to send requests to the data collection server, the json text form exists on the disk of the collection server.

Timed task, through the MR Job will be cleaned and sorted data repair classification, cloud persistence, after the data as the basis of BI deposited into the hive.

Timed task, every day to run the hive sql (wrote a dynamic rendering of sql statements engine), update the cumulative table and analyze the results of the analysis of the specific individual indicators, deposited into the

The front-end with highcharts, jquery request back-end, pass some range constraints, back-end to get the constraints of the group out of the query statement, from the ES to take out the results of the conformity of the return to the front-end.

The front-end presentation, that is, categorized indicators, screening conditions, indicators within the charts and detailed data, as well as multi-criteria and multi-indicator horizontal and vertical comparisons, export and so on.