kylin-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From vu...@viettel.com.vn
Subject Query time hierarchy and time range in Kylin cube.
Date Fri, 18 May 2018 02:55:59 GMT
My current approach to time hierarchy and partition is as follow:

-          I have partition date and hour column in Hive table to avoid full
Hive table scan. Column names are partition_date, partition_hour

-          I have separate fields in fact table name Year, Month, Day, Hour
and use these column as hierarchy dimensions in Kylin cube build. I use
dictionary encoding.

-          When I want to query time range, I have to list all combination
of time hierarchy dimensions, for example (Month, Day), in order to query. 

My query seems to be slower when the cube get bigger with the same time
range. So I want to ask the best practice to design time hierarchy and query
time range in Kylin. I see some support for timestamp in Streaming cube but
I don't see guideline for design time dimension for normal cube except
partition date and hour in Hive.

I also suspect that my time range query get slower because it currently need
to scan all segment. 

I think we need 

 


Mime
View raw message