Branch "kylin-on-parquet" has implemented and it's ready for public review and evaluation, the issue is https://issues.apache.org/jira/browse/KYLIN-3621.
However, as the discussion in https://sematext.com/opensee/m/Kylin/8WImho86LC1RxadI?subj=Evaluate+Kylin+on+Parquet, to make it more stable and with higher performance, some tasks of improvement should be done.
The key tasks may be:
1. Calculation of queries on Kylin should be all distributed
2. Remove dictionaries for dimensions
3. Store value with its data type(not binary)
4. The new build engine with Spark-SQL