I'm working on an effort to make Build Engine / Table Source / Storage more
plugin-able. The benefit is, for example, Spark can be an alternative build
engine in addition to MR, or use Spark RDD as table source, or implement
Cassandra as storage with minimal effort.

https://issues.apache.org/jira/browse/KYLIN-875

The current source modules will be refactored. The future might look like

- "core" module, merged from common, cube, dictionary, metadata, job
- "engine-mr" module, the default MapReduce build engine
- "source-hive" module, Hive table source
- "storage-hbase" module, HBase as default storage
- "query" & "server", as is

Later alternative engine / source / storage can be added.

- "engine-spark"
- "source-sparksql"
- "storage-cassandra"

Any thoughts? The change will happen on 0.8 branch gradually.


Cheers
Yang

Reply via email to