SparkCube is an open-source project for extremely fast OLAP data analysis. SparkCube is an extension of Apache Spark.
mvn -DskipTests package
The default Spark version used is 2.4.4.
mvn test
There are several configs you should add to your Spark configuration.
| config | value | comment | |
|---|---|---|---|
| spark.sql.extensions | com.alibaba.sparkcube.SparkCube | Add extension. | Required |
| spark.sql.cache.tab.display | true | To show web UI in the certain application, typically Spark Thriftserver. | Required |
| spark.sql.cache.useDatabase | db1,db2,dbn | A list of database names separated by comma. Only tables and views from these databases will be considered for cube building. | Required |
| spark.sql.cache.cacheByPartition | true/false | To store cache by partition. | Optional |
| spark.driver.extraClassPath | /path/to/this/jar | For web UI resources. | Required |
With the configurations above set in your Spark thriftserver, you should be able to see "Cube Management" Tab from the UI of Spark Thriftserver after any SELECT command is run. Then you can create/delete/build cubes from this web page.
After you have created appropriate cube, you can query the cube from any spark-sql client using Spark SQL. Note that the cube can be created against table or view, so you can join tables as view to create a complex cube.
If you want a more detailed tutorial for cube creating/building/dropping etc., please refer to https://help.aliyun.com/document_detail/149293.html
(Slides)
https://www.slidestalk.com/AliSpark/SparkRelationalCache78971
https://www.slidestalk.com/AliSpark/SparkRelationalCache2019_57927
(Blogs)
https://yq.aliyun.com/articles/703046
https://yq.aliyun.com/articles/703154
https://yq.aliyun.com/articles/713746
https://yq.aliyun.com/articles/725413
(Blogs In English)