The global dictionary is very large, and in our application a global dictionary has a size of 16 gigabytes, which will increase the time to build a dictionary, generate cuboidData, and increase the query time. Therefore, it is recommended to use cube for differentiation.Related issues:https://issues.apache.org/jira/projects/KYLIN/issues/KYLIN-3312?filter=allopenissues
------------------ ???????? ------------------ ??????: "Billy Liu"<[email protected]>; ????????: 2018??4??4??(??????) ????1:30 ??????: "user"<[email protected]>; ????: Re: The global dictionary does not specify the cube distinction. The global dictionary is designed to be shared among Cubes. So deleting the cube will not remove the global dictionary is as expected. If you want to remove the global dictionary, please use storage clean up tools. http://kylin.apache.org/docs23/howto/howto_cleanup_storage.html With Warm regards Billy Liu 2018-04-04 9:37 GMT+08:00 ?????? <[email protected]>: > In the cube design, the COUNT_DISTINCT indicator USES the global dictionary. > When building a cube, a dictionary file is generated at HDFS, and the > corresponding directory is: > "/ resources/GlobalDict/dict/database. The tableName/column/", at the same > time will generate metadata information in hbase metadata, metadata in the > hbase rowkey information:"/dict/database. The tableName/column/". The value > information specifies the corresponding HDFS path. > The problem is that the corresponding HDFS data and hbase data are not > deleted when the cube is deleted. Another HDFS data path and hbase metadata > rowkey information did not specify the concrete cube, if a table column in > more than one cube is to use global dictionary, can lead to multiple share a > dictionary data cube
