I manually set the algorithm in kylin.properties to run the sample kylin_sales_cube, and compare it to the layer one, it truely 2.5 fast than layer cubing.
If the in-mem cubing is only available from v1.5.0 , why there some source code about in-mem cubing shows in v 1.3.0 ? I suggest that u can take some comments there to mark that part will not release in the particular version. 2016-04-06 16:09 GMT+08:00 ShaoFeng Shi <[email protected]>: > in-mem cubing (also called "fast cubing") is only available from v1.5.0; > > you can check two blogs about "Fast cubing" in > https://kylin.apache.org/blog/ > > 2016-04-06 15:20 GMT+08:00 Mars J <[email protected]>: > >> Hi, >> Does the inmem cubing algorithm released in v1.3-hbase1.1.3 ? >> >> I guess the anwser is not, because I look up all the v 1.3 logs >> produced by what I execute, there is no something like 'The cube algorithm >> for...' which I found in the org.apache.kylin.engine.mr.steps (logger >> .info("The cube algorithm for " + seg + " is " + alg);) >> >> Is there some docs describ the inmem cubing algorithm ? Is this one ( >> http://www.infoq.com/cn/articles/apache-kylin-algorithm)? >> > > > > -- > Best regards, > > Shaofeng Shi > >
