Hi, Thanks for your reply. Yes, I have checked that before, but after setting this configuration, if the UHC dimension is dict encoded, the process sticks on step 3 (extract fact distinct columns, I see three reducers and the third reducer sticks on 69.78). Therefore I choose the fixed-length encoding for the dimension and it works. I look forward to hearing your opinion.
On Mon, Apr 13, 2020 at 9:11 AM Xiaoxiang Yu <[email protected]> wrote: > Hi, > Did you ever check this "Build the dict for UHC column with MR" > https://issues.apache.org/jira/browse/KYLIN-2764? Can this feature solve > your problem? The configuration can be found at > http://kylin.apache.org/docs/install/configuration.html#uhc-config . I am > glad if you can share us your opinion on this feature. > > > > > > -- > *Best wishes to you ! * > *From ：**Xiaoxiang Yu* > > > At 2020-04-11 17:26:39, "Fatemeh Ahmadi" <[email protected]> wrote: > > Dear Sir/Madam, > > I am using Apache Kylin and I have multiple real UHC dimensions > (unfortunately I can not remove them) and because of that, the query > building process requires a lot of time. Do you have any recommendations? > I am thinking of using near-real-time OLAP (because it does not have the > first two steps in batch processing) and building multiple cubes > simultaneously in short periods. But even in this case, I have to keep > user-id (which is a UHC dimension) for each cube and one other UHC > dimension as well. Do you think this solution is practical? > If you have any opinion about this problem I would deeply appreciate it. > > yours sincerely, > Fatemeh > >
