hi all! I think hao hit the point, and I also think decouple job like data from kylin metadata(model, cube) is a good choice for kylin!
best regards wangxianbin1...@gmail.com From: Hao Chen Date: 2016-02-25 15:36 To: dev Subject: Re: [jira] [Created] (KYLIN-1434) Kylin Job Monitor API: /kylin/api/jobs is too slow I think the solution KYLIN-1079 can resolve the problem by some way, though is not a perfect solution. The current row-key design of resource persistence is not very good for timeseries based data like kylin job history data, metric and so on, and the query performance and flexibility are not fast and scalable as well. Maybe it possible to consider another kinds of row-key design for such kinds of ts data for better performance and scalability like what we dose in Eagle, and what's more, kylin will never need to backup/restore such kinds of data as well like metadata so that it should be definitely decoupled from metadata store. Regards, Hao On Tue, Feb 23, 2016 at 5:13 PM, hongbin ma <mahong...@apache.org> wrote: > please check whether it is a duplicate with > https://issues.apache.org/jira/browse/KYLIN-1079 > > On Tue, Feb 23, 2016 at 3:28 PM, Hao Chen (JIRA) <j...@apache.org> wrote: > > > > > > > -- > Regards, > > *Bin Mahone | 马洪宾* > Apache Kylin: http://kylin.io > Github: https://github.com/binmahone >
