Repeat my comment in KYLIN-1049 to elaborate the low priority. "Not sure the priority of this. Most big data OLAP we see is happy with star schema, typically transformed from online snowflake schema by a ETL process. Star schema has obvious advantage in terms of simplicity and performance for big data. Even more so for streaming analysis cases."
On Tue, Sep 29, 2015 at 8:21 PM, ShaoFeng Shi <[email protected]> wrote: > Hi Tero, there was some discussion about snowflake schema in the mailing > list before, like this > < > https://mail-archives.apache.org/mod_mbox/kylin-dev/201504.mbox/%3ccahrce1ngiggi82udtaeyeosep1kjkserrproqfowh0jz5k_...@mail.gmail.com%3E > >; > but > it is still not on kylin's roadmap; the main reason is we don't see many > such cases, or they can be convert to star schema; If you can share some > M-to-N and snowflake scenario, that would help. > > 2015-09-29 4:15 GMT+08:00 Tero Paananen <[email protected]>: > > > re: https://issues.apache.org/jira/browse/KYLIN-1049 > > > > I just created this ticket. > > > > We've been evaluating several OLAP on Hadoop type tools for some time, > > and more or less found that very few tools support many-to-many > > relationships either at all or the support is buggy. Snowflake schema > > support is a little bit more common. > > > > Our data model(s) is such that it requires snowflake schema and > > many-to-many relationship support. We simply can not use OLAP tools > > that have no support for it. > > > > There's a possibility we could implement this on our own, but that's > > obviously depending on all kinds of things some of which aren't under > > my control. Even so, we'd need some help at the very least to get > > started. > > > > To that end I was hoping I could discuss what it would take to > > implement this with someone who does and whether this is something > > that's already on your roadmap. > > > > -TPP > > >
