Thanks. Really good explanation :) How is the cube data stored in HBase? Can you consider a very small example (1Fact, 1Dim, 1Measure) and explain how the final data looks like in HBase?
On Tue, Mar 31, 2015 at 5:01 PM, jason zhong <[email protected]> wrote: > BTW, only when derived column exist will add foreign key to rowkey and > aggregation group. > > On Tue, Mar 31, 2015 at 5:11 PM, jason zhong <[email protected]> > wrote: > > > Kylin will pick up* foreign key* of lookup table,and* dimension column* > > as rowkey by default. > > > > and also add these items to aggregation group ,each group has *10 items* > > at most > > > > one relative hierarchy columns are treat as one item in aggregation > group > > . > > > > e.g. > > > > We have fact table TEST_FACT with dmensions *a,b,c,d,e,f,g* primary key > > on fact table is *id* > > and lookup table TEST_LOOKUP with dimensions *h,i,j,k* > > > > the rowkey item will contain *a,b,c,d,e,f,g,h,i,j,k,id* > > > > and aggregation groups will be like: > > group 1: > > a,b,c,d,e,f,g,id,h,i > > group 2: > > j,k > > > > the auto generate group may not be the best,so you need to adjust depends > > on your own business logic. > > > > > > On Tue, Mar 31, 2015 at 2:35 PM, Abhishek Sinha <[email protected]> > > wrote: > > > >> Thanks. Good explanation :) > >> > >> By default, Kylin picks up aggregation groups and rowkey automatically > in > >> the advanced setting. How is that done? > >> Is aggregation group and rowkey connected? > >> > >> On Tue, Mar 31, 2015 at 11:55 AM, Shi, Shaofeng <[email protected]> > wrote: > >> > >> > Aggregation groups is to reduce the cube calculation and storage; Lets > >> see > >> > a simple case: > >> > > >> > Say we have a cube has 6 dimensions: > >> > > >> > [A B C D E F] It will has 2^6 cuboids: 64; > >> > > >> > If your query pattern is: [group by A, B, C] or [group by D, E, F], > then > >> > you can split the 6 dimensions to two groups: > >> > [A B C], [D E F]; > >> > > >> > Then Kylin will skip some cuboid like AD, AE, Š; > >> > > >> > The total cuboid count will be 2^3 + 2^3 = 16, which is much less than > >> the > >> > original 64; > >> > > >> > On 3/31/15, 1:49 PM, "Abhishek Sinha" <[email protected]> wrote: > >> > > >> > >Went through the slide. Doesn't give much information about Advanced > >> > >Settings. > >> > > > >> > >On Tue, Mar 31, 2015 at 9:01 AM, jason zhong <[email protected] > > > >> > >wrote: > >> > > > >> > >> Hi Abhishek, > >> > >> > >> > >> Aggregation Group is for cube optimization, > >> > >> > >> > >> Kylin will pick up dimension column and *fk on lookup table* as > >> default > >> > >> rowkey and aggregation group item. > >> > >> > >> > >> you can refer > >> > >> http://www.slideshare.net/YangLi43/design-cube-in-apache-kylin > >> > >> to know more about cube creation first. > >> > >> > >> > >> and for kylin architecture > >> > >> > >> > >> > >> > >> > >> > >> > >> > > >> > http://www.slideshare.net/YangLi43/apache-kylin-deep-dive-2014-dec?next_s > >> > >>lideshow=1 > >> > >> > >> > >> > >> > >> > >> > >> On Mon, Mar 30, 2015 at 2:16 PM, Abhishek Sinha < > >> [email protected]> > >> > >> wrote: > >> > >> > >> > >> > Can someone explain the following while designing the cube in > >> Kylin? > >> > >> > > >> > >> > 1. New Aggregation Groups > >> > >> > 2. New RowKey Column > >> > >> > > >> > >> > By default, Kylin automatically picks up the some values for > RowKey > >> > >>and > >> > >> > Aggregation groups in Advanced Settings Page. > >> > >> > a) How is that done? > >> > >> > b) When do we need add our own Aggregation Groups and RowKey? > >> > >> > c) How is RowKey related to Aggregation Groups? > >> > >> > d) What does length, dictionary and mandatory indicate? > >> > >> > > >> > >> > -- > >> > >> > Abhishek Sinha > >> > >> > Mobile: +919035191078 > >> > >> > infoworks.io > >> > >> > > >> > >> > >> > > > >> > > > >> > > > >> > >-- > >> > >Abhishek Sinha > >> > >Mobile: +919035191078 > >> > >infoworks.io > >> > > >> > > >> > >> > >> -- > >> Abhishek Sinha > >> Mobile: +919035191078 > >> infoworks.io > >> > > > > > -- Abhishek Sinha Mobile: +919035191078 infoworks.io
