The preconditions is, the candidate cubes need match with SQL query,
including join conditions (inner, left), dimension, measures; If more than
1 cubes match the condition, Kylin will select the most efficient one, for
example: Cube A has 10 dimensions, Cube B has 20 dimensions, use A;

On 6/14/15, 6:53 PM, "Vineet Mishra" <[email protected]> wrote:

>Hi Shi,
>
>Wanted to know how "Kylin will pick the ³best matched² one (with a set of
>rules)" and moreover why does cube build job runs with single M/R job. Is
>it really necessary to have single mapper and reducer job or it can be
>tweaked as well, because the cube creation takes hell lot of time, around
>2:30 hrs which is really too much in future when the data grows!
>
>It there any work around to make the M/R to spawn multiple tasks to
>expedite the running process?
>
>Thanks!
>
>On Fri, Jun 12, 2015 at 12:49 PM, Shi, Shaofeng <[email protected]> wrote:
>
>> Only 1 cube will be picked up to server a query; If you have multiple
>> active cubes with the same source table, Kylin will pick the ³best
>> matched² one (with a set of rules); You can get the info in kylin.logs;
>> for example:
>>
>> QueryRouter.findBestMatchCube(QueryRouter.java:146)] - Find candidates
>>by
>> table Š cube1, cube2, ...
>> Š. Exclude cube2 as ...
>> findBestMatchCube(QueryRouter.java:185)] - Use weak matched cube cube1
>>
>> If you have preference on a certain cube, you can disable others, only
>> leave the preferred as ³active², then it will be the only one to server
>> SQL queries;
>>
>> On 6/12/15, 3:05 PM, "Vineet Mishra" <[email protected]> wrote:
>>
>> >Hi All,
>> >
>> >I was curious to know as when I am created multiple cubes for the same
>> >table as source and later I am referring the cube to get the cube data.
>> >
>> >Which table will it refer since our reference is still limited to
>>source
>> >table but the query is made from hbase table. So I wanted to have the
>> >understanding as how/what changes need to be done so as to assure the
>> >multiple cubes are formed and queries for the same table.
>> >
>> >Urgent Call!
>> >
>> >Thanks!
>>
>>

Reply via email to