kylin-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Li Yang <liy...@apache.org>
Subject Re: Using one-to-many/many-to-many type relationships in Kylin
Date Mon, 02 Mar 2015 02:24:24 GMT
The "Second solution" is more generic and practical. It deals well with
different sizes of multi-values. True the fact table (or view) is exploded,
but the final cube will be compact as column values are dictionary
compressed.

On Fri, Feb 20, 2015 at 1:06 PM, ć‘šćƒæ˜Š <z.qianhao@gmail.com> wrote:

> Hi, Paananen,
>      The first step is to create an intermediate hive table.
>      Basically it execute some hql, you can see the exact hqls by click
> the little icon (names parameter look like a key) in the job detail. And
> then you can execute the hql through hive through hive command line to
> see if anything goes wrong.
>
> On Thu Feb 19 2015 at 2:08:10 AM Tero Paananen <teropaananen@gmail.com>
> wrote:
>
>> On Sun, Feb 15, 2015 at 1:22 AM, Zhou, Qianhao <qianzhou@ebay.com> wrote:
>> > Hi, Paananen,
>> >     Can you tell us about
>> >     1. how many distinct value count in your multi value property
>>
>> It varies a lot. We have some entities that have none, and some that
>> may have thousands.
>>
>> Normally, though, it's probably 3 - 6 values.
>>
>> For our evaluation purposes we could ignore the thousands of values cases.
>>
>> >     2. Which exactly step of the job is stopping you from the cube build
>>
>> It fails on the very first step.
>>
>> The Kylin log lists the job name as:
>>
>> Starting to launch local task to process map join;
>>
>> -TPP
>>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message