Kylin architecture is more focus on extreme large data set (e.g. billion+). In 
OLAP domain, user don't care about 100+ ms query latency and only hope to keep 
the big query good enough. If you want to compare the performance, please try 
some very large dataset that can't be kept in memory (e.g. In eBay, the data 
set is 100+billion). Besides query latency, you also need to test query 
throughput that is another side of performance.


Another thing about the benchmark, the data schema and cube design is crucial 
to the performance. Good or bad cube design will have very different 
performance result. 


Thanks
Jiang Xu
  
------------------ ???????? ------------------
??????: "Li Yang";<[email protected]>;
????????: 2015??11??16??(??????) ????3:30
??????: "dev"<[email protected]>; 

????: Re: Group by + where clause



The SQL overhead is around hundreds of ms, so don't worry about the ms
level difference.

I'm more interested in the HBase/ES comparison on big data load. HBase
design incorporates random read/write, while ES can optimize for batch load
and read-only. In that sense, HBase is born with a small burden.

Many have suggested storage other than HBase. Here we have another one ES.
 :-)

I think that's what Kylin has plugin architecture. Anyone can implement a
new storage plugin and run Kylin on anywhere if not HBase.

On Mon, Nov 16, 2015 at 2:31 PM, Sarnath <[email protected]> wrote:

> On Nov 16, 2015 5:07 AM, "Luke Han" <[email protected]> wrote:
> >
> > Hi Sarnath,     It's interesting result, could you please share your full
> testing case, data and result?
>
> Thanks. I will post the result this week. Will need to cross check
>
> > And could you please share your cube engine's design even code(is it open
> source?)? So that we also could know how the comparison goes to and help
> you to understand more.     Thanks.
>
> This is closed software. But we don't do any smart stuff. Rather straight
> forward...thanks!
> >
> > Regards!
> > Luke Han
> >
> >     _____________________________
> > From: Sarnath <[email protected]>
> > Sent: ??????, ?????? 16, 2015 01:06
> > Subject: Re: Group by + where clause
> > To:  <[email protected]>
> >
> >
> > A small correction: I meant data block encoding on hbase
>

Reply via email to