My personal opinion. Choosing the right OLAP solution is not easy. There are different options for different scales.
If it is millions of rows, then RDBMS like MySql / PostgreSQL shall fly. If it is billions of rows, then Druid / ES / Kylin will all work given you get the right hardware and software configuration. If it is trillions of rows (or more), then Kylin has a big advantage thanks to its precalculation system. Read more about precalculation here <http://www.slideshare.net/YangLi43/apache-kylin-20-from-classic-olap-to-realtime-data-warehouse> . Cheers Yang On Fri, Apr 7, 2017 at 10:39 AM, lufeng <[email protected]> wrote: > Hi Luke > > I find a comparison between Druid that can provide sub-second OLAP > queries an and ES[1]. Maybe there are some identity of views between Kylin > and ES. > > Yes, compare with HBase that ES has no standout features to invest to this > path and Kylin do lots of internal optimization on HBase storage. > > I will glad to share the comparison to community when got some progress. > > Thanks for your reply. > > > [1] http://druid.io/docs/latest/comparisons/druid-vs-elasticsearch.html > > > > 在 2017年4月6日，下午5:19，Luke Han <[email protected]> 写道： > > ES is not a target storage for Kylin so far, at least not on coming > release plan. > > There are already many storage options in Hadoop Ecosystem, I don't think > there's strong reason to invest on this path. > > And I don't remember there's any benchmark or comparison available today > for your purpose. > Please share with us if you have chance to do it:) > > Thanks. > > > Best Regards! > --------------------- > > Luke Han > > On Wed, Apr 5, 2017 at 6:17 PM, lufeng <[email protected]> wrote: > >> Hi All >> >> Now if I want to build a OLAP platform to analyze our huge data, the >> first impression is using Hive，but it can not match our real-time query >> needs. Then I found lots of companies used ES to build their OLAP engine >> before like tencent’s Hermes. So I want to know that what’s the >> difference between ES and Apache Kylin from different dimensions like >> filter, aggregation queries or etc. >> >> * Performance >> * Flexible queries >> * Data Model >> * BI integration >> * … >> >> After Kylin release latest architecture that can support different >> storage engine, I found that Kylin will support ES[1] as it’s backend >> storage engine. So I think ES is ROLAP engine that lots of people use ES as >> NoSQL DB and Kylin is a MOLAP engine. >> >> I would love to hear some feedback for my confusions. >> >> Thanks. >> >> [1] https://github.com/apache/kylin/pull/23 >> > > >
