Hi, Impala does achieve much better performance compared with hive. If source meta can be storaged in impala, we can investigate the query performance with both Kylin and impala, which is more likely what we had recently done.
[email protected] From: Sarnath Date: 2015-05-28 11:49 To: dev Subject: Re: Re: Apache Kylin versus Cloudera Impala Sorry for intruding...I am not sure how one can compare Impala with Kylin. Though Impala can create Cubes - it does not use that to serve SQL queries - in my understanding. (or) Is there something I am missing? On Thu, May 28, 2015 at 8:38 AM, [email protected] <[email protected]> wrote: > Hi , Luke > > Thanks for your prompt advice. > > Yeah, the pre-calculation cube is an obvious advantage for OLAP engine. > However, cube build cost > > may need to be leveraged over runtime scan / aggregation. We are expecting > the performance results > > between Kylin and MPP sql engine like Impala, and we are just about to do > that. Would share more > > valuable information AFAIK. > > > > [email protected] > > From: Luke Han > Date: 2015-05-28 10:30 > To: [email protected] > Subject: Re: Apache Kylin versus Cloudera Impala > Hi Sun, > There's no benchmark from our side yet,especially in prod env. I'm also > very curious to know if someone did such comparison. > > The direct advantage for Kylin over Impala (include other MPP > solution): > 1. Non-Invasive Design: you do not need to install any agent, library > or others in your existing Hadoop Cluster (Neither on Namenode or DataNode) > 2. Pre-Calculation result avoid runtime scan/aggregation, that mean you > could get result more faster in seconds latency over billions data. > > > Thanks. > > Luke > > > Best Regards! > --------------------- > > Luke Han > > 2015-05-28 10:20 GMT+08:00 [email protected] <[email protected]>: > > > Hi, team > > > > Really interested in the performance comparison and also the native > design > > advantage over Apache Kylin > > > > and Cloudera Impala. As the official saying, Cloudera Impala is a > > "Lightning-fast, distributed SQL queries > > > > for petabytes of data stored in Apache Hadoop clusters". Kylin can goes > to > > 10-1000x query efficiency over > > > > hive in the usage of MOLAP, while Cloudera Impala can also achieve much > > more performance upgrade over > > > > hive. > > > > Question is : Does Kylin do some benchmark test or performance comparison > > with Cloudera Impala in production > > > > environment? What can be the direct advantage for Apache Kylin over > > Cloudera Impala? > > > > If anyone had deployed and used both products in your usage, please > kindly > > share any available suggestions. > > > > Best regards, > > > > Sun. > > > > > > > > [email protected] > > >
