Does Cassandra have something like HBase' coprocessor which can do server side processing? This is very important for Kylin's query performance (to push down the filtering and aggregation logics to clusters); otherwise there will be too much data be transferred back and aggregated in Kylin. I remember there wasn't an equivalent technology at that time, so we didn't put effort on it.
2016-09-13 15:48 GMT+08:00 Billy(Yiming) Liu <[email protected]>: > https://community.mapr.com/docs/DOC-1558 > > There is a document introducing run Kylin on MapR. HBase is the mandatory > component. > > 2016-09-13 14:43 GMT+08:00 Something Something <[email protected]>: > >> One of my clients uses MarR distribution & they don't use HBase. We tried >> using Kylin with MapR tables & it didn't work. >> >> We could write & contribute but here's my question: >> >> It appears, the HBase key is this: >> >> <dim1>+<dim2>+....+<dimN> => values of measures for this combination, >> correct? >> >> Which means every query (except when all DIMs are grouped) would require >> scan, correct? >> >> How does Kylin do the HBase key lookup? I mean, if query requires >> grouping by only 2 DIMs, <dim1> & <dimN>, how is 'lookup' & 'scanning' >> handled? >> >> >> >> On Mon, Sep 12, 2016 at 6:28 PM, ShaoFeng Shi <[email protected]> >> wrote: >> >>> Cassandra isn't supported, as we don't see stronge need for that; if you >>> have the case, please share the scenario, thanks! >>> >>> Besides, we welcome contribution from the community, if you have the >>> willing to implement this, please also let us know. >>> >> >> > > > -- > With Warm regards > > Yiming Liu (刘一鸣) > -- Best regards, Shaofeng Shi
