Thanks Kurt, that answers my question. @nandan, id, timestamp ensures unique primary-key.
On Mon, May 21, 2018 at 2:23 PM, kurt greaves <k...@instaclustr.com> wrote: > Every column will be retrieved (that's populated) from disk and the > requested column will then be sliced out in memory and sent back. > > On 21 May 2018 at 08:34, sujeet jog <sujeet....@gmail.com> wrote: > >> Folks, >> >> consider a table with 100 metrics with (id , timestamp ) as key, >> if one wants to do a selective metric read >> >> select m1 from table where id = 10 and timestamp >= '2017-01-02 >> :00:00:00' >> and timestamp <= '2017-01-02 04:00:00' >> >> does the read on the specific node happen first bringing all the metrics >> m1 - m100 and then the metric is sliced in memory and retrieve , or the >> disk read happens only on the sliced data m1 without bringing m1- m100 ? >> >> here partition & clustering key is provided in the query, the question is >> more towards efficiency operation on this schema for read. >> >> create table { >> id : Int,. >> timestamp : timestamp , >> m1 : Int, >> m2 : Int, >> m3 Int, >> m4 Int, >> .. >> .. >> m100 : Int >> >> Primary Key ( id, timestamp ) >> } >> >> Thanks >> > >