Will there be too much performance drop if client insert records row by row？ due to many rpcs？
At 2022-06-29 21:31:02, "Mingyu Chen" <morning...@163.com> wrote: >Hi Meiyi, >Thanks for you proposal. >And for "buffer the data in the server side(BE)", Doris already support the >"begin; insert; insert; insert; ...; commit" method, >which is also a kind of "buffer data in server side" feature for "insert into >values" operations. > > >Maybe you can explain more about how it works with stream load and how to use >it. >Please register an account here[1] and tell me your account name, I will >create a DSIP for this proposal. > > >[1] https://cwiki.apache.org/confluence/display/DORIS/Home > > > > >-- > >此致！Best Regards >陈明雨 Mingyu Chen > >Email: >morning...@apache.org > > > > > >At 2022-06-29 15:54:34, "梅祎" <myime...@gmail.com> wrote: >>Hi all, >> >>In current data import ways，users generally use StreamLoad to push >>real-time data to Doris, which need to buffer multiple records in the >>client side, and then send them to Doris together. It is pretty complex for >>users. >> >>We propose to buffer the data in the server side(BE) and commit >>automatically so that it is transparent to users and users can insert >>records row by row. >> >>I'll add detailed design later.
