Hi Ashwin,
Robert Metzger

Hi Robert, I understand how the hash-joins work, but I was curious to know if the connector could actually push down all the “left side filters” down as predicates into the right side connector.

I suppose Flink does not yet do predicate pushdown and so it ends up pulling most of the data out of the Filesystem-like or Hadoop interfaces into memory and then does the hash-join in memory across multiple Flink nodes.


One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.