Hi,

Is there any best approach to reduce shuffling in spark. I have two tables
and both of them are large. any suggestions? I saw only about broadcast but
that will not work in my case.

Thanks,
Asmath

Reply via email to