+1 here The issue is that we need to de-couple network buffer capacity from parallelism (for horizontal scaling), which https://nightlies.apache.org/flink/flink-docs-stable/docs/ops/batch/batch_shuffle/#performance-tuning recommends in step 3. cc @Jason Li

@Han You @Paul Thomas Would one of you be willing to raise a FLIP and a PR to reintroduce these configurations to Flink?

@Alex sorry I haven't logged onto this Slack for very long! But yes, I will