Hello, I have a Flink Application on K8s (using Flink K8s operator) with ~100 operator, 200 parallelism producing 30TB of state on S3. On restarting an application, the checkpoint from S3 is restored pretty quickly but the full checkpoint takes a long time to complete. We've noticed that the Network Out to S3 is very low compared to the Network IN. Experimenting with the following features in

flink-conf.yaml

did not produce expected results

s3a.attempts.maximum: "25"
  s3a.threads.max: "512"
  s3a.threads.keepalivetime: "240"
  s3a.connection.maximum: "2048"
  s3a.multipart.size: "100M"
  s3a.fast.upload.active.blocks: "8"
  s3a.multipart.purge: "true"

Any recommendations to expedite the checkpointing to S3? TIA