Parquet Row Scan long wait time

I have a job that seems to be running fine, but a very long PARQUET_ROW_GROUP_SCAN and I’m not sure why. I’m running on K8s


Seems like the time is spent on receiving the blocks from the source to Dremio. Is this S3/Azure or HDFS? It could be possible the reads are slow of the target system. Can you please send us the profile and we can take a look?