Hello all, I am testing flink with a simple 1:1 table join with left and right tables having 1M rows, each row having ~2KB string in length. keys are just long variables. I am using rocksdb as the state backend.. if i run with rocksdb compression ON (the default setting - kSnappyCompression) the throughput drops to the floor (400 rows per second). However if i disable compression, throughput is a respectable 9000 rows per second. Are there any gotchas I need to be aware of?