3 ms·

Also HDFS. This bit me the first time I tried to write a Hadoop job that wrote every run to a new output dir.