In my current position, HDFS is used widely. Unfortunately, the current method for connecting (I believe JDBC) isn't as performant as other options out there. Would love to see bulk read/write options for HDFS, potentially using something like SQOOP.