Hello. I started to process video in hadoop using a open source lib HVPI, it's uses xuggler, but the VideoInputFile implementation the method isSplitable return a false. In resume to put it to work with files bigger than a HDFS block i divide the video file using ffmeg and send it into small chunks to HDFS. For now it's work so slowly. The application divide the video's file into frames, apply some operation in the frames (map phase) and recreate the video (reduce phase). I have 5 machines in the cluster spending more than 4 hours to process a 1GB video file. I have two question: first how can i solve the boundary problem in HVPI library ? How increase the cluster speed?