I want to measure the time taken by the MapReduce framework to read data from HDFS using counters. The current counters don't include that. I have looked at the MapReduce source code on Github but I am still unable to wrap my head around the whole flow. Looking at the source, I think read operation is done from the below java class and
I know I can have an empty Mapper approach to do this but since I will be running lot of experiments then that will not be very ideal.
Any pointers is highly appreciated.