I'm writing a load/store pair, and I am considering the use of two Hadoop counters (via reporter). They would could the number of records read/written, so that they would be available via the job WebUI.
Is this an overly expensive proposition? Thanks, Andreas
