Two additional counters shouldn't be expensive. But do consider if the
already available "map input records" / "(map|reduce) output records"
suit your need for such counters.
On Sat, Mar 19, 2011 at 4:33 AM, Andreas Paepcke
<[EMAIL PROTECTED]> wrote:
> 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?