Hello Raj,

On Fri, Apr 29, 2011 at 5:52 AM, Raj V <[email protected]> wrote:
> I am trying to understand the relationship,if any,  between HDFS_BYTES_READ 
> and FILE_BYTES_READ. Since HDFS is laid on top of the local file system is 
> FILE_BYTES_READ include the bytes read by HDFS_BYTES_READ.. For example in my 
> word count program
> HDFS_BYTES_READ is 1,000,000,415,344 and FILE_BYTES_READ is 679,397,984,899. 
> Does it mean that  679,397,984,899 bytes were read by the data local map jobs 
> or are the two counters completely independent of each other?

They are individual counters, independent of each other. Although HDFS
may, in lower level terms, be a layer atop the 'local filesystem', the
MapReduce components do not see it in such a granular manner.

-- 
Harsh J

Reply via email to