hadoop-mapreduce-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Pedro Costa <psdc1...@gmail.com>
Subject Cardinality of ReduceCopiers to map outputs
Date Wed, 11 Aug 2010 12:51:59 GMT

1 - I would like to know if a Map Task can produce more than 1 map
output per execution?

2 - A Map Task can't be reused, right? When a Map Task instance
produced a map outputs, this instance will end, right?

3 - I would like to know if a ReduceCopier will fetch only 1 map
output at a time and saves it in a file or in memory before merging
them? For example, if we've a ReduceCopier that fetches 4 map outputs,
the reduce will save the 4 outputs in separate and only after all map
outputs are copied to the reduce side, that the ReduceCopier will
merge them, right?



View raw message