Hi, Am I right that combiners are supposed to return key/value types that reducers expect as an input?
Lets say I have map reduce operation to calculate number of different ip that visited a resource i have log name1 ip1 name2 ip2 name1 ip2 ... map produces pairs - (Text resourceName,Text ip) reducer produces pairs (Text resourceName, IntWritable numberOfVisists) Am I right that combiner should return (Text resourceName, Text ip) pairs if so what can be optimized in combiner beside removing repeated ips (if removing will give any benefit at all). Thanks in advance, Vitaliy S
