[prev] 38 [next]

MapReduce (cont)

MapReduce makes use of (key,value) pairs
  • key values identify parts of computation
Map(key1,val1) → list(key2,val2)
  • applied in parallel to all (key1,val1) pairs
  • results with common key2 are collected in group for "reduction"
Reduce(key2,list(val2)) → val3
  • collects all values tagged with key2
  • combines them to produce result(s) val3