MapReduce Explained Simply

Core idea

MapReduce separates user logic from distributed execution. You define map/reduce, the runtime handles partitioning, scheduling, shuffle, and retries.

Map phase

Map reads input records and emits key/value pairs in parallel.

tsx
Copy
1function map(docId, text) {2  for (const word of tokenize(text)) emit(word, 1);3}

Shuffle

All values for the same key are grouped and moved to reducers. This is often the most expensive phase.

Reduce phase

Reducers aggregate grouped values.

tsx
Copy
1function reduce(word, counts) {2  emit(word, sum(counts));3}