MapReduce
- systems
Jeffrey Dean and Sanjay Ghemawat’s 2004 Google paper described MapReduce as a programming model and implementation for processing large datasets on clusters. A map stage produces intermediate key-value pairs, and a reduce stage combines values associated with the same key. The system handles substantial distribution machinery behind this model, including task scheduling and failures. Its influence came from making certain large data-processing jobs easier to express and operate across many machines. Not every computation fits map-and-reduce structure efficiently; the paper’s success concerns a useful class of workloads rather than a universal replacement for all parallel programming. **Timeline anchor:** 1 December 2004 CE. Approximate day. **References:** [Source 1](https://en.wikipedia.org/wiki/MapReduce).