Tessera

Note Figure source with adaptation.

• HDFS ＋ MapReduce

  • HDFS as distributed data storage

    • Read data from HDFS to multiple workers on different nodes (keep data close)

  • Hadoop YARN/MRv2 as the computing backend

    • Map: Process data on each worker (parallel)
    • Reduce: Shuffle data (map output) and do additional processing on each worker

• R as the front end

• Rhipe as the connector between R and Hadoop