• If all your data fits on one machine, you can have multiple replicas of the same data
    • This is still not necessarily straightforward
  • If your data does not fit in one machine, you must partition it according to some scheme
    • This is even less straightforward
    • It basically requires everything you need for replication, plus much more