Skip to Main Content
Data-parallel programming model (DPPM for short) specialized for data-intensive computing becomes considerable popular because it simplifies the development of distributed parallel programs. DPPMs are classified into two categories: 1) MapReduce, Dryad; and 2) Piccolo, Function Flow, etc. based on their maturity. We analyze and compare these typical models by deployment, application, data partition, communication, fault tolerance and so on. Finally, we pay more attention to discussing development of key technologies which are deployment of storage and computation, task partition and fault tolerance in DPPM.