Message passing is possible the most widely used parallel programming paradigm these days. It is the most portable, natural and efficient scheme for distributed memory systems. It offers natural synchronisation among the processes so that explicit synchronisation of memory access is redundant. The programmer is responsible for concluding all parallelism. In this programming model many processes across the arbitrary number of machines every one with its own local memory exchange data through send and receive communication between processes. This model can be best understood by the figure displayed in Figure:

Figure: Message passage model

As the figure suggests every processor has its own local memory. Processors execute computations with the data in their own memories and cooperate with the other processors as and when required by communication network employing message passing libraries. The message comprises the data being sent. But data isn't the just constituent of the message. The other elements in the message are: 

• The address/ identity of the processor that sending the message
• Starting address of data on the sending processor
• The type of data being sent
• The size of data
• The identity/address of processor are receiving message
• Starting address of storage for data on receiving processor.