Conventionally software has been written for serial computation in that programs are written for computers which have a single Central Processing Unit (CPU). Here the problems are solved by a series of instructions which executes one after the other and one at a time by CPU. But many complex interrelated events occurring at the same time such as galactic and planetary orbital events, tectonic plate drift and weather and ocean patterns might need super high complexity serial software. To handle these large problems and save computational time a new programming concept known as parallel programming was introduced.

To create a parallel program we should at first decide whether problem has some part that can be parallelised. There are a number of problems such as generating Fibonacci sequence in case of that there is a little scale for parallelization. Once it has been decided that problem has some section which can be parallelized we break problem in discrete portions of work which can be distributed to many tasks. This part of problem can be function-centric or data-centric. In the previous case various functions work with various subsets of data whereas in latter every function carries out a part of overall work. Depending on the kind of partition approach we need communication amid the processes. Accordingly we have to design the mode of communication and mechanisms for process synchronization.