Social Network Analysis (SNA) is aimed at studying the structure of a social network, usually represented as a graph, in order to extract the hidden knowledge about the activities and relationships of the users. With exponential increase in the volume and velocity of the data created in today's social networks like Facebook and Twitter, a main requirement for social network analysis is employing computationally efficient algorithms and methods. Since sequential and centralized approaches are far from the desired scalability, a natural solution is to distribute graph of the network on a number of processing machines and perform the execution in parallel. In this paper, existing Works on distributed large-scale graph processing are reviewed in four categories regarding their computational model. It is concluded that none of the existing categories outperforms other ones significantly, and therefore no single category addresses the requirements of all different graph algorithms. This highlights the need to research on identifying the types of algorithms for which each category of the computational models is more suitable, and also on how to customize the model for the corresponding type.