###GraphAnalysis And Parallelization #Introduction to Problem The basic problem is to calculate the total number of connected components using parallel graph algorithm. The dataset used in this project is Facebook dataset which contains 88000 edges and 4000 vertices. Secondly we are using parallel approach instead of sequential one because we need to reduce our time as dataset can be large containing billions or millions of node. Moreover in real time systems data processing needs to be fast, so working on a parallel approach is an appropriate problem to consider. The libraries that incorporate graph analysis are available for example Graph-Stream (Java), JUNG (Java universal Networks and Graphs) for java developers and BGL (Boost Graph Library), Lemon and SNAP for C++ developers are available in the market. But all these libraries don’t have parallelization facility available with them. There is a library called PBGL (Parallel Boost Graph Library) which incorporates parallel features demands distributed graph as an input, but that totally deviates to a different approach. So after exploring most of these libraries came to a conclusion of writing our own code from scratch in Java and exploit features of multithreading available with Java. But for Visualization we picked Graph-Stream as an extra tool. There is a difference between parallel and distributed graph Algorithm. We have picked parallel approach keeping in mind the scope of the project. Parallel graph algorithms are the one in which computation is distributed among systems whereas in distributed one computation can be distributed among geographically isolated systems. While pre-processing Facebook dataset which is given as an edge-list, we are pre-processing to convert first into adjacency matrix and later on dividing into chunks. The subsequent section discusses upon algorithms and data structures used for this project. Following images show results of different threads working on individual forests. Number of threads is equal to the number of forests.