Networks are ubiquitous in our world. Prominent examples are the WWW and social networks. Many of the networks are very big and complex consisting of over millions of nodes and links. Therefore, pattern mining from big networks is a daunting task. In this talk we focus on mining two kinds of structural patterns including community structures and functional roles of nodes. More specifically, community structures are densely connected groups of nodes, with only sparser connections between groups. An example of community structures in social networks is a group of like-minded people. Many algorithms find community structures. But they tend to fail to identify and isolate two kinds of nodes that play special roles – nodes that bridge communities (hubs) and nodes that are marginally connected to communities (outliers). Recently, we proposed a novel algorithm called SCAN (Structural Clustering Algorithm for Networks), which detects community structures, hubs and outliers in networks. The algorithm is fast, visiting each node only once. An empirical evaluation of the method using both synthetic and real datasets demonstrates superior performance over other methods such as the modularity-based algorithms. Last but not least, we present a MapReduce/Hadoop implementation of SCAN for big social networks like Twitter.