This study proposes a practical directed-graph drawing method that can be applied to large-scale relational data by modifying multidimensional scaling (MDS). We modified the mutual distance between the nodes to incorporate the link direction information in the distance matrix, from which the coordinate of the vertex is obtained. As the application of our method, we visualized the big data of real economy of Japan which includes a million companies and millions of relations, and we particularly deal with how firms are connected. Moreover, we discovered interesting features of the economic network. Firms are connected by links into tightly knit groups with high intragroup density and low intergroup connectivity community structures. We also found that the features of community structures are specific to individual industrial sectors, such as manufacturing, retail, wholesale, and construction.