Graph aggregation and summarization is to obtain a concise supergraph covering the most information of the underlying input graph, and it is used to extract summarization, solve storage consumption and protect privacy in social networks. This paper investigates current graph aggregation and summarization techniques and further reviews and classifies their partitioning/grouping methods. Based on the consistency of grouping information, five grouping criteria are specified: The consistency of attribute information, the consistency of neighborhood group, the consistency of connection strength, the consistency of neighborhood vertex and reconstruction zero error. From the top level view, graph aggregation and summarization techniques can be classified into three types, namely, attribute similarity, structure cohesiveness and the hybrid of both. This paper comprehensively summarizes the state of art of current research works, and explores the research directions in the future.