Abstract: Graph summarization aims to search a group of simple hypergraphs or sparse graphs, which illustrate the main structural information or change trend of the original graph. Based on the application field and background of original graph, different graph summarization techniques are used to construct a specific summary graph, which can solve the problems of information overload, query optimization, spatial compression, impact analysis, social network visualization and so on. According to the classification criteria of the main purpose of the summary, the existing graph summarization techniques are divided into four categories: the graph summarization based on spatial compression, the graph summarization based on query optimization, the graph summarization based on pattern visualization and the graph summarization based on impact analysis. The partial graph summarization algorithms of non-attribute graphs and attribute graphs are tested on real data sets to analyze the indexes of information retention rate, compression rate, information entropy and running time experimentally. At last, not only the development trends of the graph summarization are highlighted, but also the challenges and the future research directions that can be explored in depth are pointed out. Combining with the popular deep learning technology, some valuable and potential Macro coutermeasures are put forward to solve these challenges.