Graph sampling is a technique to pick a subset of vertices and/ or edges from original graph. Among various graph sampling approaches, Traversal Based Sampling (TBS) are widely used due to low cost and feasibility for many cases, in which Simple Random Walk (SRW) and its variants share a large proportion in TBS. We illustrate the foundation SRW and presents the problems of SRW. Based on the problems, we provide a taxonomy of different Random Walk (RW) based graph sampling methods and give an insight to the reason why and how they revise SRW. our summary includes classical methods and state-of-art RW-based methods. There are 3 ways to propose new algorithms based on SRW, including SRW and its combinations, modified selection mechanisms, and the graph topology modification. We explained the ideas behind those algorithms, and present detailed pseudo codes. In addition, we add the mathematics behind random walk, and the essence of random walk variants, which is not mentioned in detail in many research papers and literature reviews. Apart from RW-based methods, SRW also has related with the non-RW and non-TBS methods, we discuss the relationships between SRW and non-RW methods, and the relationships between SRW and non-TBS methods. The relations between these approaches are formally argued and a general framework to bridge theoretical analysis and practical implementation is provided.