With the development of cloud computing and big data technology, privacy protection draws more people's attention. Data encryption is a common way to protect data privacy, but solely using encryption cannot resist all types of attacks. Adversary can observe the access pattern on how users access to the data, to infer the private information including the importance of the data, the relevance between the data and even the plaintext of encrypted data. Oblivious RAM (ORAM) is an important method to protect the access pattern, including access operations and access locations, by obscuring an actual access, which makes adversary unable to distinguish it from a random one. ORAM makes an important role in designing secure cloud storage systems and secure computation. ORAM can reduce the possibility of the adversary inferring the private information through the access pattern and reduce the attack surface of the system, so as to provide a safer and more complete service. This paper summarizes the researches and application settings of the ORAM, mainly introducing the relevant concepts of the model as well as design methods with focus placed on analyzing and summarizing common strategies to optimize the model and their advantages and disadvantages, as well as optimizations for amortized and worst-case bandwidth between client and server, storage overheads reduction and round-trips reduction. Moreover, this paper discusses general issues of the ORAM used for secure storage system designing including data integrity and concurrent accesses for multi-clients. The paper also discusses some issues of the ORAM used for secure computation, including secure computation protocols designing and oblivious data structure designing, and finally makes a conclusion for the future research directions of the ORAM.