As cloud computing becomes more and more popular, Virtual Machine Introspection (VMI) technologies and tools are evolving to better monitor the behavior of virtual machines residing in the IaaS (Infrastructure-as-a-Service) cloud. With those technologies, a lot of applications came out. But all these applications that utilize VMI need to overcome the problem of disconnection between the low-level state that the hypervisor sees and its semantics within the guest, which is called the semantic gap. In this paper, the authors mainly introduced VMI's type, VMI's key problem, and analyzed six methods aiming to solve the semantic gap. They also proposed a combined method to deal with the semantic gap better.