It is of great importance to study real evolution features of power system cascading failures to well understand the continuing spread of events and related system behaviors. Based on the academic review of previous paper, eight blackouts caused by cascading failures with great impact over the world are chosen as examples to carefully analyze the evolution procedure, critical event, controlling spread mode and other common characteristics of cascading failure. According to above studies, some drawbacks of the existing techniques and algorithms such as low practical applicability and over-simplification of related factors or system behaviors are presented. Furthermore, considering the development of progressive information monitoring system, we discuss about power system cascading failures research directions from the academic and technical views of integrated vulnerability analysis algorithm, cascading faults on-line diagnose and intelligent alarm technique, device state monitoring and modeling and so on, which are expected to offer some general ideas for the future power system cascading failures research.