System capability is one of the core capabilities of the students majored in computer science, and the operating system courses are the key to cultivating it. There are some typical problems in the traditional teaching process. Firstly, it is difficult to distinguish the ability gap among students in a fine grained manner, so we cannot provide the students detailed guidance. Secondly, teaching activity design is too general and coarsegrained, and there is lack of iterative feedback mechanism during and after each semester. Thirdly, without finegrained learning behavior data, the rules of competence acquisition cannot be analyzed and discovered at the micro level. We design an integrated experiment environment for operating systems, which can not only realize automatic experiment releasing, code committing and evaluation, but also collect and analyze the behavior data of students during the experimentation process. Thus, we can discover the problems in the teaching process in time. Only 38% of students completed at least 4 labs before the experiment environment came into use in 2014, which reached 62% in 2015. However, by analyzing the learning behavior data in 2015, we found problems such as code copying, lack of preliminary knowledge, inadequate code reading and the lab2 phenomenon. In response to these problems, we take measures both in the current semester and in the longterm period. Finally, quantitative analysis of learning behavior data for 2015, 2016 and 2017 shows that, these improvement measures achieved significant effect.
