With the increase of internet services, internet has been closely connected with people's daily life. Experience in human-computer interaction has a variety of effects on people, such as emotions and behaviors. One of the major factors that affect user experience is the server delay. This paper adopts a passive measurement method in large-scale network, so as to collect the information of users and servers, including a large amount of time information. Then it analyzes the distribution characteristics of the server delay through these data, in order to find the factors that affect the delay. Furthermore, we also analyze the stability of server, through studying the correlation between server delay and network load. This study is conducted based on the big data collected from network traffic monitoring system, covering tens of thousands of the main Internet Content Providers (ICP). The results show that the server delay is affected by many aspects. This paper hopes that suggestions can help ICP find a solution to shorten the server delay.