With the general popularity and ubiquitous availability of Short Message Service (SMS), numerous service providers use SMS as underlying data distributing carrier, such as the short-message based authentication system and emergency notification. For this kind of application services, the reliability and delay of SMS are critical factors to the user satisfaction of the service. This paper presents statistic analysis results of the trace data in a SMAP (Short Message-based Authentication Platform) system. The analysis results show that the delay is caused by the short-message distributing latency (from SM-SC to user's mobile phone) in most cases. And a decision tree-based solution is introduced to diagnose the failure in SMAP system automatically. Preliminary results illustrate the effectiveness of the decision tree-based approach.