Abstract Emergency communication networks that serve various public safety personnel, including medical responders, police, and hazard and fire fighters, play a critical role in routing call services and managing voice traffic. In such networks, these incoming voice calls from the affected population contribute to the maximum volume of traffic, which may not be supported by existing infrastructure because of logistical constraints. Using the traces of the Mongolia Emergency Information Network (EIN), we propose a new method of flexibly allocating the service time for voice calls, ensuring significantly improved performance. Our first contribution involves a comprehensive analysis of the experimental traces obtained from EIN, including a study of the impact of the call holding times on the system throughput, comparing quiet periods with high traffic due to emergency events. Observations from our study include: Traffic congestion occurs in the evening, especially for medical responders and police traffic, as opposed to quieter morning hours, and the frequency of emergency calls for hazard and fire fighters exhibits a flat response all day. The second contribution lies characterizing the system capacity and formulating a service policy for peak demand situations, where operators get overloaded, causing a system breakdown. By carefully integrating the bursty nature of the calls, our dynamic service time allocation method based on the MVA (Mean Value Analysis) concept shows a visible improvement in the robustness of EIN, allowing much higher throughput for reduced call holding times.