The data explosion in large-scale distributed simulations cripples the performance of simulation and restricts the scalability and persistence. Because HLA (high level architecture) has inherent limitations to define the level of demands for data, RTI (runtime infrastructure) cannot utilize the simulation features effectively to make congestion control. The LOI (layer of interest) depicts the differences of demands for data in various receivers and provides a way to make some encouraging technologies feasible, such as QoS (quality of service), congestion control and layered DDM (data distribution management) strategy. In this paper, the LOI is combined into RTI抯 working mechanisms and a phase filtering based on the LOI is put forward to control congestions in RTI. Finally, experimental results are presented to demonstrate the efficient control over congestions in RTI using the LOI-based phase filtering, together with the critical data and transportation stability guarantee.