As the IPTV user number continues to grow, the IPTV Application Servers increases and eventrally stats to overload. In this case, the cluster system has become an effective application server that provides high performance and a high quality of services. It keeps the load in balance, and it can enhance the system throughput effectively. It can also improve the IPTV quality of service. For the load balancing problem, a two-phase adaptive load-balancing strategy, which is based on the characteristics of IPTV users’ behaviors, is proposed. This load balancer will dynamically modify the load dispatch weight of each node in the application server cluster system by a real time watch of each node’s loading degree. This algorithm keeps a load-balanced situation among all the application servers. It maintains the integrality of users’ session, and it also ensures the QoS (Quality of Services). The experimental results of this paper show that this algorithm has a good performance, and can improve the system’s throughput on the basis of guaranteed QoS.