IP multicast is not widely deployed over Internet. One of the reasons which prevents its deployment is multicast forwarding state scalability and control explosion. In this paper, we propose an algorithm called Q-STA (QoS Scalable Tree Aggregation) which reduces the number of forwarding states by allowing several groups to share the same tree (the less trees, the less forwarding states). Q-STA performs fast aggregations by evaluating few trees for each aggregation while increasing the number of accepted groups. Q-STA builds trees by using links having a low utilization in order to achieve load balancing. Moreover, groups are aggregated to trees having a minimum number of links in order to minimize the network load. By extensive simulations, we show that Q-STA outperforms the previous algorithm.