Call admission has become important to guarantee the QoS, since there are limitations of capacities of resources as links on network. We direct our attention to the online call admission algorithm that aims to maximize the total of profits of accepted calls, subject to capacity constraints of links on the network. The call admission problem has been often studied by focusing on the network resource. However, like GRID computing, when the traffic changes according to the throughput rather than the distance between computers used for communication, it is important to pay attention to the change of the traffic. Therefore, we experimentally analyze the relationship between the change of traffic patterns and the behavior of a call admission algorithm. Consequently, experimental results show a characteristic of the algorithm such that there are two types of variations in the performance for the change of traffic demands; sensitive and stable.