Traffic prediction based wavelength resource management considering holding time

Optical WDM networks are expected for future multimedia services. The multimedia services require stable connections with various length of holding time. While some multimedia services use optical path connections in only minute, other multimedia services use them in several hours. As network perspective, the overhead of connection setup has much influence in the case of short-term connections. On the other hand, longterm connections occupy assigned network resources for long periods. Thus, it is important to assign the appropriate wavelength considering the holding time. Furthermore, the network traffic is large and rapidly changes in time. Therefore, the method that accurately estimates the state of the current traffic and predicts the state of future traffic becomes more important. In this paper, we propose a novel traffic prediction method that can deal with large changes in traffic in optical WDM networks. The proposed prediction method is classified into two categories; “short-term traffic state estimation” and “longterm traffic trend prediction”. Traffic is predicted by using these two methods cooperatively and complementarily. We also propose a wavelength assignment method exploiting this traffic prediction. The proposed wavelength assignment method sets optical paths considering the prediction information and the holding time of connections. Each connection with different holding time is allocated wavelength resources in the limited range based on the traffic prediction information dynamically. We evaluate the performance of the proposed method by computer simulations and confirm that the proposed method can achieve better performance than the traditional method does.