Today``s Internet provides only best-effort services. However, future Internet is required to support various quality of services (QoSs). Many technologies such as Differentiated Services (Diff-Serv) and Multi-protocol Label Switching (MPLS) have been proposed to meet various QoS requirements. Different classes of traffic can be controled using different Label Switched Paths (LSPs), and therefore various types of traffic engineering is possible.
In order to estimate QoS``s for the allocated resources, it is crucial to adopt an appropriate traffic model which represents the patterns of real Internet traffic. It is important to investigate the relation between the required QoS and allocated resources. Many authors have demonstrated the presence of long-range dependence in the traffic of Wide Area Network (WAN), Local Area Network (LAN), World Wide Web (WWW). It has been shown that these traffic patterns are well modeled using self-similar processes. Several specific traffic models have been proposed to characterize these traffic patterns, such as fractional Gaussian noise (fGn), fractional ARIMA and multi-fractal wavelet model. However, it is difficult to predict the performance when input traffic shows self-similar characteristics and multiple classes of traffic are served in various scheduling environments. We propose fluid simulation for reducing simulation times instead of packet-level simulation with First-In First-Out (FIFO) discipline and Generalized Processor Sharing (GPS) scheduling. Many scheduling policies are based on GPS to fairly determine the order of packet transmission. A number of service disciplines have been introduced to reduce computational complexity and to guarantee fairness of different queues for packet-level fair queueing. Fluid simulation algorithms in scheduling systems are proposed and the complexity of the algorithm is analyzed. Compared to packet-level simulation, the proposed fluid simulation is simpler and more tractable because ...