Currently existing models of network traffic have been developed using network traffic traces gathered via traditional network monitoring systems. These models are then used to develop new network protocols and networking enhancements.
However, the kind of results typified by Figure 8 indicate that network traffic demands of applications are not accurately reflected by traffic on the network wire (since the traffic pattern has already been modified by the current network protocol). Hence, while current network models may accurately reflect current network-wire traffic, they are not useful in optimizing application communications.
We have seen (in Section 4.3) that recursively passing traffic through TCP stacks results in decreased network performance. Yet the current tools available to network researchers (such as tcpdump) only allow traffic collection after that traffic has passed through at least one TCP stack. Thus, the results of simulations using this traffic data are corrupt; the input data has already been modulated. This is a critical observation for the development of next-generation, high-speed network protocols.
Traffic traces generated by MAGNeT provide a realistic picture of the protocol-independent traffic demands generated by applications running on today's networks. Thus, MAGNeT provides network researchers and developers with a better understanding of applications' actual communication needs.