On applying the method of “system of systems” in robustness analysis and autonomous control of dynamics-aware internet architecture

Owing to the advantages of the “system of systems” that has been widely used in enormous fields ranging from science to engineering, in this paper the method of the “system of systems” is applied to the analysis of the network architecture of the Internet and the corresponding controller of the Internet flow where the robustness is one of the most crucial factors for the performance analysis of the autonomous architecture of the Internet. In our research, the underlying network architecture of the internet is formalized by a model of a generalized networked control system (NCS) where the autonomous distributed information processing units are constrained by the end-to-end rule of the Internet. With the formalization, the factors of nonlinear controllers, communication channels, and communication protocols are interacted in the model based on the principles of the “system of systems”. Accordingly, the technology of “software systems engineering” is applied to design and test the new algorithm proposed by us for the Internet flow control, which is based on the nonlinear filtering mechanism of the TCP congestion control, to provide a new Internet flow control protocol. The higher stability of the nonlinear controller for the Internet flows is observed in the experiments compared with the linear ones. The robustness of the Internet flow control is achieved even under the unexpected dynamical environment that causes the uncertainty of bandwidth, delay, and loss of the communication. In order to optimize the design of the network architecture, the corresponding analysis of the complex behavior of the flow dynamics of the Internet has been carried out. The significance of the results shows that a new design principle of the dynamics-aware protocols for the future Internet can be innovated in terms of the designed mathematical model of the “system of systems”. From our practice, it can be infer- ed that the performance evaluation of the Internet flow control under the cross layers can be extended to reconfigurable overlay networks.