Internet congestion optimization and control: Transmission protocols for wire-line and wireless networks

This study addresses the problem of congestion control and bandwidth allocation in both wired and wireless networks. We use Explicit Congestion Notification as a criterion for measuring the congestion in data packet networks. The first model suites wired networks. In this model a new economic approach is presented for modeling the Internet links and users. The approach implements both ECN and packet losses as a congestion measure. The objective function was shown to be concave over a convex set of constraints. A constraint relaxation procedure was held to solve the problem in a decentralized manner. The relaxation involves an exponential probability penalty function that depends on the constraints deviation. The routers will mark the packets based on the probability distribution while the end users will monitor the packet marks and losses. A decentralized transmission rate control law is implemented at the data sources to find the new rate allocations.; Another model that depends on the queuing delay and a target number of packets is presented. The objective function will maximize the aggregate users' utilities given a set of delay and upper bound constraints. The model was solved using the optimization duality approach. It presents a new class of utility curves that are strictly concave with a new closed set of delay constraints. The solution demonstrates excellent convergence characteristics. Further, the model involves minimum roundtrip estimation using the ECN marks. The presented results show that implementing the ECN marks for better roundtrip computations is very promising and can significantly improve the estimates.; The last part includes congestion and flow control in wireless environments. We proposed the idea of new agents to monitor the channel status and model the packet bursts along with the wireless channel fading at the receivers. A FSMC model is used to find the wireless channel status. The model presents further, a discrimination scheme between the congestion and wireless lost packets. The estimated wireless loss fraction is conveyed to the end users via marking. A new control law that does not penalize the data sources for the random wireless losses is used at the application transport layer.