ABSTRACT
The notable approaches to address the congestion control and resource allocation issues from a macroscopical perspective are the primal algorithm, the dual algorithm, and the primal-dual algorithm. All these algorithms result from the network utility maximization model. The significance of these algorithms is the modeling of the congestion measuring and source actions. It is presented in the following: Links can feed back to data sources about the information of congestion in resources being used. It is assumed that each network link measures its congestion by a scalar variable and that sources have information of the aggregate price of links in their path; and With the congestion information, flows can take actions by using their control mechanisms to react to the congestion along the path to adjust their sending rate. These assumptions are implicitly presented in many variants of the transmission control protocols (TCPs). The price signal being used in these protocols can be, for example, loss probability and queuing delay.
