Price-based Congestion Control Algorithms

In recent years, with the developing of the computer and communication technologies, large strides have been taken in analytical approaches to network congestion control. In this paper, we reviewed the "dual"congestion control algorithm based on nonlinear optimization theory, view network links and sources as processors of a distributed computation system to solve the dual problem using gradient projection algorithm. We study the solution algorithm of 'shadow prices'and a shadow price-based flow control(PFC) framework, leading to a new Congestion Control algorithm , eXplicit Price-based Congestion Control algorithm(XPCP). Based on the work of Kelly et al. And Low and Lapsely, a price-based flow control(PFC) framework is defined and explained. This framework can be used to analyze and develop distributed flow control algorithms for packet-switched data networks under the premise that as much control as possible should be exercised by the end nodes. Important resource allocation goals are described and compared, notably social welfare maximization (SWM) and weighted max-min fairness (WMMF) both of which consider individual demand functions of different users and are therefore suitable for the heterogeneous requirements found in modern telecommunication networks. A detailed analysis of a shadow price-based flow control (PFC) framework is performed. Using the price as the primary control variable, many existing characterizations of the widely researched max/min fairness concept are generalized to the case of weighted max-min fairness. The necessary condition of the WMMF optimization problem gives a theoretical foundation to the widely used characteristics that (weighted) max-min fairness is achieved in a network where sources take feedback only from the most congested link along their path. A practical application of the XPCP framework is the idea of charging explicit congestion prices as determined by a price-based flow control algorithm. This idea is aligned with existing research on congestion pricing and compared to other congestion pricing schemes. Important problems are identified and possible solutions are shown. In this paper, the price-based flow control framework are supplemented and developed, and mainly study the Price-based Congestion Control Algorithms. The mainly works are: First, consummated and expanded the price-based control framework. This is the fundamental method frame of the price-based congestion control arithmetic, so when research the price-based congestion control arithmetic, we must based on this frame. This frame can explain and analysis the most congestion control arithmetic, through the control theory and nonlinear optimization theory. But by now, these researches are dispersed and principium state. Hasn't become an integrated system. The second part of this paper concentrates on the studied of this frame. First, we introduce the primary framework of the price-based optimization flow control. Then we showed the arithmetic of DFP Method(Davidon-Fletcher-Powell method)of shadow price in link, added the capability of arithmetic, parameter discuss, network sign structure (SumNet vs. MaxNet)and analysis the end system et. And improved the REM arithmetic,offered a new arithmetic base on explicit price feedback, formed an integrated framework of price-based congestion control arithmetic. In this process we gained integrated price-based congestion control arithmetic. Second, we show you a new way of figure out the shadow price in link, according to which we improve REM arithmetic based on implicit price feedback. The shadow price of link has always be got by gradient projection algorithm. However convergence of gradient projection algorithm too slowly, sometimes it doesn't convergence at all. That is why it doesn't full fill the requirement of quick responsibility by high bandwidth network. DFP Method is presented in our paper, which is characteristic of quick convergence .REM arithmetic improved according to this simulation analysis was proved to have more stability and convergence quickly. Third, we advanced a XPCP arithmetic based on dominance price feedback. Most of studies are concentrates on implicit price feedback arithmetic at present. Sources need to estimate the price of link again. Obviously, that brings out some difficulties and no-accuracy to count. And nobody involved in the arithmetic basis on explicit price feedback. At the same time, studied on optimization flow control always used the SumNet structure. It doesn't involve in the max/min fairness concept MaxNet structure. XPCP is the MaxNet structure and base on explicit price feedback. simulation study indicated that the new arithmetic has the quick respond and high using rate of bandwidth.