Network traffic management based on differentiated and integrated services

Abstract

The problem of traffic control and management for the purpose of forecasting and prevention of congestion of a telecommunication network is considered. The definition of the concept of traffic policy is given. The solution of traffic control problems is based on the application of a policy that allows the harmonious distribution of information flow. A promising way to prevent congestion is the management of traffic parameters ‒ policing and shaping, i.e the management of statistical characteristics of traffic. The necessity of using probabilistic and statistical methods of traffic control, based on correlation and regression analysis, is substantiated. The concept of end-to-end network diagnostics is considered. End-to-end management includes network management as a whole, congestion management and traffic parameters management. Further development of traffic policy is the use of differentiated and integrated services. The method involves the dynamic allocation of bandwidth of transmission channels between users based on their priorities. This approach avoids periodic local congestion of routes that have been identified as optimal at the current stage. The algorithm involves the selection of routes according to the criterion of equable loading of transmission lines. To implement a policy of traffic with differentiated services, it is necessary to solve the problem of flow distribution between locally optimal or quasi-optimal routes in real time. The process model of the total flow formation due to the imposition of separate flows from different sources is considered. It is shown that, under the conditions of maintaining the balance of network flows the intensities of flows to nodes with limited buffer memory exceed the corresponding intensity of the network with unlimited memory. This saves storage and reduces packet delays due to latency in switching buffers.