This thesis discusses the centralized and decentralized congestion control strategies in differentiated services of a networked system. First a sensor-decision maker-actuator (3 nodes) network structure is designed and traffic flow is divided into three classes: Premium, Ordinary and Best effort. For each node we design a control strategy based on the dynamic fluid flow model and M/M/1 queuing theory to allocate bandwidth to different class traffic for avoiding congestion. For the whole network, we design control algorithms based on both centralized and decentralized methods. The stability of the closed loop system is analyzed and some comparisons between centralized and decentralized schemes are made. Several examples and simulation results are included to illustrate the design methodology.