Congestion Control

Congestion refers to demand exceeding service capability. It is characterized by its duration and extent. This chapter discusses two major approaches to deal with congestion: proactive congestion control and reactive congestion control. It introduces the architecture of congestion control in the Internet, which is implemented in the Transmission Control Protocol (TCP). The chapter outlines several versions of the TCP congestion control. It describes a fluid model of TCP congestion control and addresses generalization to network-level modeling. The chapter establishes simple relationships between the steady-state throughput, congestion window size and packet loss of a long-lived, greedy TCP connection. It presents a classic analysis of a TCP connection with a single bottleneck link, using a fluid approximation. The chapter also includes a discussion on network utility maximization and provides information on the challenges to TCP.