Other meanings of Network congestion
Computer Networking
Network congestion occurs when a network link or node carries more data than it can handle, leading to reduced quality of service such as increased latency, packet loss, and degraded throughput. It is a fundamental problem in packet-switched networks, including the Internet, and is managed through congestion control algorithms and traffic engineering.
Network congestion arises when the aggregate demand for bandwidth exceeds the capacity of a link or switch, often due to bursty traffic, inadequate provisioning, or routing inefficiencies. In TCP/IP networks, congestion manifests as queue buildup in router buffers, leading to increased queuing delay and eventually packet drops when buffers overflow. The primary mechanism is the mismatch between the sending rate of sources and the available capacity, exacerbated by the lack of global coordination among senders. Congestion can also be triggered by network failures that reroute traffic onto already saturated paths, or by malicious activities such as denial-of-service attacks that flood links with unwanted traffic. Unlike flow control, which manages the sender-receiver window, congestion control is a network-wide issue requiring cooperative algorithms.
The most influential congestion control algorithm is TCP's additive increase/multiplicative decrease (AIMD), introduced by Van Jacobson in 1988, which dynamically adjusts the sending window in response to packet loss or delay signals. Modern variants include TCP Reno, NewReno, and Cubic (the default in Linux), which use loss-based detection, while delay-based algorithms like Vegas and BBR (Bottleneck Bandwidth and Round-trip propagation time) aim to operate at the point where the pipe is full without inducing queue buildup. Active Queue Management (AQM) techniques, such as Random Early Detection (RED) and CoDel, help by signaling congestion before buffers overflow, improving fairness and reducing latency. In addition, explicit congestion notification (ECN) allows routers to mark packets instead of dropping them, enabling faster and less disruptive feedback. These algorithms are critical for maintaining Internet stability, as uncontrolled congestion can lead to congestion collapse, where throughput plummets to near zero.
Congestion degrades user experience by increasing latency (e.g., in video conferencing and online gaming) and reducing effective throughput for bulk transfers. In severe cases, it can cause timeouts and retransmissions that further worsen the situation, leading to a positive feedback loop. Network operators mitigate congestion through capacity planning, traffic shaping, and load balancing, as well as by deploying content delivery networks (CDNs) to bring data closer to users. Quality of Service (QoS) mechanisms prioritize delay-sensitive traffic, while multipath protocols like MPTCP can utilize multiple paths to bypass congested links. On the Internet, inter-domain congestion is managed through peering agreements and transit pricing, though disputes can arise, as seen in the 2010 Comcast–Level 3 conflict. Research continues on new architectures, such as information-centric networking, to address congestion more fundamentally.
Beyond the classic TCP/IP context, congestion appears in other domains: in road traffic, it is studied using the fundamental diagram of traffic flow, and in data centers, it is exacerbated by the incast pattern where many servers send to one switch simultaneously. The term 'congestion collapse' was first observed in the early Internet, notably during the 1986 'congestion collapse' events that motivated Jacobson's work. A lesser-known fact is that the original TCP specification did not include congestion control; it was added later. Also, the concept of 'bufferbloat'—excessive buffering in routers—was popularized by Jim Gettys in 2010, showing that large buffers can paradoxically increase latency. In wireless networks, congestion is often confused with interference, but they are distinct: congestion is due to overload, while interference is due to signal degradation. Finally, congestion pricing, a concept from economics, has been applied to Internet traffic management, though rarely implemented.
This article focuses on the networking sense of network congestion, not road traffic or other uses.
Help improve the encyclopedia. Reports go straight to the site manager.