throttling

Throttling is the deliberate slowing or limiting of resource usage by an entity to control consumption, protect resources, or enforce policies. It can occur across computing, networking, software services, and hardware.

In networks, throttling refers to limiting bandwidth or data transfer for users or applications. Techniques include traffic shaping, rate limiting, and policing. Shaping smooths traffic over time; policing enforces a fixed cap by dropping or marking excess packets. Throttling can be applied by Internet service providers, enterprises, or network devices, and can raise concerns about net neutrality and equal access.

In software and APIs, throttling restricts the rate of requests or operations. Common approaches use token bucket or leaky bucket algorithms to grant or deny actions based on tokens in a window. Clients may receive error responses such as an HTTP 429 Too Many Requests and should implement backoff strategies to recover gracefully.

In hardware, throttling refers to reducing a component’s performance, such as CPU throttling, where clock speed is lowered to limit power consumption or heat. This is often part of energy management and thermal protection, and may be triggered automatically by hardware sensors or fixed policies.

Policy implications and effects: throttling can improve stability and fairness by preventing resource exhaustion, but may degrade user experience if overly aggressive. It is often governed by service level agreements, terms of service, and regulatory considerations in areas like net neutrality and consumer protection.

Related concepts include quality of service, traffic shaping, rate limiting, admission control, and congestion control.