Bandwidth throttling consists in the limitation of the communication speed (bytes or kilobytes per second), of the ingoing (received) or outgoing (sent) data in a network node or in a network device such as computers and mobile phones.

The data speed and rendering may be limited depending on various parameters and conditions.

Bandwidth throttling should be done along with rate limiting pattern to minimize the number of throttling errors.

Limiting the speed of data sent by a data originator (a client computer or a server computer) is much more efficient than limiting the speed in an intermediate network device between client and server because while in the first case usually no network packets are lost, in the second case network packets can be lost / discarded whenever ingoing data speed overcomes the bandwidth limit or the capacity of device and data packets cannot be temporarily stored in a buffer queue (because it is full or it does not exist); the usage of such a buffer queue is to absorb the peaks of incoming data for very short time lapse.

In the second case discarded data packets can be resent by transmitter and received again.

When a low-level network device discards incoming data packets usually can also notify that fact to data transmitter in order to slow down the transmission speed (see also network congestion).

NOTE: Bandwidth throttling should not be confused with rate limiting which operates on client requests at application server level and/or at network management level (i.e. by inspecting protocol data packets). Rate limiting can also help in keeping peaks of data speed under control.

These bandwidth limitations can be implemented: