||This article may contain original research. (June 2011)|
A bandwidth cap, also known as a band cap, limits the transfer of a specified amount of data over a period of time. Internet service providers commonly apply a cap when a channel intended to be shared by many users becomes overloaded, or may be overloaded, by a few users. Implementation of a bandwidth cap is sometimes termed a Fair Access Policy, Fair Usage Policy or Usage-based billing.
Standard cap 
In many situations, each user of a network is expected to use high speed transmission for only a short time, for example to download a megabyte web page in less than a second. When use is continuous, as it might be in the case of file sharing, Internet radio or streaming video, a few users who use the connection at high rates for hours at a time may seriously impair the service of others. The concept is more relevant in cable internet where both the core network and the access network are shared, than in DSL where the core network is shared but the access network is not. It is most relevant in wireless internet, particularly satellite internet, where both the core network and the access network are shared and total network bandwidth is relatively narrow.
One type of bandwidth cap, administered by an Internet service provider (ISP), simply limits the bitrate or speed of data transfer on a broadband Internet connection. The purpose of bandwidth capping is to prevent individual users from consuming the entire transmission capacity of the cable, a shared resource. Critics have charged that it is a method to charge consumers more by introducing tiered bandwidth caps.
Capping might be handled by the user's cable modem. Knowledge of capping has led to attempts at uncapping. When uncapping succeeds, the resulting data transfer rate is supposed to be extremely fast, but users who are caught are said to be banned permanently by broadband ISPs. Uncapping is considered theft of service by ISP's.
Sophistication is possible, and even required in limiting bandwidth. The simplest approach simply limits the data rate. The problem with the simplest approach is, a very active user could consume the maximum bandwidth continuously, theoretically imposing an excess burden on the ISP and possibly reducing the performance of other users.
Channel capacity is a finite resource, using huge amounts of it can be deemed an abuse in countries with poor communications infrastructure. Dial-up ISPs often published policies that tried to clarify the difference between "unmetered" and "unlimited".
A more sophisticated approach is called "bursting", the administrator would specify a "peak rate limit", a lower "sustained rate limit", and a "credit limit". If you continuously saturate your connection, you will only get the sustained rate. While you are idle or use less than the sustained rate, you accumulate a credit, in bytes, up to some limit. If you try to saturate your connection after idling, you will get the peak rate until your credit runs out, at which point you will again be running at the sustained rate limit. The transition from peak rate to sustained rate could be abrupt, gradual, or even an arbitrarily designed curve. If you alternately idle and saturate, your long-term average data rate will never exceed the sustained rate limit, and your short-term data rate will never exceed the peak rate limit. This is known as a Token bucket.
Other schemes or models are possible to regulate bursting. Running at the peak rate could accumulate a debt. Once the debt reaches a limit, the user is held to the sustained rate limit and the debt does not change. Only running below the sustained limit (or idle time) pays back the debt. The behavior is similar.
Lowered cap 
Another type of capping refers to the reported phenomenon of an ISP reducing an individual user's bandwidth cap, without notifying that user, as a defensive measure and/or as a punishment for heavy use, especially for upstream traffic. "Servers" tend to use upstream bandwidth heavily, and violate most service agreements. (Cable and other broadband services tend to be asymmetrical, making upstream capacity scarcer than downstream.) Somehow the ISP detects that the user is an offender — perhaps by analyzing traffic to detect the activity of a server, or perhaps by comparing the user's long-term data rate against an unpublished limit. If a user gets tagged as an offender, then the ISP imposes a lower bandwidth cap upon that user, and/or restricts other services. e.g. NTT Communications in Japan imposes a 30GB/day upload cap with a warning for a first violation and disconnection for repeat offenders.
The implementation of a sustained rate limit might appear as a reduced cap. If the data rate improves automatically after idle time, this would confirm the operation of a sustained rate limit.
Reports claim that the intervention of an administrator is required to remove a lowered cap.
A number of broadband Internet Service Providers in North America and Europe introduced bandwidth caps in the early 2000s. The same practice has been in place in Australia, New Zealand, Malaysia (All others except ADSL broadband) and South Africa since the release of broadband. Data caps on internet in New Zealand result from many popular websites being hosted in the United States, hence local ISPs must lease international capacity; a proponent of a second international cable to the United States claims that cheaper international capacity will result in ISPs raising their data caps.
Download quota 
A download quota is a restriction fixed by some ISPs to limit the volume of data downloaded by the end user during a fixed period, usually a month. Once a fixed download cap (e.g. 250 gigabytes/month) has been reached by the end user, the speed at which they access the internet is usually throttled to a slower speed, cut off, or they are charged for excess data usage.