Control of best effort traffic

The Internet is almost exclusively based on the concept of best effort transmission. The simplicity of this model certainly played a key role in the deployment of , a ubiquitous Internet service. However, a completely uncontrolled network may suffer from congestion collapse, which occurs when the offered load locally exceeds the available hand­ width. Such collapses already occurred in the mid-'80s, when the Internet had no deployed congestion avoidance. Repeated congestion problems triggered the definition and implementation of congestion control functions in TCP (thus, end to end), similar to those found in DECNET. Their original design required avoiding modifications to routers; congestion control was introduced as a set of functions at sources and destinations, whereas routers continued to handle all traffic in an aggregate way. With some minor modifications, these mechanisms arc still used to: • Protect the Internet from congestion collapse • Make all users share the available bandwidth in a "fair" way The efficiency of these mechanisms depends heavily on user equipment correctly implementing congestion control functions.