Abstract: The Internet research community is promoting active queue management in routers as a proactive means of addressing congestion in the Internet. Active queue management refers to managing the length of an outbound queue in a router by selectively dropping packets during times of congestion. Active queue management mechanisms such as Random Early Detection (RED) have been shown to work well for TCP flows but can fail in the presence of unresponsive UDP flows such as high-bandwidth video flows. Recent proposals extend RED to strongly favor TCP and TCP-like flows and to actively penalize unresponsive or so-called misbehaving flows. This is problematic for multimedia flows that, although potentially well-behaved, do not, or can not, satisfy the definition of a TCP-like flow.
In this talk I discuss the congestion control problem for multimedia traffic on the Internet and describe an extension to RED active queue management called Class-Based Thresholds (CBT). CBT attempts to realize a "better than best effort" service for well-behaved multimedia flows by active queue management. The goal of CBT is to reduce congestion in routers and to protect TCP from all UDP flows while also ensuring acceptable throughput and latency for well-behaved UDP flows. Experimental results show that CBT realizes these goals with lower state requirements and implementation complexity than other active queue management mechanisms.