180 likes | 340 Views
Promoting the Use of End-To-End Congestion Control in the Internet. IEEE/ACM Transaction on Networking Vol.7 No.4 August 1999 Sally Floyd and Kevin Fall. Abstract. Potentially negative impacts of an increasing deployment of non-congestion-controlled best-effort traffic on the Internet
E N D
Promoting the Use of End-To-End Congestion Control in the Internet IEEE/ACM Transaction on Networking Vol.7 No.4 August 1999 Sally Floyd and Kevin Fall
Abstract • Potentially negative impacts of an increasing deployment of non-congestion-controlled best-effort traffic on the Internet • These impacts range from extreme unfairness against competing TCP traffic to the potential for congestion collapse.
Introduction • Promote the end-to-end congestion control: • Router mechanisms • Restrict the bandwidth of selected high-bandwidth best-effort flows • Approaches: • identify a high bnadwidth flow as Unresponsive • Not TCP-friendly • Using disproportionate bandwidth
Possible Approaches • Involves the deployment of packet scheduling disciplines in routers that isolate each flow(per-flow scheduling mechanisms) • (in this paper)for routers to support the continued use of end-to-end congestion control to share scarce bandwidth • Rely on financial incentives or pricing mechanisms to control sharing
TCP UDP Definition: GoodPut: A Flow as the bandwidth delivered to the recerver,excluding duplicate packets. Unresponsive flows: flows that do not use end-to-end congestion control,and do not reduce their load on the network when subjected to packet drops.
WRR:each flow assigned an equal weight in units of bytes/s 25%
Danger of Congestion Collapse • Classical congestion collapse • Largely due to TCP connection unnecessarily retransmitting packets • Corrected by the timer improvements and congestion-control mechanism in modern implementations of TCP
(cont’d) • Congestion collapse from undelivered packets • Bandwidth is wasted by delivering packet through the network that are dropped before reaching their ultimate destination • (different from classical)the degraded condition is not stable,but returns to normal once the load is reduced
Set bandwidth 128 kb/s, 9% Bandwidth wasted by the packets that never reach dest. UDP flow restricted to 25% of the link bandwidth
10% 35%
Prevent Congestion Collapse from undelivered packets • To Succeed,through incentives at routers,in maintaining an environment characterized by end-to-end congestion control • Maintain a virtual-circuit-style environment where packets are prevented from entering the network unless the network has sufficient reources to deliver
Identify flow to regulate(A) • TCP-friendly flows:if its arrival rate does not exceed the arrival of conformant TCP connection in the same circumstances. • Assume a flow should not use more bandwidth than would the most aggressive conformant TCP implementation in the same circumstances.
Maximum sending rate:T bytes/s • Maximum packet size:B bytes • Minimum round-trip time R • Packet drop rate p • Can Calculate the max arrival rate • Limitations:Can only be applied to a flow at the level of granularity of a single TCP connection (hard to determine B,R)
Identify flow to regulate(B) • Unresponsive:its arrival rate decreases appropriately in response to an increased packet drop rate. • If the packet drop rate of the connection increases by a factor of x, then the arrival rate from the source should decrease by a factor of roughly
(Cont’d) • If steady-state drop rate increases by a factor x and the presented load for a high-bandwidth flow does not decrease by a factor reasonably close to or more,then the flow can be deemed not to be using congestion control(unresponsive)
Conclusion • In this paper has argued on the need for end-to-end congestion control, and further,on the need for mechanisms in the network to detect and restrict unreoponsive or high bandwidth best-effort flows in times of congestion • In this paper have not yet oulined a specific proposal for mechanisms for identifying and controlling unresponsive flows.
Future Work • The most important issue is not the precise functioning of the mechanisms to restrict the bandwidth of unresponsive best-effort flows,but simply that such mechanisms be deployed. • Mechanisms such as these would go a long way to making concrete the essential role played by congestion control for best-effort traffic in the Internet.