A collaborative congestion control method for high-speed network

Abstract

The invention discloses a collaborative work congestion control method used in high-speed networks. The network congestion state is judged by detecting the RTT delay information at the sending end and the 1-bit prediction information fed back by the router, and the congestion window is adaptively adjusted. The method of the present invention adjusts the congestion window under the high-speed network by combining the load factor of the network path and the number of data packets connected to the buffers of all routers on the path to adjust the congestion window under the high-speed network, which can effectively adapt to the transmission characteristics of the high-speed network, so as to ensure that the network Get better link utilization, TCP friendliness, and flow-to-flow fairness. Moreover, the unified integration of the two congestion information mechanisms can avoid misjudgment information generated by a single congestion detection mode in some environments, and reduce the direct impact of unilateral information errors on the reliability of the method.

Description

technical field [0001] The invention relates to a collaborative work congestion control method used in high-speed networks. Background technique [0002] Since the TCP Reno method was proposed in 1988, it has been considered as a very effective Internet network transmission control method, and has been accepted by the majority of scientific researchers and has been used until now. But after entering the 21st century, with the continuous rise and popularization of Gigabit networks, wireless networks, sensor networks and satellite networks, traditional TCP congestion control methods are facing great challenges in these network environments. Especially with the rapid development of network technology and the continuous improvement of access performance, today's interconnected backbone networks around the world present a network characteristic of high bandwidth and high latency. In this network environment, with the continuous increase of network bandwidth and round-trip delay,...

AI Technical Summary

Problems solved by technology

The high-speed flow seizes a large amount of available bandwidth of the bottleneck link, while the traditional TCP Reno flow only obtains very little throughput and is almost starved to death. This is what the method designer does not want to see

[0007] (2) RTT unfairness is serious

These methods not only improve the utilization rate of network bandwidth, but also indirectly increase the packet loss rate of the network.

[0013] (4) Unable to accurately judge network congestion

However, these methods are too complicated, and often need to insert a section of very long control information bits in the data packet header, which will be a big problem in realization for the TCP / IP protocol stack with little remaining space in the data packet. question

Images