RTT-based congestion control method and device combined with proportional-integral-derivative control

Abstract

The invention discloses a congestion control method and device based on RTT and combined with proportional-integral-derivative control, and the method comprises the steps that a sender judges whether RTT feedback timeout occurs or not, and measures the RTT time delay of a network path at the current moment; if RTT feedback is overtime, the sender actively decelerates; otherwise, according to whether the current RTT is within the tolerable queue oscillation range or not, different rate updating modes, including a method based on a fixed threshold value and a proportional integral differential control method, are employed to update the rate; and then the rate obtained by calculation is taken as the current target rate, and the time interval between the data packets is controlled so as to reach the target rate. According to the method, proportional integral differential control is carried out in a tolerable oscillation range, and RTT feedback timeout information is introduced, so the problems that large-scale incast cannot be processed and the queue oscillation range is uncontrollable in an existing congestion control method based on RTT are solved, and a stable and low-delay congestion queue is maintained.

Description

technical field [0001] The invention relates to the field of data communication, in particular to a method and equipment for controlling network congestion in a data center. Background technique [0002] With the development of hardware technology and the evolution of application requirements, data center networks have higher and higher requirements for microsecond-level ultra-low latency and high bandwidth. On the one hand, the development of hardware technology has made the underlying link bandwidth of the data center develop from the earliest 1Gbps to the current 100Gbps, and even has a 200Gbps link. High-speed links have given birth to a new data center architecture with resource decoupling. The high-speed network connects hardware resources such as CPU, GPU, memory, and disk, which facilitates the unified management and use of these hardware. On the other hand, with the emergence of new storage elements NVMe and new hardware such as GPUs and TPUs dedicated to machine ...

AI Technical Summary

Problems solved by technology

[0004] In view of the deficiencies of the above-mentioned prior art, the purpose of the present invention is to provide an RTT congestion control method that maintains stability and low delay, so as to solve the problem that the existing congestion control methods cannot be deployed on a large scale, cannot handle large-scale incasts, and rely on the underlying network Equipment and other issues; the present invention is a congestion control method based on RTT combined with proportional integral differential control implemented at the sending end. The whole process does not depend on the additional functions of the switch, and is still applicable to some network architectures without switches

Images