Congestion control method and system based on active queue debt compensation

Abstract

The invention provides a congestion control method and system based on active queue debt compensation, and the method introduces a queue debt mechanism in a BBR frame, and comprises the steps: initializing a state variable, and carrying out the zero setting of a last-round bandwidth estimation value and an accumulated queue debt; identifying a bandwidth detection behavior by comparing new and old bandwidth estimated values, calculating a single queue debt and accumulating the single queue debt to an accumulated queue debt; calculating a bandwidth delay product based on the current bandwidth and the minimum round-trip delay, and correcting the bandwidth delay product by using the cumulative queue debt to obtain a target in-transit data volume; limiting the in-transit data volume to the target in-transit data volume in an emptying stage of the BBR algorithm so as to actively repay the debt; and after emptying is finished, the debt is reset to finish the compensation cycle. According to the method, accurate quantification and active compensation of queue debt generated by bandwidth detection are realized, the inherent defects of queue accumulation and delay growth caused by periodic detection of a traditional BBR algorithm are overcome, and the network queuing delay is remarkably reduced while high throughput is maintained.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer network transmission control, and particularly relates to a congestion control method and system based on active queue debt compensation. BACKGROUND

[0002] With the popularity of modern network applications such as real-time video conferencing, interactive live streaming and cloud gaming, users' demand for low latency and high fluency is increasingly urgent. These applications are extremely sensitive to queuing delay, tail delay and short-term throughput fluctuation in network transmission. Although traditional congestion control algorithms based on packet loss feedback, such as TCP Reno and CUBIC, can guarantee high throughput, their passive response mechanism often leads to deep buffer queuing, causing unacceptable data transmission delay, which cannot meet the current demand.

[0003] Therefore, the research paradigm of congestion control has shifted to a model-driven route, with BBR (Bottleneck Bandwidth and Round-trip propagation time) algorithm proposed by Google as a representative. BBR congestion control algorithm no longer relies on packet loss as a congestion signal, but actively probes the bottleneck bandwidth and minimum round-trip time to control the data sending rate close to the ideal working point of network capacity, in order to balance high throughput and low delay. However, the periodic speed-up mechanism used by BBR congestion control algorithm to achieve bandwidth probing will inevitably inject traffic exceeding the instantaneous processing capacity into the network in practice, resulting in queue accumulation at the bottleneck link. If these "queue debts" caused by probing behavior are not completely emptied within the period, they will continue to accumulate in multiple probing periods, causing the minimum round-trip time measurement value to drift, and then causing delay inflation and throughput fluctuation.

[0004] To solve this problem, the subsequent versions of BBR congestion control algorithm, BBRv2 and BBRv3, have made important improvements. BBRv2 introduces a response mechanism to explicit congestion notification (ECN) and packet loss signals to alleviate congestion by dynamically adjusting the upper limit of in-transit data volume; BBRv3 further optimizes the probing duration and key parameters to improve bandwidth utilization and protocol fairness. However, these improvements still do not solve the fundamental "probing-accumulation" contradiction. Although the ProbeRTT mechanism used by BBRv3 can periodically empty the queue, it still only adopts a non-discriminatory speed-down strategy due to the lack of quantitative perception of historical accumulated debt, resulting in throughput fluctuation and inability to accurately clear the queue.

[0005] Therefore, there is an urgent need in the art for a new congestion control scheme capable of accurately perceiving and actively compensating for the queue debt caused by bandwidth probing to overcome the inherent defects of the existing BBR congestion control algorithm and its improved versions, such as the minimum round-trip time measurement drift, delay jitter, and throughput fluctuation caused by periodic probing. SUMMARY

[0006] In view of this, the embodiments of the present application provide a congestion control method and system based on active queue debt compensation to solve the inherent defects of the existing BBR congestion control algorithm and its improved versions, such as the minimum round-trip time measurement drift, delay jitter, and throughput fluctuation caused by periodic probing.

[0007] One aspect of the present application provides a congestion control method based on active queue debt compensation, which is implemented based on the network sending end in the BBR congestion control algorithm framework. The method comprises the following steps: initializing state variables, including setting the last round bandwidth estimation value and the cumulative queue debt to the initial value of zero; after completing each round of bandwidth estimation, comparing the new bandwidth estimation value with the last round bandwidth estimation value, and when the new bandwidth estimation value is greater than the last round bandwidth estimation value, calculating a single queue debt according to the product of the bandwidth increment and the minimum round-trip time, and accumulating the single queue debt to the cumulative queue debt; when it is necessary to set a target in-transit data amount, calculating a bandwidth delay product based on the current bandwidth estimation value and the minimum round-trip time, and correcting the bandwidth delay product according to the cumulative queue debt to determine the target in-transit data amount; when entering the emptying phase of the BBR congestion control algorithm, limiting the in-transit data amount of the sending end to the target in-transit data amount to actively repay the cumulative queue debt; after the end of the emptying phase, resetting the cumulative queue debt to the initial value.

[0008] In some embodiments of the present application, the emptying phase is the ProbeRTT phase in the BBR congestion control algorithm; and the step of obtaining the target in-transit data amount further comprises performing in the ProbeDown phase of the BBR congestion control algorithm.

[0009] In some embodiments of the present application, in the ProbeRTT phase, the target in-transit data amount is adjusted based on the cumulative queue debt, and the expression is: ; wherein, the target in-transit data amount is represented by D; the current bandwidth delay product is represented by BDP; the cumulative queue debt is represented by Q; and the minimum round-trip time is represented by RTT. represents an adjustment factor for controlling the compensation strength of the queue debt, and is greater than 0.

[0010] In some embodiments of the present application, the minimum round-trip delay is a minimum value obtained by a filtering algorithm based on the round-trip delay samples measured within a historical time window; and the filtering algorithm is a sliding window minimum filtering or a percentile filtering.

[0011] In some embodiments of the present application, the step of calculating the single queue debt and accumulating is performed after confirming that the new bandwidth estimation value is continuously higher than the last round bandwidth estimation value for more than a preset threshold time.

[0012] In some embodiments of the present application, the method further comprises: setting a dynamic upper limit threshold for the accumulated queue debt, the threshold being a predetermined proportion multiple of the current bandwidth delay product, the predetermined proportion multiple being greater than 0 and less than 1; when the accumulated queue debt exceeds the dynamic upper limit threshold, triggering the step of subtracting the accumulated queue debt from the bandwidth delay product to obtain the target in-transit data amount, and controlling data transmission according to the target in-transit data amount, regardless of whether the emptying phase is entered.

[0013] In some embodiments of the present application, the method further comprises: when the accumulated queue debt reaches a preset alarm threshold, generating alarm information related to the accumulation of the queue debt; sending the alarm information to a target object through a communication link, and storing the alarm information in a log database; wherein the log database also stores network state parameters associated with the alarm information at the time when the alarm is triggered.

[0014] In another aspect, the present application also provides a congestion control system based on active queue debt compensation, the system comprising: a data sending end configured to generate and send a data stream; a data receiving end configured to receive the data stream from the data sending end and return an acknowledgement information; wherein the data sending end is further configured to integrate a congestion control module for performing the steps of the method according to any one of the above.

[0015] In another aspect, the present application also provides a computer readable storage medium having a computer program or instructions stored thereon, the computer program or instructions being executed by a processor to implement the steps of the method according to any one of the above.

[0016] This invention provides a congestion control method and system based on proactive queue debt compensation. The method is implemented within the BBR congestion control algorithm framework and includes: initializing state variables, setting the previous round bandwidth estimate and accumulated queue debt to zero to establish an accurate benchmark for subsequent debt tracking; identifying bandwidth probing behavior by comparing old and new bandwidth estimates, calculating single queue debt and adding it to the accumulated queue debt to establish a complete debt tracking ledger, achieving precise quantification of the impact of historical probes; calculating the bandwidth-delay product based on the current bandwidth and minimum round-trip time, and using the accumulated queue debt to correct the bandwidth-delay product to obtain the target amount of data in transit, ensuring that the transmission rate is precisely matched with the actual network carrying capacity; limiting the amount of data in transit to the target value during the BBR emptying phase to proactively repay debt and effectively eliminate historical queue backlog; resetting the debt after emptying to complete the compensation cycle, ensuring that each probe cycle is independent and does not interfere with each other.

[0017] This invention achieves accurate quantification and proactive compensation for queue debt caused by bandwidth probing, overcoming the inherent defects of traditional BBR algorithms that lead to queue accumulation and latency inflation due to periodic probing. By significantly reducing queuing latency while maintaining high throughput, it effectively improves the network experience for latency-sensitive applications.

[0018] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the description, or may be learned by practice of the invention. The objects and other advantages of the invention can be realized and obtained by means of the structures specifically pointed out in the specification and drawings.

[0019] Those skilled in the art will understand that the objectives and advantages achievable with the present invention are not limited to those specifically described above, and that the above and other objectives achievable with the present invention will become clearer from the following detailed description. Attached Figure Description

[0020] The accompanying drawings, which are provided to further illustrate the invention and form part of this application, are not intended to limit the scope of the invention.

[0021] Figure 1 This is a flowchart illustrating a congestion control method based on active queue debt compensation according to an embodiment of the present invention.

[0022] Figure 2 This is a technical framework diagram of a congestion control method based on active queue debt compensation according to an embodiment of the present invention.

[0023] Figure 3 This is a comparison diagram of the round-trip delay distribution between the traditional BBR congestion control algorithm and the congestion control algorithm based on active queue debt compensation, as described in an embodiment of the present invention.

[0024] Figure 4 This is a comparison chart of throughput and tail round-trip time of the traditional BBR congestion control algorithm, Cubic congestion control algorithm, and active queue debt compensation-based congestion control algorithm under deep buffer conditions, according to an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0026] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.

[0027] Modern model-driven BBR congestion control algorithms rely on accurate estimations of bottleneck bandwidth and minimum round-trip time (RTT) to avoid deep queuing while maintaining high throughput. However, in practice, BBR congestion control algorithms often send data at a rate exceeding the estimated bandwidth for short periods to discover higher available bandwidth. While this proactive probing strategy helps improve bandwidth utilization, it also introduces brief but significant queue accumulation. These short-lived queues may not be completely cleared within a single probing cycle and accumulate gradually over multiple cycles, forming residual queues. This ultimately leads to an offset in the RTT measurement, increases tail latency, and impairs the experience of latency-sensitive applications such as video streaming, online gaming, and interactive live streaming. Existing ProbeRTT mechanisms lack awareness of queue accumulation during bandwidth probing and therefore cannot truly clear historical queues.

[0028] In view of this, one aspect of the present invention provides a congestion control method based on active queue debt compensation, the method being implemented at the network transmitter within the BBR congestion control algorithm framework, such as... Figure 1 As shown, the method includes the following steps S101~S105: S101: Initialize state variables, including setting the previous round's bandwidth estimate and accumulated queue debt to the initial value of zero.

[0029] S102: After completing each round of bandwidth estimation, compare the new bandwidth estimate with the previous bandwidth estimate. When the new bandwidth estimate is greater than the previous bandwidth estimate, calculate the single queue debt based on the product of the bandwidth increment and the minimum round-trip delay, and add the single queue debt to the cumulative queue debt.

[0030] S103: When it is necessary to set the target amount of data in transit, calculate the bandwidth-delay product based on the current bandwidth estimate and the minimum round-trip time, and correct the bandwidth-delay product according to the accumulated queue debt to determine the target amount of data in transit.

[0031] S104: When entering the emptying phase of the BBR congestion control algorithm, the amount of data in transit at the sending end is limited to the target amount of data in transit in order to actively repay the accumulated queue debt.

[0032] S105: After the emptying phase ends, reset the accumulated queue debt to its initial value.

[0033] In step S101, before starting the active queue debt compensation, the sending end establishes a definite initial state for the algorithm's state variables.

[0034] In this invention, the core state variables include, but are not limited to: the bandwidth estimate of the previous round and the accumulated queue debt.

[0035] The previous bandwidth estimate is used to record the previous bandwidth value and serves as a benchmark for determining whether bandwidth has increased.

[0036] Cumulative queue debt is used to quantify the total historical burden, and its value is equal to the sum of the "single queue debt" that has not yet been repaid due to bandwidth probes. This value reflects the total amount of residual data accumulated in the network due to historical probe activities.

[0037] The specific initialization process is as follows: the bandwidth estimate from the previous round is initialized to 0, indicating that no effective historical bandwidth benchmark has been established when the algorithm starts; the accumulated queue debt is initialized to 0, indicating that the sender starts running from a zero-debt state and there is no historical backlog.

[0038] In step S102, by quantifying each bandwidth increase into a specific single queue debt value, accurate accounting of historical detection behavior is achieved, providing core input parameters for precise compensation control based on the total debt scale in subsequent steps.

[0039] Specifically, after the BBR congestion control algorithm completes a round of bandwidth probing and obtains a new estimate, the sender immediately compares it with the stored bandwidth estimate from the previous round. If the new estimate is greater than the previous estimate, it indicates that a bandwidth probing action has occurred. At this point, the sender immediately calculates the queue debt generated by this probe, as follows: ; in, Indicates a single queue debt; This represents the new bandwidth estimate, which is the latest available bandwidth obtained during the current probe period; This represents the previous round of bandwidth estimation, serving as a historical bandwidth as a comparison benchmark. This represents the minimum round-trip time, and this represents the basic propagation time of the link.

[0040] The calculated single-queue debt is added to the cumulative queue debt in real time, thus completing the update of the debt ledger.

[0041] In some embodiments, the minimum round-trip time (RTT) is the minimum value obtained by filtering a sample of RTT measured within a historical time window; the filtering algorithm is either a sliding window minimum value filter or a percentile filter. By maintaining a historical time window and applying the filtering algorithm, network instantaneous jitter and measurement noise can be effectively resisted, and the minimum RTT sample that best represents the basic propagation delay of the link can be selected. This provides a stable and reliable benchmark for queue debt calculation and avoids debt estimation deviations caused by a single abnormal measurement.

[0042] In some embodiments, the step of calculating and accumulating single queue debt is performed after confirming that the new bandwidth estimate has been higher than the previous bandwidth estimate for a preset threshold time.

[0043] Specifically, by setting a preset threshold time, the system distinguishes between real, continuous bandwidth increases and short-term bandwidth estimation fluctuations, thereby improving the accuracy of debt tracking and the overall robustness of the system, and preventing unnecessary compensation operations due to instantaneous fluctuations.

[0044] In step S103, when it is necessary to update the target in-transit data volume, the sending end performs the core debt compensation calculation.

[0045] Specifically, the bandwidth-delay product is calculated based on the current bandwidth and the minimum round-trip time, and then corrected according to the accumulated queue debt to determine the target amount of data in transit. This operation is a direct manifestation of "debt repayment." By setting a lower target amount of data in transit, the sending end will correspondingly reduce the scheduling rate and congestion window, thereby proactively reducing the data transmission rate and creating space for network devices to clear historical accumulated queues.

[0046] In some embodiments, the bandwidth-delay product is calculated as follows: ; in, Represents the bandwidth-delay product; This represents the current bandwidth estimate; This represents the minimum round-trip time.

[0047] This calculation defines the ideal amount of data that fills the transmission channel without causing queuing under current network conditions.

[0048] In some embodiments, the target in-transit data volume is calculated as follows: ; in, This indicates the amount of data in transit for the target. This represents the accumulated queue debt.

[0049] This calculation is the core of debt compensation. By subtracting accumulated queue debt from the ideal data volume, it accurately calculates the target amount of data in transit that should be set to achieve queue clearance.

[0050] In some embodiments, an adjustment factor may be introduced in the step of subtracting the accumulated queue debt from the bandwidth-delay product described above. This is used to control the compensation level for queue debt, and its expression is: ; in, This represents the adjustment factor, used to control the level of compensation for queue debt, and its value is greater than 0.

[0051] By adjusting Values ​​can enable differentiated compensation strategies, allowing for flexible trade-offs between low latency and high throughput to adapt to the specific network performance requirements of different application scenarios. Specifically: when At this time, the sending end performs partial compensation. This strategy repays part of the historical debt while retaining a certain amount of data in transit to maintain higher throughput, making it suitable for scenarios with more stringent throughput requirements.

[0052] when When this happens, the sending end performs full compensation. This strategy can completely clear accumulated queue debt and is the standard solution for achieving the lowest queue latency.

[0053] when At this time, the sending end performs overcompensation. This strategy not only fully repays the debt but also further reduces the amount of data in transit, reserving more capacity for the network. It is suitable for demanding environments that are extremely sensitive to latency or need to prevent sudden congestion, but may cause a slight decrease in throughput.

[0054] In some embodiments, the step of obtaining the target amount of data in transit further includes performing the ProbeDown sub-phase of the BBR congestion control algorithm.

[0055] Specifically, in the BBR congestion control algorithm, the ProbeDown phase is a sub-phase of the ProbeBW state. This phase empties queues generated during the current probe period by moderately reducing the scheduling rate and congestion window. By initiating debt compensation in advance during this phase, the repayment pressure is distributed, resulting in smoother queue emptying and effectively avoiding large fluctuations in the transmission rate.

[0056] In step S104, when the BBR congestion control algorithm enters the emptying phase, the sending end actively repays the accumulated queue debt by setting the congestion window to the compensated target in-transit data amount calculated in step S103.

[0057] Specifically, the sender achieves this limitation by dynamically adjusting its congestion window (CWND) and pacing rate. The congestion window sets an upper limit on the amount of data in transit, while the pacing rate controls the time interval between packet transmissions to smooth traffic. By setting the congestion window to the target amount of data in transit and adjusting the pacing rate, the sender ensures that the actual amount of data in transit does not exceed this target value. This operation results in a data injection rate lower than the available bandwidth of the link, thus creating time and space for bottleneck devices in the network, such as routers, to empty their buffers.

[0058] In some embodiments, the emptying phase is the ProbeRTT phase in the BBR congestion control algorithm.

[0059] Specifically, ProbeRTT is a periodically triggered dedicated phase in the BBR congestion control algorithm used to measure minimum round-trip time (RTT). Traditional BBR congestion control algorithms crudely limit the amount of data in transit to an extremely low value, such as 4 packets, and maintain this for 200ms to reset the latency state. This invention, by using the target amount of data in transit as the new congestion window limit target at this phase, transforms the draining behavior from "blindly slowing down" to "slowing down on demand." This debt-aware draining strategy can maintain higher throughput performance while ensuring effective resetting of the RTT measurement, thus avoiding the precipitous drop in throughput caused by the excessively small fixed window in traditional mechanisms.

[0060] In step S105, once the emptying phase is successfully completed, the current round of proactive debt repayment based on the queue is considered finished. At this point, the sending end resets the accumulated queue debt to its initial value. This periodic debt reset operation ensures a closed-loop management of debt accumulation and repayment cycles. Simultaneously, it provides a clean initial state for the next round of bandwidth probing and debt tracking. This design, from a mechanism perspective, avoids stale debt data interfering with subsequent congestion control decisions, effectively guaranteeing the accuracy and stability of the algorithm during long-term operation.

[0061] In some embodiments, the above method further includes: A dynamic upper limit threshold is set for accumulated queue debt. This threshold is set as a predetermined proportional multiple of the current bandwidth-delay product, with a value greater than zero and less than one. This design allows the safety threshold to be dynamically adjusted according to the actual capacity of the network path. When the accumulated queue debt exceeds this dynamic upper limit threshold, regardless of the current state of the BBR congestion control algorithm, the sender will be forcibly interrupted from its normal process, immediately entering the emptying phase and executing step S104. This design effectively prevents debt from accumulating indefinitely under abnormal network conditions, significantly enhances the robustness of the algorithm, and ensures that network latency does not spiral out of control.

[0062] In some embodiments, the above method further includes a monitoring and alarm process: When the accumulated queue debt reaches a preset alarm threshold, alarm information related to the accumulation of queue debt is generated; the alarm information is sent to the target object through the communication link and stored in the log database; the log database also stores network status parameters related to the alarm information and the time of alarm triggering.

[0063] On the other hand, the present invention also provides a congestion control system based on active queue debt compensation, the system comprising: The data sender is configured to generate and send data streams; The data receiver is configured to receive the data stream from the data sender and return an acknowledgment message; The data sending end is further configured to integrate a congestion control module for performing the steps of any of the methods described above.

[0064] On the other hand, the present invention also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implement the steps of any of the methods described above.

[0065] The present invention will now be described with reference to a specific embodiment: This embodiment details the design and implementation of a congestion control method based on Active Queue Debt Compensation (AQDC-BBR), which is a lightweight extension solution to address the structural problem of queue accumulation during periodic bandwidth probing in existing BBR congestion control algorithms.

[0066] 1. Problem Modeling and Core Ideas Modern model-driven BBR congestion control algorithms, in their core ProbeBW (Bandwidth Probe) state, periodically enter a transmission rate increase phase (ProbeUp sub-state) to probe available bandwidth. This injects additional data into the network beyond its instantaneous processing capacity, creating queue backlog. If these short-term queues are not completely cleared, they accumulate over multiple cycles as "residual queues," causing a shift in the minimum round-trip time measurement and leading to tail latency growth, which impairs the user experience for latency-sensitive applications such as video streaming and online gaming. Existing ProbeRTT mechanisms lack awareness of historical queue accumulation and cannot completely clear these residual queues.

[0067] To address this issue, this embodiment proposes the concept of Queue Debt, which is formally defined as follows: In a single ProbeBW state cycle, at the ProbeUp sub-state, due to the bandwidth estimate from... Upgraded to The amount of additional data injected within a minimum round-trip time window is expressed as: ; in, Indicates the first Single queue debt resulting from secondary bandwidth probe; Indicates the first Estimated wheel bandwidth.

[0068] After multiple probing cycles, the debt from these individual cycles will accumulate to form the total queue debt, expressed as: ; in, This represents the accumulated queue debt.

[0069] This accumulated queueing debt occupies the link buffer, effectively increasing the network latency observed at the sender. Therefore, the measured round-trip time increases proportionally with the total queueing debt, as expressed by: ; in, This represents the observed round-trip time, which is the actual round-trip time measured by the sending end, including queuing delay. This indicates the bottleneck link bandwidth.

[0070] If the sending end continues to probe without compensation, the round-trip latency will increase exponentially with each cycle.

[0071] Therefore, the traditional ProbeRTT phase should not only be used to resample to minimize round-trip time, but its more critical function is to proactively repay the queue debt accumulated in the ProbeBW cycle.

[0072] Specifically, if a new delivery rate is observed in a flow during the ProbeUp sub-phase... ,and Compared to the last ProbeRTT If there is an improvement, the additional queues injected during this period can be explicitly quantified as follows: .

[0073] In subsequent ProbeBW cycles, this type of debt continued to accumulate, and its corresponding This reflects the buffer space indirectly occupied by the stream due to bandwidth contention. During the ProbeRTT phase, this is achieved by reducing the amount of data in transit for the target. This can drive the link to clear these historical backlogs, thereby allowing the RTT measurement to return to the true minRTT level.

[0074] The fundamental purpose of this method is to optimize the trade-off between latency and throughput. When accumulated queue debt exists, its effective throughput can be approximated as: ; in, Indicates throughput.

[0075] As queue backlogs lead to increased RTT (Round-Trip Time), throughput may not only fail to improve but could even decrease if the amount of data in transit is not reduced accordingly. This invention addresses this issue by maintaining high throughput while stabilizing RTT through a debt repayment mechanism, thereby significantly improving the performance of latency-sensitive applications such as real-time video streaming.

[0076] The core idea of ​​this method is to explicitly identify, quantify, and repay the "queue debt" caused by bandwidth probing, thereby achieving more precise control over latency while maintaining high throughput.

[0077] 2. System Architecture and Working Mechanism This design introduces two lightweight functional modules into the BBR framework: a queue debt tracker and an in-transit data compensator. Figure 2 As shown.

[0078] The queue debt tracker is responsible for entering the ProbeUp (uplink probing) sub-phase in the ProbeBW state. Once a bandwidth increase is observed, it calculates the single queue debt and continuously accumulates it to the cumulative queue debt. The in-transit data compensator is responsible for dynamically adjusting the target in-transit data volume using the cumulative debt in the ProbeDown (downlink probing) sub-phase of the ProbeBW state or the independent ProbeRTT phase. By limiting the congestion window of the sender to this target in-transit data volume, it actively repays historical debt and empties the queue.

[0079] 3. Verification of beneficial effects To verify the practical effectiveness of this invention, the proposed AQDC-BBR congestion control algorithm was deployed and run in a network test environment, and compared with the standard BBR congestion control algorithm. The test environment simulated a deep buffer scenario in a real network, which is prone to queue accumulation and latency inflation.

[0080] 1) Latency performance comparison The delay control effect of the present invention is quantitatively verified by statistically analyzing the cumulative distribution function (CDF) and probability density function (PDF) of round-trip time (RTT).

[0081] Figure 3 Round-trip delay distributions for BBRv1, BBRv2, BBRv3, and the AQDC-BBR proposed in this invention are presented. Results show that, among all tested BBR versions, the AQDC-BBR congestion control algorithm, which integrates an active queue debt compensation mechanism, consistently reduces tail latency.

[0082] Compared to the unmodified version, AQDC-BBR achieves an improvement in round-trip latency at the 99th percentile. This improvement is particularly significant in BBRv2 and BBRv3.

[0083] 2) Trade-off between throughput and latency Figure 4 The results demonstrate a comparison of throughput and tail round-trip time under deep buffer conditions. The test results further confirm the advantages of this invention in complex network environments.

[0084] When faced with extremely large buffers, the baseline BBR version systematically fills them, resulting in a 99th percentile round-trip time (RRT) of up to 230 milliseconds, four times the network baseline propagation delay. While these versions can maintain high throughput, this severe self-induced high latency makes them unsuitable for interactive applications. In contrast, the AQDC-BBR congestion control algorithm of this invention, under the same adverse network conditions, has a 99th percentile RRT of only 150 milliseconds, significantly lower than the standard BBR congestion control algorithm, and is comparable to the algorithm in tight buffers. The performance remains almost consistent across different environments. This demonstrates the consistency and robustness of the algorithm's latency control capabilities across various network scenarios.

[0085] More importantly, while achieving the aforementioned significant latency reduction, the throughput loss caused by the AQDC-BBR congestion control algorithm is negligible. Compared with the standard BBR congestion control algorithm, the throughput difference between the two is less than 0.5%.

[0086] This set of data strongly demonstrates that the present invention can achieve precise and effective control of queuing delay, especially tail delay, with almost no sacrifice in throughput. It breaks through the inherent trade-off of "reducing delay must sacrifice throughput" in traditional congestion control, and reflects the technical advancement and practical value of the method of the present invention.

[0087] In summary, this invention provides a congestion control method and system based on proactive queue debt compensation. The method is implemented within the BBR congestion control algorithm framework and includes: initializing state variables, setting the previous round's bandwidth estimate and accumulated queue debt to zero to establish an accurate benchmark for subsequent debt tracking; identifying bandwidth probing behavior by comparing old and new bandwidth estimates, calculating single queue debt and adding it to the accumulated queue debt to establish a complete debt tracking ledger, achieving precise quantification of the impact of historical probes; calculating the bandwidth-delay product based on the current bandwidth and minimum round-trip time, and using the accumulated queue debt to correct this bandwidth-delay product to obtain the target amount of data in transit, ensuring that the transmission rate accurately matches the actual network capacity; limiting the amount of data in transit to the target value during the BBR emptying phase to proactively repay debt and effectively eliminate historical queue backlog; resetting the debt after emptying to complete the compensation cycle, ensuring that each probe cycle is independent and does not interfere with each other.

[0088] This invention achieves accurate quantification and proactive compensation for queue debt caused by bandwidth probing, overcoming the inherent defects of traditional BBR algorithms that lead to queue accumulation and latency inflation due to periodic probing. By significantly reducing queuing latency while maintaining high throughput, it effectively improves the network experience for latency-sensitive applications.

[0089] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the desired tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave.

[0090] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0091] In this invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.

[0092] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A congestion control method based on active queue debt compensation, characterized in that, The method is implemented on the network transmitter within the BBR congestion control algorithm framework, and includes the following steps: Initialize state variables, including setting the previous round's bandwidth estimate and accumulated queue debt to their initial values ​​of zero; After each round of bandwidth estimation, the new bandwidth estimate is compared with the previous bandwidth estimate. When the new bandwidth estimate is greater than the previous bandwidth estimate, the single queue debt is calculated based on the product of the bandwidth increment and the minimum round-trip delay, and the single queue debt is added to the cumulative queue debt. When it is necessary to set the target amount of data in transit, the bandwidth-delay product is calculated based on the current bandwidth estimate and the minimum round-trip time, and the bandwidth-delay product is corrected according to the accumulated queue debt to determine the target amount of data in transit. When entering the emptying phase of the BBR congestion control algorithm, the amount of data in transit at the sending end is limited to the target amount of data in transit in order to actively repay the accumulated queue debt. After the emptying phase is completed, the accumulated queue debt is reset to the initial value.

2. The congestion control method based on active queue debt compensation according to claim 1, characterized in that, The emptying phase is the ProbeRTT phase in the BBR congestion control algorithm; and the step of obtaining the target amount of data in transit also includes the ProbeDown phase of the BBR congestion control algorithm.

3. The congestion control method based on active queue debt compensation according to claim 2, characterized in that, During the ProbeRTT phase, the target in-transit data volume is compensated and adjusted based on the accumulated queue debt, expressed as: ； in, The target data volume in transit is represented by BDP; BDP represents the current bandwidth-delay product. This represents the accumulated queue debt; This represents the adjustment factor, used to control the level of compensation for queue debt, and its value is greater than 0.

4. The congestion control method based on active queue debt compensation according to claim 1, characterized in that, The minimum round-trip time is the minimum value obtained by filtering a sample of round-trip time measured within a historical time window; the filtering algorithm is either sliding window minimum value filtering or percentile filtering.

5. The congestion control method based on active queue debt compensation according to claim 1, characterized in that, The step of calculating and accumulating single queue debt is performed after confirming that the new bandwidth estimate has been continuously higher than the previous bandwidth estimate for a preset threshold time.

6. The congestion control method based on active queue debt compensation according to claim 1, characterized in that, The method further includes: A dynamic upper limit threshold is set for the accumulated queue debt, which is a predetermined ratio multiple of the current bandwidth-delay product, and the predetermined ratio multiple is greater than zero and less than one. When the accumulated queue debt exceeds the dynamic upper limit threshold, regardless of whether it is in the emptying phase, the step of subtracting the accumulated queue debt from the bandwidth delay product to obtain the target amount of data in transit is triggered, and data transmission is controlled according to the target amount of data in transit.

7. The congestion control method based on active queue debt compensation according to claim 3, characterized in that, The method further includes: When the accumulated queue debt reaches a preset alarm threshold, alarm information related to the accumulation of queue debt is generated; the alarm information is sent to the target object through a communication link and stored in a log database; wherein, the log database also stores network status parameters related to the alarm information and the alarm trigger time.

8. A congestion control system based on active queue debt compensation, characterized in that, The system includes: The data sender is configured to generate and send data streams; The data receiver is configured to receive the data stream from the data sender and return an acknowledgment message; The data sending end is further configured to integrate a congestion control module for performing the steps of the method as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 7.

Images