Available bandwidth estimation method and system for cloud desktop scene, and storable medium

Abstract

The invention discloses a cloud desktop scene-oriented available bandwidth estimation method and system and a storable medium. The method comprises the following steps of: sampling a bandwidth, capturing a link state at a current moment and performing statistical analysis; link evaluation: estimating a current available broadband value and guiding a video encoder according to a link evaluation result; feedback is adjusted, and the sending rate is adjusted to be matched with the load capacity of the current link. The method has the advantages of improving the link utilization rate and realizing high-precision and high-throughput bandwidth estimation on the premise of ensuring the advantages of light weight and low time delay based on the heuristic algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of cloud desktop available bandwidth estimation, and in particular to a cloud desktop scenario-oriented available bandwidth estimation method, system and storable medium. Background Art

[0002] At present, with the gradual popularization of application scenarios such as real-time interactive video transmission and cloud desktop, applications such as online courses and online office meetings have been promoted unprecedentedly. Network communications continue to develop steadily, network infrastructure is becoming increasingly complete, and emerging businesses such as cloud computing and big data are showing a rapid growth trend. The number of IoT users has exceeded 2.3 billion, and the number of 5G base stations has reached 3.377 million. Looking to the future, with the further promotion of 5G and 6G mobile communication technologies, as well as the widespread application of new data transmission applications such as 3D volumetric video interaction technology and remote robot collaboration technology, accurate network bandwidth estimation will put forward higher requirements.

[0003] Bandwidth estimation algorithms can be divided into deep learning-based and heuristic algorithms. The bandwidth estimation algorithm based on deep learning estimates the available network bandwidth through neural networks and system models. Through the training of a large amount of offline data, the model can often learn the potential characteristics of the link, thereby achieving high-precision link available bandwidth estimation. This type of algorithm has strong generalization ability and high bandwidth utilization, but often has high overhead, high latency and poor robustness. At the same time, there is an inevitable problem that a large amount of data needs to be trained in advance. This obviously cannot meet the requirements of low latency and fast startup of real-time interactive data transmission in cloud desktop scenarios. Therefore, the present invention will focus on the study of bandwidth estimation algorithms based on heuristics. Traditional bandwidth estimation algorithms based on heuristics usually rely on empirical rules and prior knowledge, and form general rules for bandwidth estimation by summarizing the laws of bandwidth usage in historical data. It is based on statistical characteristics in specific network scenarios, such as time, user behavior patterns, latency, packet loss rate, etc., to derive the bandwidth changes at the current moment. The heuristic bandwidth estimation algorithm has low overhead and high computational efficiency, but in order to avoid congestion, it often adopts a more conservative strategy, so the accuracy and bandwidth utilization are relatively low. The WebRTC GCC (Google congestion control) algorithm proposed by G Carlucc et al. is mainly based on delayed congestion detection and uses adaptive control strategies to optimize the packet sending rate. However, when the link changes rapidly, the convergence speed of the algorithm is often slow, which is obviously not enough to cope with the current growing network demand and high-precision and low-latency user experience requirements. Therefore, the focus of the present invention is how to improve link utilization and realize a high-precision and high-throughput bandwidth estimation algorithm while ensuring the lightweight and low-latency advantages of the heuristic algorithm.

[0004] Available bandwidth estimation for cloud desktop scenarios requires low-latency communication between the sender and receiver because of real-time interactive data transmission. Secondly, the dynamic changes in network link status are more complex than those in static networks. The system is required to understand the relationship between network load and actual available bandwidth at each moment after capturing the network status in real time. At the same time, for different network conditions, the system must be able to adopt the optimal sending strategy to maximize bandwidth utilization while avoiding congestion. N Cardwell et al. proposed the BBR (Bottleneck Bandwidth and Round-trip propagation time) algorithm, which established a congestion control mechanism based on bandwidth and delay estimation by comprehensively considering bottleneck bandwidth and round-trip delay. It ensures that the network utilization is close to the maximum available bandwidth through continuous estimation and adjustment, thereby avoiding the queuing phenomenon in traditional congestion control algorithms. However, due to the periodic restriction of transmission to detect available bandwidth, it will cause higher delays during the minRTT detection period.

[0005] In summary, there is at least one of the following technical problems:

[0006] Although the bandwidth estimation algorithm based on deep learning has advantages in estimation accuracy and generalization ability, the complexity of the dynamic network environment places higher requirements on the data volume, and its complex computing model leads to high computing delay and resource consumption, which makes it difficult to meet the cloud desktop scenarios with high real-time requirements.

[0007] Heuristic bandwidth estimation algorithms rely on empirical rules and statistical characteristics. Although they have high computational efficiency and low resource overhead, they often use conservative estimation methods in complex or volatile network environments, which reduces bandwidth utilization. Their accuracy is relatively insufficient, making it difficult to fully utilize the available bandwidth of the network. Most congestion control algorithms have a conflict between bandwidth utilization and congestion control, and it is difficult to completely eliminate high latency during bandwidth detection.

[0008] How to improve link utilization and implement a high-precision, high-throughput bandwidth estimation algorithm while ensuring the lightweight and low-latency advantages of heuristic algorithms. Summary of the invention

[0009] The main purpose of the present invention is to provide an available bandwidth estimation method, system and storable medium for cloud desktop scenarios, so as to solve the problem in the prior art of how to improve link utilization and realize a high-precision and high-throughput bandwidth estimation algorithm while ensuring the lightweight and low-latency advantages of heuristic algorithms.

[0010] In order to achieve the above object, according to one aspect of the present invention, a method for estimating available bandwidth for a cloud desktop scenario is provided, comprising:

[0011] Step 1: Bandwidth sampling, capturing the current link status and performing statistical analysis;

[0012] Step 2: Link evaluation, estimating the current available bandwidth value and guiding the video encoder based on the link evaluation result;

[0013] Step 3: Adjust the feedback and adjust the sending rate to match the load capacity of the current link.

[0014] Preferably, step 1 includes: constructing a bandwidth sampler to capture the network status at the current moment, inputting the captured link status information into a link evaluator, and generating a bandwidth sampling result; step 2 includes: using a link evaluator with frame-level grouping dispersed statistical data to analyze the relationship between the link bandwidth sampling value at each moment and the existing sending strategy, accurately determine the saturation state of the current link, discover the hidden features of the link, and input them into an adaptive module; step 3 includes: the adaptive module autonomously selects the optimal sending strategy adjustment method based on the specific network results of the link evaluation and the estimated available bandwidth value, so that the sending rate at the next moment matches the actual available bandwidth.

[0015] Preferably, the core objective of the bandwidth sampler is to obtain link status information and discover additional available bandwidth by dispersing statistical data of frame-level grouping, thereby maximizing bandwidth utilization.

[0016] Preferably, in step 1, the bandwidth sampling of the current transmission bit rate is performed by a bandwidth sampler based on frame-level grouping distributed statistical data to obtain a bandwidth sampling value S and a small frame sampling value S small , the specific calculation method is:

[0017] Bandwidth sampling value

[0018] Where F is the size of the frame, S last and S start Respectively represent the actual sending time of the first split small frame and the last small frame in each complete frame;

[0019] The small frame sampling value is the average sending rate of multiple small frames.

[0020]

[0021] Among them, F small Indicates the data size of each small frame, Send i and Send i-1 Respectively represent the actual sending time of the current small frame and the previous small frame.

[0022] Preferably, in step 2, the bandwidth sampling value S and the small frame sampling value S small And the current transmission rate B = F / I, evaluate the current link status, which specifically includes three conditions:

[0023] Condition 1: C 1 =S / S small , Condition 2: C 2 =S / B, condition 3: C 3 =S small / αβ

[0024] Where α is the sensitivity adjustment coefficient. If any condition C 1 , C 2 or C 3 If it is less than 1, the link is determined to be in a saturated state, and the feedback flag P is set to 0. Otherwise, it is an undersaturated state, and the feedback flag P is set to 1.

[0025] Preferably, in step 3, according to the feedback result of the link evaluation, the system adjusts the sending rate according to the continuity of the current state:

[0026] When a link is in a saturated or undersaturated state for many consecutive times, the current sending rate obviously does not match the actual network bandwidth. In this case, a more aggressive jump adjustment method is adopted, which includes significantly reducing the sending rate in a saturated state to avoid link congestion, and quickly injecting traffic to fill the link bandwidth in an undersaturated state.

[0027] When saturation and undersaturation states appear alternately, the link load fluctuates greatly and the sending rate is close to the actual available bandwidth. A conservative linear adjustment strategy is adopted to balance the load through small linear adjustments to avoid drastic rate fluctuations.

[0028] Preferably, bandwidth estimation and rate adjustment are performed to record the bandwidth sampling value S when the link is in a saturated state. list And the small frame sampling value S small_list , by calculating the average value of the sampled values, the current available bandwidth can be obtained as:

[0029] min(average(S list ),average(S small_list ))

[0030] Based on the estimated available bandwidth, the system can use jump regulation to quickly adjust the transmission rate to match the actual bandwidth when in an oversaturated state, thereby avoiding link congestion while ensuring high bandwidth utilization.

[0031] Preferably, link status recording and cleaning optimizes bandwidth utilization. After each link evaluation, the system will add the current feedback result to the feedback list, and add the bandwidth sampling value and small frame sampling value to S list and S small_list In the process, when the list length reaches the specified threshold, the system cleans up the old data according to the latest link evaluation results and retains the latest status information for the next decision.

[0032] According to another aspect of the present invention, there is provided an available bandwidth estimation system for a cloud desktop scenario, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method according to any one of claims 1 to 8 when executing the program.

[0033] According to another aspect of the present invention, there is provided a storable medium for estimating available bandwidth for a cloud desktop scenario, comprising: a computer program stored thereon, characterized in that when the program is executed by a processor, the steps of the method described in any one of claims 1-8 are implemented.

[0034] The application of the technical solution of the present invention has the following technical effects:

[0035] A frame-level bandwidth sampler is used to capture the sending time interval of each packet frame, and the link hidden characteristics are captured by the delay changes of the sending queue. It does not depend on specific network conditions, can flexibly and quickly adapt to changes in network conditions and provide more accurate bandwidth estimation.

[0036] Bandwidth utilization is maximized by dispersing statistics through frame-level grouping to discover additional available bandwidth.

[0037] Dynamic adjustments are made between frames based on the estimated network status and available bandwidth. By analyzing the scattered statistics of frame-level grouping, that is, the intervals between the transmission of multiple small frames, the bandwidth sampler can discover the additional available bandwidth on the link. When each group of complete frames is sent, a bandwidth sampling value can be generated. By comparing and analyzing the time intervals between the transmission of small frames of the complete frame at the current moment, the trend of link status changes during the transmission of this complete frame can be inferred.

[0038] Set the link estimator to be more sensitive to congestion conditions.

[0039] By flexibly applying aggressive and conservative adjustment modes in the sending strategy, the efficiency and stability of network transmission can be effectively balanced, which not only avoids resource waste, but also can cope with sudden network conditions and prevent network congestion caused by excessive adjustment.

[0040] The dynamic adjustment mechanism combining linear and jump adjustment modes can flexibly respond to network fluctuations under different network conditions. When the network load is relatively stable, the linear adjustment mode can smoothly adjust the sending bit rate to maintain a high network utilization rate. When the network status changes significantly, the jump adjustment mode can respond quickly to avoid bandwidth waste or congestion, ensuring the efficiency and stability of data transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The drawings constituting a part of the present application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0042] Figure 1 A method flow chart of the available bandwidth estimation method for cloud desktop scenarios according to the present invention is shown;

[0043] Figure 2 Shows Figure 1 Schematic diagram of the overall communication framework of the available bandwidth estimation method for cloud desktop scenarios;

[0044] Figure 3 Shows Figure 1 The overall flow chart of the available bandwidth estimation method for cloud desktop scenarios in ;

[0045] Figure 4 Shows Figure 1 The link state graph captured by the bandwidth sampler of the available bandwidth estimation method for cloud desktop scenarios in ;

[0046] Figure 5 Shows Figure 1 Figure 2. Link evaluator model diagram of the available bandwidth estimation method for cloud desktop scenarios in FIG.

[0047] Figure 6 Shows Figure 1 The linear adjustment model diagram of the available bandwidth estimation method for cloud desktop scenarios in FIG.

[0048] Figure 7 Shows Figure 1 Figure 1 shows a jump adjustment model of the available bandwidth estimation method for cloud desktop scenarios;

[0049] Figure 8 Shows Figure 1 Feedback detection model diagram of the available bandwidth estimation method for cloud desktop scenarios;

[0050] Fig. 9 Shows Figure 1 Bandwidth sampling model diagram of the available bandwidth estimation method for cloud desktop scenarios. DETAILED DESCRIPTION

[0051] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0052] like Figures 1 to 9 As shown, an embodiment of the present invention provides an available bandwidth estimation method for a cloud desktop scenario, including: step 1: bandwidth sampling, capturing the link status at the current moment and performing statistical analysis; step 2: link evaluation, estimating the current available bandwidth value and guiding the video encoder according to the link evaluation result; step 3: adjusting feedback, adjusting the sending rate to match the load capacity of the current link.

[0053] In this embodiment, the link status at the current moment is captured and statistically analyzed, the current available bandwidth value is estimated, and the video encoder is guided according to the link evaluation result to adjust the sending rate to match the load capacity of the current link. It mainly includes the following three parts:

[0054] First, a bandwidth sampler is constructed to capture the current network status, and the captured link status information is input into the link evaluator to generate bandwidth sampling results.

[0055] Second, the link evaluator that uses frame-level grouping dispersed statistical data analyzes the relationship between the link bandwidth sampling value at each moment and the existing sending strategy, accurately determines the saturation state of the current link, discovers the hidden features of the link, and inputs them into the adaptive module.

[0056] Third, the adaptive module autonomously selects the optimal sending strategy adjustment method based on the specific network results of the link evaluation and the estimated available bandwidth value, so that the sending rate at the next moment matches the actual available bandwidth.

[0057] This embodiment provides a low-latency and high-precision adaptive bandwidth estimation method based on frame-level grouping distributed statistical data. The method includes three parts:

[0058] First, a bandwidth sampler based on frame-level packet dispersion statistics.

[0059] First, in order to address the problems of low bandwidth utilization, poor accuracy, and weak generalization ability of traditional heuristic algorithms, while retaining the advantages of low latency and low overhead of heuristic algorithms, the present invention adopts a frame-level bandwidth sampler to capture the sending time interval of each packet frame, and captures the hidden characteristics of the link by changing the delay of the sending queue. It does not depend on specific network conditions, can flexibly and quickly adapt to changes in network conditions, and provide more accurate bandwidth estimation.

[0060] The core goal of the bandwidth sampler is to obtain the link status information and discover additional available bandwidth by dispersing statistical data through frame-level grouping, thereby maximizing bandwidth utilization. For an application that generates frames at a fixed frame rate for data transmission, a frame with a data size of F is sent every fixed time interval I. Then, the average frame transmission bit rate at this time can be expressed as:

[0061] B=F / I

[0062] Among them, B represents the average frame transmission bit rate; F represents the size of a data sent; I represents each fixed time interval;

[0063] On this basis, each frame is divided into multiple small frames for transmission, and the transmission time of the small frames is set to be transmitted within the time interval I. When the current bandwidth is not fully utilized or is just fully utilized, the link at this time is in an undersaturated or just saturated state, so no additional transmission queue or congestion will occur between frames. In this case, the small frames can be fully transmitted before the next frame arrives, so the small frames split from the previous complete frame will not affect the original transmission plan of the next frame. When the bandwidth exceeds its carrying capacity, that is, the frame transmission rate exceeds the bandwidth carrying capacity of the link, the link is in an oversaturated state, and congestion will occur between frames, resulting in additional transmission queue delays. At this time, the small frames at the current moment will be delayed in transmission due to the failure of the small frames at the previous moment to be fully transmitted.

[0064] The bandwidth samples collected at each moment during the end-to-end transmission process can be expressed by the following formula:

[0065]

[0066] Among them, S last and S start They represent the actual sending time of the first split small frame and the last small frame in each complete frame, so the actual sending time of each complete frame is S last -S start ; S represents the bandwidth sample value; F represents the size of a data sent.

[0067] When the link is in an undersaturated state, since no transmission queue is generated, the bandwidth sample value S will be greater than the average transmission code rate B, that is, greater than the throughput of the link at that moment, and the code rate can be increased. When the link is in an oversaturated state, transmission lag occurs, and due to the transmission queue, the bandwidth sample value S will be less than the average transmission code rate B. On the other hand, due to the accumulation at the sending end, the consumption rate of the data queue will be equal to the actual available bandwidth, and the bandwidth sample value at this time will be equal to the actual available bandwidth.

[0068] One of the advantages of this design is that it can dynamically adjust between frames based on the estimated network status and available bandwidth. By analyzing the scattered statistics of frame-level grouping, that is, the sending intervals of multiple small frames, the bandwidth sampler can discover the additional available bandwidth on the link. Another advantage is that a bandwidth sampling value can be generated when each group of complete frames is sent. By comparing and analyzing the small frame sending time intervals of the complete frame at the current moment, the changing trend of the link status during the sending of this complete frame can be inferred, which is recorded as the small frame sampling value, that is, the average sending rate of small frames during the complete frame:

[0069]

[0070] Among them, F small Indicates the data size of each small frame, Send i and Send i-1 Respectively represent the actual sending time of the current small frame and the previous small frame. By capturing the link change trend during the sending of the complete frame, that is, the bandwidth sampling value S and the small frame sampling value S small As well as the sending code rate B of the current complete frame, this information can be passed to the link evaluation module to perform corresponding network status evaluation and sending code rate adjustment.

[0071] Second, a link evaluator is constructed to identify potential network congestion and link characteristics.

[0072] Based on the bandwidth sampler, the present invention further captures the bandwidth sampling value S and the small frame sampling value S during the transmission of the complete frame. small The transmission code rate B of the complete frame is analyzed to evaluate the saturation state of the current link and the actual network load capacity, thereby maximizing the network bandwidth utilization.

[0073] Specifically, the main function of the link evaluator is to accurately evaluate the match between the actual load capacity of the current link and the transmission throughput. During this process, the bandwidth sampler provides captured link status information, including bandwidth sampling values ​​and small frame sampling values. By comparing the bandwidth sampling value, small frame sampling value and transmission rate, the link evaluator can determine whether the current link has reached the maximum available bandwidth and whether the link is in a saturated state. Link evaluation is usually performed at the end of the transmission of each complete frame, and the results of each evaluation are recorded. Ultimately, the trend of the link evaluator's results can provide a basis and guidance for adaptively adjusting the data transmission rate.

[0074] The link assessment module can obtain the following three conditions by analyzing the bandwidth sampling value:

[0075] Condition 1: C 1 =S / S small , Condition 2: C 2=S / B, condition 3: C 3 =S small / αB

[0076] Among them, α is the link congestion sensitivity adjustment coefficient, which is used to balance the sensitivity in the evaluation process.

[0077] In particular, considering that in practical applications, data congestion has a more significant impact on user experience than low throughput, the link estimator is set to be more sensitive to congestion. 1 , C 2 or C 3 If the value is less than 1, the evaluation module will determine that the current link is saturated or oversaturated, that is, the load capacity of the current link is insufficient to digest the real-time data flow. In other words, the bandwidth of the link is close to or exceeds its actual carrying capacity, resulting in potential queue accumulation and data delay. On the contrary, if all conditions are greater than or equal to 1, the link evaluator believes that the link is undersaturated, that is, the load capacity of the link has not been fully utilized, and there is available bandwidth redundancy.

[0078] At the same time, the current link is evaluated after each bandwidth sampling. The evaluation results include link saturation and unsaturation. If the link appears multiple saturation states or unsaturation states in succession, it means that the current bit rate is seriously mismatched with the network. A more aggressive strategy adjustment method is adopted to quickly adapt to changes in network conditions. If the current link saturation state and unsaturation state appear alternately, a more conservative strategy adjustment method is adopted. This conservative strategy helps to avoid link congestion caused by excessive adjustment, thereby maintaining the stability of data transmission. By flexibly applying adjustment strategies based on the evaluation results of the link status, the utilization of network resources can be better optimized, transmission efficiency can be improved, and the risk of network congestion can be reduced.

[0079] Third, dynamically adjust the sending rate based on the adaptive network status.

[0080] Based on the aforementioned bandwidth sampler and link evaluator, the present invention uses two different adjustment methods to adjust the current transmission code rate. In an actual network environment, if there is a significant gap between the currently estimated available bandwidth and the actual transmission rate, a more aggressive transmission strategy adjustment method, i.e., jump adjustment, is usually adopted to maximize bandwidth utilization. The purpose of this strategy is to quickly adapt to the current link conditions and make full use of unoccupied bandwidth resources. If the gap between the estimated available bandwidth and the current transmission rate is small, it is necessary to consider the volatility problem in the actual network. In this case, the link status, especially the saturation of the link, and the current transmission rate should be comprehensively evaluated. In order to avoid link congestion caused by an overly aggressive adjustment strategy, a more conservative transmission strategy adjustment mode, i.e., linear adjustment, should be adopted. This conservative strategy aims to reduce the risk of performance degradation caused by network fluctuations or sudden congestion and ensure the stability of the transmission process. By flexibly applying aggressive and conservative adjustment modes in the transmission strategy, the efficiency and stability of network transmission can be effectively balanced, which can avoid resource waste, cope with sudden network conditions, and prevent network congestion problems caused by excessive adjustment.

[0081] Specifically, when the linear adjustment mode is used, the adjustment of the sending code rate follows the principle of gradual change. After each link evaluation, the algorithm will linearly increase or decrease the sending code rate according to the current link status to ensure that the data transmission rate matches the network load capacity. This method is relatively smooth and suitable for use when the network fluctuation is small to avoid sudden adjustments from interfering with the link. When jump adjustment is used, the adjustment of the sending code rate is faster and more aggressive. When the link is continuously in an undersaturated state, it means that the current sending code rate is seriously lower than the actual carrying capacity of the network, and the available bandwidth is not fully utilized. To this end, the algorithm will significantly increase the sending code rate by quickly injecting more traffic, so that the link reaches the saturation state as soon as possible, thereby maximizing the bandwidth utilization. When the link is continuously in an oversaturated state, it indicates that the current sending code rate is significantly higher than the actual load capacity of the link, resulting in link congestion. At this time, the algorithm needs to quickly reduce the sending rate to avoid further congestion. In order to accurately estimate the available bandwidth, the algorithm calculates the bandwidth sampling values ​​and small frame sampling values ​​recorded in multiple link saturation states. The estimated value of the available bandwidth can be calculated by the following formula:

[0082] min(average(S list ),average(S small_list ))

[0083] Among them, S list and S small_list The bandwidth sampling values ​​and small frame sampling values ​​are recorded under multiple link saturation states, and finally the sending bit rate is adjusted according to the estimated available bandwidth value.

[0084] Among them, S list Indicates a list of bandwidth sampling values ​​recorded under link saturation conditions, S small_list is the corresponding small frame sampling value list. By calculating the average of these values, a more conservative estimate of the available bandwidth under the current network state can be obtained. Based on this estimate, the algorithm will adjust the sending bit rate to make it more consistent with the actual load capacity of the link.

[0085] This dynamic adjustment mechanism that combines linear and jump adjustment modes can flexibly respond to network fluctuations under different network conditions. When the network load is relatively stable, the linear adjustment mode can smoothly adjust the sending bit rate to maintain a high network utilization rate. When the network status changes significantly, the jump adjustment mode can respond quickly to avoid bandwidth waste or congestion, ensuring the efficiency and stability of data transmission.

[0086] According to another aspect of the present invention, there is provided an available bandwidth estimation system for a cloud desktop scenario, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method according to any one of claims 1 to 8 when executing the program.

[0087] According to another aspect of the present invention, there is provided a storable medium for estimating available bandwidth for a cloud desktop scenario, comprising: a computer program stored thereon, characterized in that when the program is executed by a processor, the steps of the method described in any one of claims 1-8 are implemented.

[0088] like Figure 2 As shown, the present invention deploys the available bandwidth estimation method at the transmitting end, and the specific flow chart is as follows Figure 3 As shown, the present invention proposes an available bandwidth estimation method based on frame-level bandwidth sampling, which aims to maximize bandwidth utilization and reduce congestion by dynamically adjusting the sending rate through real-time evaluation of network status. The method mainly includes the following steps:

[0089] S1: Bandwidth sampling of the current sending rate.

[0090] First, during the transmission process, the data of each complete frame is split into multiple small frames for transmission. The bandwidth sampling of the current transmission bit rate is performed through the bandwidth sampler based on the frame-level grouping dispersed statistical data to obtain the bandwidth sampling value S and the small frame sampling value S small , the specific calculation method is:

[0091] Bandwidth sampling value Where F is the size of the frame, S last and S startThey represent the actual sending time of the first split small frame and the last small frame in each complete frame respectively, and the small frame sampling value is the average sending rate of multiple small frames. Among them, F small Indicates the data size of each small frame, Send i and Send i-1 Respectively represent the actual sending time of the current small frame and the previous small frame.

[0092] S2: The link evaluator performs evaluation and judgment.

[0093] Through the bandwidth sampling value S and the small frame sampling value S small And the current transmission rate B = F / I, evaluate the current link status. Specifically, it includes three conditions:

[0094] Condition 1: C 1 =S / S small , Condition 2: C 2 =S / b, condition 3: C 3 =S small / αB

[0095] Where α is the sensitivity adjustment coefficient. If any condition C 1 , C 2 or C 3 If it is less than 1, the link is determined to be in a saturated state, and the feedback flag P is set to 0. Otherwise, it is an undersaturated state, and the feedback flag P is set to 1.

[0096] S3: Adaptive adjustment strategy.

[0097] According to the feedback results of the link evaluation, the system will adjust the sending rate according to the continuity of the current state. When the link is in a saturated or undersaturated state for many consecutive times, it means that the current sending rate obviously does not match the actual network bandwidth. At this time, a more aggressive jump adjustment method is adopted, which is specifically divided into a significant reduction in the sending rate in the saturated state to avoid link congestion, and a rapid injection of traffic to fill the link bandwidth in the undersaturated state. When the saturated and undersaturated states appear alternately, it means that the link load fluctuates greatly at this time, and the sending rate is close to the actual available bandwidth. A conservative linear adjustment strategy is adopted to balance the load through small linear adjustments to avoid drastic rate fluctuations.

[0098] S4: Bandwidth estimation and rate adjustment.

[0099] When the link is oversaturated, in order to more accurately evaluate the available bandwidth, the sending rate is reduced to match the actual network. Record the bandwidth sampling value S when the link is in a saturated state list And the small frame sampling value S small_listBy calculating the average value of the sampled values, the current available bandwidth can be obtained as:

[0100] min(averagr(S list ),average(S small_list ))

[0101] Based on the estimated available bandwidth, the system can use jump regulation to quickly adjust the transmission rate to match the actual bandwidth when in an oversaturated state, thereby avoiding link congestion while ensuring high bandwidth utilization.

[0102] S5: Link status recording and cleaning.

[0103] In order to optimize bandwidth utilization, the system will add the current feedback result P to the feedback list P after each link evaluation. list At the same time, the bandwidth sampling value and the small frame sampling value are added to S list and S small_list When the list length reaches the specified threshold, the system will clean up the old data according to the latest link evaluation results and retain the latest status information for the next decision.

[0104] From the above description, it can be seen that the above embodiments of the present invention achieve the following technical effects:

[0105] A frame-level bandwidth sampler is used to capture the sending time interval of each packet frame, and the link hidden characteristics are captured by the delay changes of the sending queue. It does not depend on specific network conditions, can flexibly and quickly adapt to changes in network conditions and provide more accurate bandwidth estimation.

[0106] Bandwidth utilization is maximized by dispersing statistics through frame-level grouping to discover additional available bandwidth.

[0107] Dynamic adjustments are made between frames based on the estimated network status and available bandwidth. By analyzing the scattered statistics of frame-level grouping, that is, the intervals between the transmission of multiple small frames, the bandwidth sampler can discover the additional available bandwidth on the link. When each group of complete frames is sent, a bandwidth sampling value can be generated. By comparing and analyzing the time intervals between the transmission of small frames of the complete frame at the current moment, the trend of link status changes during the transmission of this complete frame can be inferred.

[0108] Set the link estimator to be more sensitive to congestion conditions.

[0109] By flexibly applying aggressive and conservative adjustment modes in the sending strategy, the efficiency and stability of network transmission can be effectively balanced, which not only avoids resource waste, but also can cope with sudden network conditions and prevent network congestion caused by excessive adjustment.

[0110] The dynamic adjustment mechanism combining linear and jump adjustment modes can flexibly respond to network fluctuations under different network conditions. When the network load is relatively stable, the linear adjustment mode can smoothly adjust the sending bit rate to maintain a high network utilization rate. When the network status changes significantly, the jump adjustment mode can respond quickly to avoid bandwidth waste or congestion, ensuring the efficiency and stability of data transmission.

[0111] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for estimating available bandwidth for cloud desktop scenarios, characterized in that: include: Step 1: Bandwidth sampling, capturing the current link status and performing statistical analysis; Step 2: Link evaluation, estimating the current available bandwidth value and guiding the video encoder based on the link evaluation result; Step 3: Adjust the feedback and adjust the sending rate to match the load capacity of the current link.

2. The available bandwidth estimation method for cloud desktop scenarios according to claim 1, characterized in that: The step 1 includes: constructing a bandwidth sampler to capture the network status at the current moment, inputting the captured link status information into a link evaluator, and generating a bandwidth sampling result; the step 2 includes: using a link evaluator with frame-level grouping dispersed statistical data to analyze the relationship between the link bandwidth sampling value at each moment and the existing sending strategy, accurately determine the saturation state of the current link, discover the hidden features of the link, and input them into an adaptive module; the step 3 includes: the adaptive module autonomously selects the optimal sending strategy adjustment method according to the specific network results of the link evaluation and the estimated available bandwidth value, so that the sending rate at the next moment matches the actual available bandwidth.

3. The available bandwidth estimation method for cloud desktop scenarios according to claim 1, characterized in that: The core goal of the bandwidth sampler is to obtain link status information and discover additional available bandwidth by dispersing statistical data through frame-level grouping, thereby maximizing bandwidth utilization.

4. The available bandwidth estimation method for cloud desktop scenarios according to claim 1, characterized in that: In step 1, the bandwidth sampling of the current transmission bit rate is performed by a bandwidth sampler based on frame-level grouping distributed statistical data to obtain a bandwidth sampling value S and a small frame sampling value S small , the specific calculation method is: Bandwidth sampling value Where F is the size of the frame, S last and S start Respectively represent the actual sending time of the first split small frame and the last small frame in each complete frame; The small frame sampling value is the average sending rate of multiple small frames. where F small Indicates the data size of each small frame, Send i and Send i-1 Respectively represent the actual sending time of the current small frame and the previous small frame.

5. The available bandwidth estimation method for cloud desktop scenarios according to claim 1, characterized in that: In step 2, the bandwidth sampling value S and the small frame sampling value S are used. small And the current transmission rate B = F / I, evaluate the current link status, which specifically includes three conditions: Condition 1: C1 = S / S small , Condition 2: C2 = S / B, Condition 3: C3 = S small / αB Wherein, α is the sensitivity adjustment coefficient. If any condition C1, C2 or C3 is less than 1, the link is judged to be in a saturated state, and the feedback flag P is set to 0. Otherwise, it is an undersaturated state, and the feedback flag P is set to 1.

6. The available bandwidth estimation method for cloud desktop scenarios according to claim 1, characterized in that: In step 3, according to the feedback result of the link evaluation, the system adjusts the sending rate according to the continuity of the current state: When a link is in a saturated or undersaturated state for many consecutive times, the current sending rate obviously does not match the actual network bandwidth. In this case, a more aggressive jump adjustment method is adopted, which includes significantly reducing the sending rate in a saturated state to avoid link congestion, and quickly injecting traffic to fill the link bandwidth in an undersaturated state. When saturation and undersaturation states appear alternately, the link load fluctuates greatly and the sending rate is close to the actual available bandwidth. A conservative linear adjustment strategy is adopted to balance the load through small linear adjustments to avoid drastic rate fluctuations.

7. The available bandwidth estimation method for cloud desktop scenarios according to claim 1, characterized in that: Bandwidth estimation and rate adjustment, recording the bandwidth sampling value S when the link is in a saturated state list and small frame sampling value S small_list , by calculating the average value of the sampled values, the current available bandwidth can be obtained as: min(average(S list ),average(S small_list )) Based on the estimated available bandwidth, the system can use jump regulation to quickly adjust the transmission rate to match the actual bandwidth when in an oversaturated state, thereby avoiding link congestion while ensuring high bandwidth utilization.

8. The method for estimating available bandwidth for a cloud desktop scenario according to claim 1, characterized in that: Link status recording and cleaning optimizes bandwidth utilization. After each link evaluation, the system will add the current feedback results to the feedback list, and add the bandwidth sampling value and small frame sampling value to S list and S small_list In the process, when the list length reaches the specified threshold, the system cleans up the old data according to the latest link evaluation results and retains the latest status information for the next decision.

9. A system for estimating available bandwidth for cloud desktop scenarios, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images