Method for generating ecn threshold configuration information and congestion control system

Abstract

The embodiment of the application discloses a kind of ECN threshold configuration information generation method and congestion control system, it is related to computer technical field.The ECN threshold configuration information generation method includes: receiving the current traffic state information sequence of network node device;Based on the traffic model file of the network node device, whether the traffic state information sequence is the traffic model in the traffic model file is identified;When the traffic state information sequence is not the traffic model in the traffic model file, the traffic state information sequence is input into the ECN threshold prediction model corresponding to the network node device to carry out ECN threshold configuration information prediction, the ECN threshold configuration information predicted by the ECN threshold prediction model is determined as the current ECN threshold configuration information of the network node device;Wherein, the ECN threshold prediction model corresponding to the network node device takes the sample traffic state information sequence of the network node device as input, takes the sample ECN threshold configuration information corresponding to the sample traffic state information sequence as output label, is trained to obtain LSTM model, and the sample ECN threshold configuration information is the ECN threshold configuration information that can take into account rat flow and elephant flow.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of computers, and in particular to an ECN threshold configuration information generation method and a congestion control system. BACKGROUND

[0002] At present, the ECN threshold value of the intelligent lossless network RoCEv2 is usually configured manually according to expert experience, and is manually modified according to the change of the network model in the later period.

[0003] The ECN threshold value configured manually cannot take into account the time-delay-sensitive mouse flow and the throughput-sensitive elephant flow existing in the network at the same time. When the ECN threshold is set too low, the ECN congestion mark can be triggered as soon as possible to inform the source server to reduce the speed, so as to maintain a low cache depth (i.e. a low queue delay), which is beneficial to the time-delay-sensitive mouse flow. However, the ECN threshold that is too low will affect the throughput-sensitive elephant flow, limit the traffic bandwidth of the elephant flow, and cannot meet the high throughput of the elephant flow. When the ECN threshold is set too high, the time for triggering the ECN congestion mark can be prolonged to guarantee the burst absorption capacity of the queue and meet the traffic bandwidth of the throughput-sensitive elephant flow. However, when the queue is congested, due to the large cache, queue queuing will occur, causing a large queue delay, which is not beneficial to the time-delay-sensitive mouse flow.

[0004] In addition, the ECN threshold value configured manually cannot cope with the burst traffic in the Incast scenario. The ECN threshold value configured manually is usually set to a fixed ECN threshold according to the maximum link bandwidth accessed in order to guarantee no packet loss, which makes it easy for the uplink and downlink queues in the same switch in the high-performance distributed application to appear the Incast scenario of N to 1. SUMMARY

[0005] The embodiments of the present application provide an ECN threshold configuration information generation method and a congestion control system to dynamically generate a suitable ECN threshold, take into account the time-delay-sensitive mouse flow and the throughput-sensitive elephant flow existing in the network at the same time, and cope with the burst traffic in the Incast scenario.

[0006] In a first aspect, the embodiments of the present application provide an ECN threshold configuration information generation method, which comprises:

[0007] receiving a current traffic state information sequence of a network node device; wherein the traffic state information sequence is a time sequence sequence composed of traffic state information at each time point of the network node device within a specified time length up to the current time point, and the traffic state information at each time point includes the traffic state information of each virtual channel of the network node device at the time point;

[0008] identify, based on a traffic model file of the network node device, whether the traffic state information sequence is an existing traffic model in the traffic model file; wherein the traffic model file stores a plurality of existing traffic models, and each traffic model is a traffic state information sequence;

[0009] when the traffic state information sequence is not an existing traffic model in the traffic model file, input the traffic state information sequence into an ECN threshold prediction model corresponding to the network node device to perform ECN threshold configuration information prediction, and determine the ECN threshold configuration information predicted by the ECN threshold prediction model as the current ECN threshold configuration information of the network node device; wherein the ECN threshold configuration information includes an ECN threshold value of each virtual channel of the network node device; the ECN threshold prediction model corresponding to the network node device is trained by taking a sample traffic state information sequence of the network node device as input and taking sample ECN threshold configuration information corresponding to the sample traffic state information sequence as output label, and the sample ECN threshold configuration information is ECN threshold configuration information that can take into account both mouse flow and elephant flow.

[0010] In a second aspect, the embodiments of the present application provide a congestion control system, which comprises a network node device, a collector and an ECN threshold configuration automatic generation component.

[0011] The network node device is configured to periodically push traffic state information of the network node device at a current time point to the collector.

[0012] The collector is configured to combine the traffic state information of the network node device at the current time point with traffic state information of the network node device at other time points within a specified time length before the current time point, to obtain a current traffic state information sequence of the network node device, and send the traffic state information sequence to the ECN threshold configuration automatic generation component.

[0013] The ECN threshold configuration automatic generation component is configured to detect, based on a traffic model file of the network node device, whether the traffic state information sequence is an existing traffic model in the traffic model file, and when the traffic state information sequence is not an existing traffic model in the traffic model file, input the traffic state information sequence into an ECN threshold prediction model corresponding to the network node device to perform ECN threshold configuration information prediction, and determine the ECN threshold configuration information predicted by the ECN threshold prediction model as the current ECN threshold configuration information of the network node device and return the ECN threshold configuration information to the network node device; wherein the traffic model file stores a plurality of existing traffic models, and each traffic model is a traffic state information sequence; the ECN threshold prediction model corresponding to the network node device takes a sample traffic state information sequence of the network node device as input, takes sample ECN threshold configuration information corresponding to the sample traffic state information sequence as output label, is trained on an LSTM model, and the sample ECN threshold configuration information is ECN threshold configuration information that can take both mouse flow and elephant flow into account;

[0014] The network node device is configured to configure the ECN threshold of the virtual channel of the network node device based on the current ECN threshold configuration information of the network node device to perform congestion control on the virtual channel of the network node device.

[0015] In a third aspect, an embodiment of the present application provides an ECN threshold configuration information generation device, and the device comprises:

[0016] An information receiving module is configured to receive a current traffic state information sequence of a network node device; wherein the traffic state information sequence is a time sequence composed of traffic state information of the network node device at each time point within a specified time length up to a current time point, and the traffic state information at each time point comprises traffic state information of each virtual channel of the network node device at the time point;

[0017] An identification module is configured to identify, based on a traffic model file of the network node device, whether the traffic state information sequence is an existing traffic model in the traffic model file; wherein the traffic model file stores a plurality of existing traffic models, and each traffic model is a traffic state information sequence;

[0018] predict a ECN threshold configuration information by inputting the traffic state information sequence into the ECN threshold prediction model corresponding to the network node device, and determine the ECN threshold configuration information predicted by the ECN threshold prediction model as the current ECN threshold configuration information of the network node device; wherein the ECN threshold configuration information comprises an ECN threshold value of each virtual channel of the network node device; the ECN threshold prediction model corresponding to the network node device is trained by taking a sample traffic state information sequence of the network node device as input and taking a sample ECN threshold configuration information corresponding to the sample traffic state information sequence as output label, and the sample ECN threshold configuration information is an ECN threshold configuration information capable of considering both mouse flow and elephant flow.

[0019] In a fourth aspect, an electronic device is provided, which comprises a processor and a memory. The memory stores a computer program. The computer program is executed by the processor to implement the method for generating ECN threshold configuration information according to the first aspect of the embodiments of the present application.

[0020] In a fourth aspect, an electronic device is provided, which comprises a processor and a memory. The memory stores a computer program. The computer program is executed by the processor to implement the method for generating ECN threshold configuration information according to the first aspect of the embodiments of the present application.

[0021] In the embodiments of the present application, the ECN threshold prediction model corresponding to the network node device is trained by taking a sample traffic state information sequence of the network node device as input and taking a sample ECN threshold configuration information corresponding to the sample traffic state information sequence as output label, and the sample ECN threshold configuration information is an ECN threshold configuration information capable of considering both mouse flow and elephant flow. When the current traffic state information sequence of the network node device is received, whether the traffic state information sequence is a traffic model already existing in the traffic model file of the network node device is identified based on the traffic model file of the network node device. When the traffic state information sequence is not a traffic model already existing in the traffic model file, the traffic state information sequence is input into the ECN threshold prediction model corresponding to the network node device to predict an ECN threshold configuration information, and the ECN threshold configuration information predicted by the ECN threshold prediction model is determined as the current ECN threshold configuration information of the network node device. The ECN threshold prediction model of the network node device can dynamically generate a suitable ECN threshold configuration information capable of considering both mouse flow and elephant flow for the network node device, which facilitates the network node device to consider both mouse flow and elephant flow based on the dynamic suitable ECN threshold configuration information and cope with burst traffic in the Incast scenario. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the one or more embodiments of the present application or the prior art, the drawings needed to be used in the description of the embodiments will be briefly introduced. Obviously, the drawings in the following description are only drawings of the one or more embodiments of the present application, and other drawings can be obtained by those of ordinary skill in the art without any creative effort on the basis of these drawings.

[0023] Figure 1 is a schematic diagram of an application scenario of an ECN threshold configuration information generation method provided by some embodiments of the present application;

[0024] Figure 2 is a schematic diagram of a flow of an ECN threshold configuration information generation method provided by some embodiments of the present application;

[0025] Figure 3 is a schematic diagram of a training flow of an ECN threshold prediction model provided by some embodiments of the present application;

[0026] Figure 4 is a schematic diagram of an acquisition flow of a training data set provided by some embodiments of the present application;

[0027] Figure 5 is a schematic diagram of a structure of a congestion control system provided by some embodiments of the present application;

[0028] Figure 6 is a schematic diagram of a module of an ECN threshold configuration information generation apparatus provided by some embodiments of the present application;

[0029] Figure 7 is a schematic diagram of a structure of an electronic device provided by some embodiments of the present application. DETAILED DESCRIPTION

[0030] The embodiments of the present application provide an ECN threshold configuration information generation method and an intelligent lossless network management platform, which can be used in an intelligent lossless network based on a RoCEv2 protocol, to dynamically generate suitable ECN threshold configuration information for network node devices in the network, so that the ECN threshold of the network can take into account both latency-sensitive mouse flows and throughput-sensitive elephant flows existing in the network at the same time, and cope with burst traffic in an Incast scenario.

[0031] In order to enable a person skilled in the art to better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be clearly and completely described in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by a person skilled in the art without creative labor should be within the scope of protection of the present application.

[0032] Among them, the terms "first", "second" and the like in the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances to enable the embodiments of the present application to be implemented in an order other than those illustrated or described here.

[0033] Figure 1 is a schematic diagram of an application scenario of the ECN threshold configuration information provided by some embodiments of the present application. As shown in Figure 1 , in this application scenario, it includes a plurality of network node devices 110 of an intelligent lossless network and an intelligent lossless network management platform 120, wherein the intelligent lossless network management platform 120 further includes a collector 121 and an ECN threshold configuration automatic generation component 122.

[0034] The generation method of the ECN threshold configuration information provided by the embodiments of the present application can be applied in the ECN threshold configuration automatic generation component 122.

[0035] In implementation, the network node device 110 can periodically push the traffic state information of the network node device 110 at the current time point to the collector 121 in the intelligent lossless network management platform 120.

[0036] The collector 121 can combine the traffic state information of the network node device 110 at the current time point with the traffic state information of the network node device 110 at other time points within a specified length of time before the current time point to obtain the traffic state information sequence of the network node device 110 at the current time point, and send the traffic state information sequence to the ECN threshold configuration automatic generation component 122 in the intelligent lossless network management platform 120.

[0037] The ECN threshold configuration automatic generation component 122 can detect whether the traffic status information sequence is an existing traffic model in the traffic model file of the network node device 110. If the traffic status information sequence is not an existing traffic model in the traffic model file, the component inputs the traffic status information sequence into the ECN threshold prediction model corresponding to the network node device to predict the ECN threshold configuration information. The component then determines the ECN threshold configuration information predicted by the ECN threshold prediction model as the current ECN threshold configuration information of the network node device and returns it to the network node device 110.

[0038] The network node device 110 can configure the ECN threshold of its virtual channel based on the current ECN threshold configuration information of the network node device to perform congestion control on its virtual channel.

[0039] Figure 2 This is a flowchart illustrating a method for generating ECN threshold configuration information according to some embodiments of this application. The following is a summary of the process. Figure 2 The method for generating ECN threshold configuration information provided in the embodiments of this application will be described in detail.

[0040] like Figure 2 As shown, in some embodiments of this application, the method for generating ECN threshold configuration information includes:

[0041] S201, Receive the current traffic status information sequence of the network node device.

[0042] In this embodiment, the current traffic status information sequence of the network node device is a time-series sequence composed of traffic status information of the network node device at each time point within a specified time length up to the current time point. The traffic status information at each time point includes: the traffic status information of each virtual channel of the network node device at that time point. The traffic status information may include: cache space size, cache occupancy rate, bandwidth throughput, and other information.

[0043] For example, assuming network node device 1 has 3 virtual channels, the current traffic status information sequence of network node device 1 can be:

[0044]

[0045] in, This represents the traffic status information of network node 1 at time T.

[0046]

[0047] in, the flow state information of the third virtual channel of the network node 1 at the Tth time point.

[0048] S202, identifying whether the flow state information sequence is an existing flow model in the flow model file based on the flow model file of the network node device.

[0049] In the embodiments of the present application, the flow model file stores a plurality of existing flow models, and each flow model is a flow state information sequence.

[0050] In specific implementation, the current flow state information sequence of the network node device can be compared with the existing flow models in the flow model file. If the current flow state information sequence is the same as a certain flow model in the flow model file, it is determined that the current flow state information sequence is the existing flow model in the flow model file. If the current flow state information sequence is not the same as each flow model in the flow model file, it is determined that the current flow state information sequence is not the existing flow model in the flow model file.

[0051] S203, when the flow state information sequence is not the existing flow model in the flow model file, inputting the flow state information sequence into the ECN threshold prediction model corresponding to the network node device to perform ECN threshold configuration information prediction, and determining the ECN threshold configuration information predicted by the ECN threshold prediction model as the current ECN threshold configuration information of the network node device.

[0052] In the embodiments of the present application, the ECN threshold configuration information of the network node device includes the ECN threshold value of each virtual channel of the network node device.

[0053] For example, according to the above example, inputting the current flow state information sequence of the network node device 1 into the ECN threshold prediction model 1 of the network node device 1, and the ECN threshold configuration information output by the ECN threshold prediction model 1 is:

[0054]

[0055] wherein, the ECN threshold value of the first virtual channel of the network node device 1, the ECN threshold value of the second virtual channel of the network node device 1, the ECN threshold value of the second virtual channel of the network node device 1.

[0056] In the embodiments of the present application, the ECN threshold prediction model corresponding to the network node device is trained by taking the sample traffic state information sequence of the network node device as input, and taking the sample ECN threshold configuration information corresponding to the sample traffic state information sequence as output label.

[0057] In specific implementation, the traffic state information sequence can be input into the ECN threshold prediction model corresponding to the network node device. After receiving the traffic state information sequence, the ECN threshold prediction model can perform ECN threshold configuration information prediction based on the traffic state information sequence, obtain predicted ECN threshold configuration information, and then output the predicted ECN threshold configuration information. Then, the predicted ECN threshold configuration information output by the ECN threshold prediction model can be determined as the current ECN threshold configuration information of the network node device. By continuously receiving the current traffic state information sequence of the network node device, inputting the traffic state information sequence into the ECN threshold prediction model corresponding to the network node device for ECN threshold configuration information prediction, and determining the predicted ECN threshold configuration information as the current ECN threshold configuration information of the network node device, the ECN threshold configuration information of the network node device can be dynamic ECN threshold configuration information.

[0058] In the embodiments of the present application, the ECN threshold prediction model of the network node device is trained by taking the sample traffic state information sequence of the network node device as input, and taking the sample ECN threshold configuration information corresponding to the sample traffic state information sequence as output label, which can enable the ECN threshold prediction model to predict suitable ECN threshold configuration information that can take into account both mouse flow and elephant flow. After receiving the current traffic state information sequence of the network node device, inputting the current traffic state information sequence of the network node device into the ECN threshold prediction model for ECN threshold configuration information prediction, the current suitable ECN threshold configuration information of the network node device can be obtained, which takes into account both mouse flow and elephant flow. By continuously inputting the current traffic state information sequence of the network node device into the ECN threshold prediction model for ECN threshold configuration information prediction, the dynamic suitable ECN threshold configuration information of the network node device can be obtained, which can cope with burst traffic in the Incast scenario.

[0059] Figure 3 is a training flow diagram of the ECN threshold prediction model provided by some embodiments of the present application. As shown in Figure 3 In some embodiments of the present application, the ECN threshold prediction model of the network node device can be trained by the following method.

[0060] S301, input the sample traffic state information sequence of the current sample in the training data set of the network node device into the LSTM model, to obtain the predicted ECN threshold configuration information output by the LSTM model.

[0061] In the embodiments of the present application, the current sample is the sample of this iteration, which is updated at each iteration. For example, at the first iteration, the current sample is the first sample in the training data set, and at the second iteration, the current sample is the second sample in the training data set.

[0062] In the embodiments of the present application, before step S301, the training data set of the network node device can be obtained. Figure 4 is the flowchart of the training data set acquisition process provided by some embodiments of the present application. As shown, the training data set of the network node device can be obtained by the following method. Figure 4

[0063] S401, based on the threshold screening condition, the historical ECN threshold configuration information of the network node device is screened to obtain the target historical ECN threshold configuration information that can take into account both mouse flow and elephant flow.

[0064] In the embodiments of the present application, the threshold configuration screening condition is used to screen the ECN threshold configuration information that can take into account both the time-sensitive mouse flow and the throughput-sensitive elephant flow. In some embodiments, the threshold configuration screening condition can specifically include the following conditions:

[0065] Condition 1: the sum of the PFC threshold values of all uplink port input direction virtual channels of the network node device is greater than the ECN threshold value of the downlink port output direction virtual channel of the network node device. That is:

[0066]

[0067] wherein, represents the PFC threshold value of all uplink port input direction virtual channels of the node device, represents the ECN threshold value of the corresponding virtual channel of the downlink port output direction of the node device.

[0068] ​By setting the condition, the screened ECN threshold configuration information can ensure that the network node device downlink port output direction virtual channel does not trigger the PFC flow control of all uplink port input direction virtual channels of the network node device before triggering the ECN congestion mark, reducing the number of PFC flow control triggers. And the screened ECN threshold configuration information can not trigger ECN frequently, resulting in a large number of CNP (Congestion Notification Packets) congestion notification packets and too low throughput.

[0069] Condition 2: The Service buffer threshold of the network node device downlink port output direction virtual channel is greater than the sum of the PFC threshold value of all uplink port input direction virtual channels of the network node device and the Headroom buffer threshold of the buffer space of all uplink port input direction virtual channels of the network node device. That is:

[0070]

[0071] Wherein represents the Service buffer threshold of the node device downlink port output direction virtual channel, represents the PFC threshold value of all uplink port input direction virtual channels of the node device, represents the Headroom buffer threshold of the buffer space of all uplink port input direction virtual channels of the node device.

[0072] By setting the condition, the screened ECN threshold configuration information can make the buffer space of the network node device downlink port output direction virtual channel sufficient before triggering the PFC flow control of all uplink port input direction virtual channels of the network node device, so that there is no packet loss.

[0073] Condition 3: The value of the ECN-PFC Buffer Gap between the ECN threshold and the PFC threshold of each virtual channel of the network node device is such that the traffic in the virtual channel does not trigger the PFC threshold within the effective time of the ECN threshold (that is, the time difference from marking ECN to source end speed reduction).

[0074] By setting the condition, the screened ECN threshold configuration information can make the traffic of the network node device not trigger the PFC threshold within the time difference from marking ECN to source end speed reduction, avoiding RDMA throughput reduction.

[0075] In a specific implementation, the historical ECN threshold configuration information of the network node device can be screened based on the threshold screening condition, and the historical ECN threshold configuration information that meets all the above conditions is determined as the target historical ECN threshold configuration information.

[0076] S402, the target historical ECN threshold configuration information is determined as sample ECN threshold configuration information, and the historical traffic state information sequence corresponding to the target historical ECN threshold configuration information is determined as the sample traffic state information sequence corresponding to the sample ECN threshold configuration information.

[0077] S403, the total data set of the network node device is obtained based on the sample ECN threshold configuration information and the sample traffic state information sequence corresponding to the sample ECN threshold configuration information.

[0078] In the embodiments of the present application, there are multiple sample ECN threshold configuration information that meet the threshold screening condition, and each sample ECN threshold configuration information and its corresponding sample traffic state information sequence is a sample, so that there are multiple samples based on the threshold screening condition, and the set composed of the multiple samples is the total data set of the network node device.

[0079] S404, the total data set is data cleaned, and the training data set of the node device is obtained based on the total data set after data cleaning.

[0080] In some embodiments of the present application, in order to ensure the diversity and standardization of the samples in the data set and facilitate the subsequent training of the model, the multiple samples can be further data cleaned by removing repeated, missing and abnormal data, and the total data set is established based on the multiple samples after data cleaning.

[0081] For example, the samples can be de-duplicated according to the unique identifier of the data in the sample, so as to de-duplicate the samples; the missing values in the samples can be filled by mean interpolation, so as to process the missing data in the samples; the standard deviation can be used to detect whether the data is abnormal according to the threshold, and the detected abnormal data is removed, so as to process the abnormal samples.

[0082] After the above processing, the total data set of the network node device after data cleaning can be obtained. Then, the training data set can be obtained based on the total data set. For example, 80% of the data in the total data set can be divided into the training data set, so as to obtain the training data set of the network node device. The training data set can include multiple samples, and each sample includes a sample traffic state information sequence and its corresponding sample ECN threshold configuration information.

[0083] In addition, the remaining 20% of the data in the total data set can be divided into a test data set to test the ECN threshold prediction model.

[0084] In S302, a loss value of the predicted ECN threshold configuration information compared with the sample ECN threshold configuration information of the current sample is calculated based on a target loss function.

[0085] In the embodiments of the present application, the target loss function can be the mean squared error (MSE). The predicted ECN threshold configuration information and the sample ECN threshold configuration information of the current sample can be substituted into the target loss function to calculate the loss value corresponding to the current sample.

[0086] In S303, the weights of the LSTM model are updated based on the loss value.

[0087] In the embodiments of the present application, the gradient descent method based on the neural network is used to find the optimal weight parameters of the LSTM model that minimize the target function. The gradient descent optimization algorithm can select the adam optimizer to improve the learning speed of the traditional gradient descent (optimizer='adam'). Each sample in the training data set can be taken as the current sample, and the above processes S301-S303 can be iteratively performed until the weights of the LSTM model minimize the target loss function.

[0088] In S304, the LSTM model whose weights minimize the target loss function is determined as the ECN threshold prediction model of the network node device.

[0089] In the embodiments of the present application, the LSTM model whose weights minimize the target loss function is the optimal weight model after the above iteration, and the optimal weight model can be verified based on the test data set. The model that passes the verification is determined as the ECN threshold configuration prediction model of the network node device.

[0090] Through the above processes, the ECN threshold prediction model of the network node device can be obtained. In specific implementation, the above processes can be performed before step S201 to obtain the ECN threshold prediction model of the network node device in advance.

[0091] The above describes in detail the generation method of the ECN threshold configuration information provided by the embodiments of the present application by taking one network node device as an example. In specific implementation, there can be multiple network node devices in the intelligent lossless network. For each network node device in the intelligent lossless network, the ECN threshold configuration information of each network node device can be generated by the generation method of the ECN threshold configuration information provided by the above embodiments. Since the principles are the same, they will not be described here.

[0092] Further, based on the same inventive concept, the application also provides a congestion control system, as shown in Figure 5 . Figure 5 is a structural schematic diagram of the congestion control system provided by some embodiments of the application. As shown in Figure 5 , in some embodiments of the application, the congestion control system 500 includes a network node device 501, a collector 502, and an ECN threshold configuration automatic generation component 503.

[0093] The network node device 501 is configured to periodically push the traffic state information of the network node device at the current time point to the collector.

[0094] The collector 502 is configured to combine the traffic state information of the network node device at the current time point with the traffic state information of the network node device at other time points within a specified length of time before the current time point to obtain a traffic state information sequence of the network node device at the current time point, and send the traffic state information sequence to the ECN threshold configuration automatic generation component.

[0095] The ECN threshold configuration automatic generation component 503 is configured to detect, based on a traffic model file of the network node device, whether the traffic state information sequence is an existing traffic model in the traffic model file, and when the traffic state information sequence is not an existing traffic model in the traffic model file, input the traffic state information sequence into an ECN threshold prediction model corresponding to the network node device to perform ECN threshold configuration information prediction, and determine the ECN threshold configuration information predicted by the ECN threshold prediction model as the current ECN threshold configuration information of the network node device and return it to the network node device. The traffic model file stores a plurality of existing traffic models, each traffic model being a traffic state information sequence. The ECN threshold prediction model corresponding to the network node device is trained by taking a sample traffic state information sequence of the network node device as input and taking sample ECN threshold configuration information corresponding to the sample traffic state information sequence as output label, and is obtained by training an LSTM model. The sample ECN threshold configuration information is ECN threshold configuration information that can take both mouse flow and elephant flow into account.

[0096] The network node device 504 is configured to configure the ECN threshold of the virtual channel of the network node device based on the current ECN threshold configuration information of the network node device, and perform congestion control on the virtual channel.

[0097] In some embodiments of the present application, the network node device 501 pushes the traffic state information of the network node device at the current time point to the collector 502 in a Telemetry push mode, so as to push the traffic state information of the current time point to the collector 502 in real time at a high speed.

[0098] In some embodiments of the present application, the ECN threshold configuration automatic generation component 503 is further configured to store the traffic state information sequence as a traffic model and the ECN threshold configuration information predicted by the ECN threshold prediction model as the ECN threshold configuration information corresponding to the traffic model in the traffic model file.

[0099] In some embodiments of the present application, the ECN threshold configuration automatic generation component 503 is further configured to:

[0100] When the traffic state information sequence is a traffic model already in the traffic model file, the ECN threshold configuration information corresponding to the traffic state information sequence in the traffic model file is determined as the current ECN threshold configuration information of the network node device and returned to the network node device.

[0101] The congestion control system provided by the embodiments of the present application takes a sample traffic state information sequence of a network node device as input and takes sample ECN threshold configuration information capable of considering both mouse flow and elephant flow corresponding to the sample traffic state information sequence as output label, trains an LSTM model to obtain an ECN threshold prediction model corresponding to the network node device. When receiving a current traffic state information sequence of a network node device, the traffic model file of the network node device is used to identify whether the traffic state information sequence is a traffic model already in the traffic model file; when the traffic state information sequence is not a traffic model already in the traffic model file, the traffic state information sequence is input into the ECN threshold prediction model corresponding to the network node device to perform ECN threshold configuration information prediction, and the ECN threshold configuration information predicted by the ECN threshold prediction model is determined as the current ECN threshold configuration information of the network node device. The ECN threshold prediction model of the network node device can dynamically generate suitable ECN threshold configuration information capable of considering both mouse flow and elephant flow for the network node device, so as to facilitate the network node device to consider both mouse flow and elephant flow based on the dynamic suitable ECN threshold configuration information and cope with burst traffic in an Incast scenario.

[0102] It should be noted that the congestion control system 500 provided by the embodiments of the present application and the generation method of the ECN threshold configuration information provided by the foregoing embodiments are based on the same inventive concept. For the specific processing method in the congestion control system 500, reference can be made to the corresponding steps in the foregoing embodiments, which will not be described herein again.

[0103] In addition, with the above Figure 1 Corresponding to the method for generating ECN threshold configuration information shown, this application also provides an apparatus for generating ECN threshold configuration information, such as... Figure 6 As shown. Figure 6 As shown, in some embodiments of this application, the ECN threshold configuration information generation device 600 includes:

[0104] The information receiving module 601 is used to receive the current traffic status information sequence of the network node device; wherein, the traffic status information sequence is a time sequence composed of the traffic status information of the network node device at each time point within a specified time length up to the current time point, and the traffic status information at each time point includes: the traffic status information of each virtual channel of the network node device at that time point.

[0105] The identification module 602 is used to identify whether the traffic status information sequence is an existing traffic model in the traffic model file based on the traffic model file of the network node device; wherein, the traffic model file stores multiple existing traffic models, and each traffic model is a traffic status information sequence.

[0106] The threshold prediction module 603 is used to, when the traffic state information sequence is not an existing traffic model in the traffic model file, input the traffic state information sequence into the ECN threshold prediction model corresponding to the network node device to predict ECN threshold configuration information, and determine the ECN threshold configuration information predicted by the ECN threshold prediction model as the current ECN threshold configuration information of the network node device; wherein, the ECN threshold configuration information includes the ECN threshold value of each virtual channel of the network node device; the ECN threshold prediction model corresponding to the network node device is obtained by training an LSTM model with the sample traffic state information sequence of the network node device as input and the sample ECN threshold configuration information corresponding to the sample traffic state information sequence as output label, and the sample ECN threshold configuration information is ECN threshold configuration information that can take into account both mouse flow and elephant flow.

[0107] In some embodiments of this application, the device 600 further includes:

[0108] The storage module is used to store the traffic status information sequence as a traffic model and the ECN threshold configuration information predicted by the ECN threshold prediction model as the ECN threshold configuration information corresponding to the traffic model in the traffic model file.

[0109] In some embodiments of this application, the device 600 further includes:

[0110] a threshold obtaining module, configured to, when the sequence of traffic state information is a traffic model that already exists in the traffic model file, determine, as current ECN threshold configuration information of the network node device, ECN threshold configuration information corresponding to the sequence of traffic state information in the traffic model file.

[0111] In some embodiments of the present application, the apparatus 600 further comprises a training module of an ECN threshold prediction model, configured to:

[0112] input a sequence of sample traffic state information of a current sample in a training data set of the network node device into the LSTM model to obtain predicted ECN threshold configuration information output by the LSTM model; the training data set comprises a plurality of samples, and each sample comprises a sequence of sample traffic state information of the network node device and sample ECN threshold configuration information corresponding to the sequence of sample traffic state information;

[0113] calculate, based on a target loss function, a loss value of the predicted ECN threshold configuration information compared with sample ECN threshold configuration information of the current sample;

[0114] update weights of the LSTM model based on the loss value;

[0115] iteratively perform the above steps until the weights of the LSTM model minimize the target loss function;

[0116] determine the LSTM model with the weights that minimize the target loss function as an ECN threshold prediction model of the network node device.

[0117] In some embodiments of the present application, the apparatus 600 further comprises a training data set obtaining module, configured to:

[0118] filter historical ECN threshold configuration information of the network node device based on a threshold filtering condition to obtain target historical ECN threshold configuration information that can take into account both rat flows and elephant flows;

[0119] determine the target historical ECN threshold configuration information as sample ECN threshold configuration information and determine historical traffic state information corresponding to the target historical ECN threshold configuration information as sample traffic state information corresponding to the sample ECN threshold configuration information;

[0120] obtain a total data set of the network node device based on the sample ECN threshold configuration information and the sample traffic state information corresponding to the sample ECN threshold configuration information;

[0121] Data cleaning is performed on the total data set, and a training data set of the node device is obtained based on the total data set after data cleaning.

[0122] In some embodiments of the present application, the threshold screening condition includes the following conditions:

[0123] Condition 1: The sum of the PFC threshold values of all input direction virtual channels of the network node device is greater than the ECN threshold value of the output direction virtual channel of the network node device;

[0124] Condition 2: The Service buffer threshold of the output direction virtual channel of the network node device is greater than the sum of the PFC threshold values of all input direction virtual channels of the network node device and the Headroom buffer threshold of the buffer space of all input direction virtual channels of the network node device;

[0125] Condition 3: The value of the ECN-PFC Buffer Gap between the ECN threshold and the PFC threshold of each virtual channel of the network node device is such that the traffic in the virtual channel does not trigger the PFC threshold within the effective time of the ECN threshold.

[0126] The ECN threshold configuration information generation device provided by the embodiments of the present application can serve as the execution subject of the ECN threshold configuration information generation method shown in Figure 2 , and can realize the functions of the ECN threshold configuration information generation method shown in Figure 2 . Since the principles are the same, they will not be described here.

[0127] In addition, based on the same inventive concept, the embodiments of the present application also provide an electronic device. Figure 7 is a structural schematic diagram of an electronic device 700 provided by some embodiments of the present application. As Figure 7 shown, the electronic device 700 includes a processor 701 and a memory 702.

[0128] In some embodiments, the memory 702 stores a computer program, and the processor 701 executes the computer program to realize:

[0129] receiving a current traffic state information sequence of a network node device; wherein the traffic state information sequence is a time sequence sequence composed of traffic state information of the network node device at each time point within a specified time length up to the current time point, and the traffic state information at each time point includes traffic state information of each virtual channel of the network node device at the time point;

[0130] identify whether the traffic state information sequence is an existing traffic model in a traffic model file of the network node device, wherein the traffic model file stores a plurality of existing traffic models, and each traffic model is a traffic state information sequence;

[0131] when the traffic state information sequence is not the existing traffic model in the traffic model file, input the traffic state information sequence into an ECN threshold prediction model corresponding to the network node device to perform ECN threshold configuration information prediction, and determine the ECN threshold configuration information predicted by the ECN threshold prediction model as the current ECN threshold configuration information of the network node device; wherein the ECN threshold configuration information includes an ECN threshold value of each virtual channel of the network node device; the ECN threshold prediction model corresponding to the network node device is trained by taking a sample traffic state information sequence of the network node device as input and taking sample ECN threshold configuration information corresponding to the sample traffic state information sequence as output label, and the sample ECN threshold configuration information is ECN threshold configuration information that can take into account both mouse flow and elephant flow.

[0132] In some embodiments of the present application, the memory 702 stores a computer program, and the processor 701 executes the computer program to further implement:

[0133] store the traffic state information sequence as a traffic model and the ECN threshold configuration information predicted by the ECN threshold prediction model as ECN threshold configuration information corresponding to the traffic model into the traffic model file.

[0134] In some embodiments of the present application, the memory 702 stores a computer program, and the processor 701 executes the computer program to further implement:

[0135] when the traffic state information sequence is the existing traffic model in the traffic model file, determine the ECN threshold configuration information corresponding to the traffic state information sequence in the traffic model file as the current ECN threshold configuration information of the network node device.

[0136] In some embodiments of the present application, the memory 702 stores a computer program, and the processor 701 executes the computer program to further implement:

[0137] inputting a sample traffic state information sequence of a current sample in a training data set of the network node device into an LSTM model to obtain predicted ECN threshold configuration information output by the LSTM model; the training data set contains a plurality of samples, and each sample includes a sample traffic state information sequence of the network node device and sample ECN threshold configuration information corresponding to the sample traffic state information sequence;

[0138] based on the target loss function, calculating a loss value of the predicted ECN threshold configuration information compared with the sample ECN threshold configuration information of the current sample;

[0139] updating weights of the LSTM model based on the loss value;

[0140] iteratively performing the above steps until the weights of the LSTM model minimize the target loss function;

[0141] determining the LSTM model with the weights minimizing the target loss function as an ECN threshold prediction model of the network node device.

[0142] In some embodiments of the present application, the memory 702 stores a computer program, and the processor 701 executes the computer program to further implement:

[0143] filtering historical ECN threshold configuration information of the network node device based on a threshold filtering condition to obtain target historical ECN threshold configuration information capable of considering both mouse flow and elephant flow;

[0144] determining the target historical ECN threshold configuration information as the sample ECN threshold configuration information and determining a historical traffic state information sequence corresponding to the target historical ECN threshold configuration information as a sample traffic state information sequence corresponding to the sample ECN threshold configuration information;

[0145] obtaining a total data set of the network node device based on the sample ECN threshold configuration information and the sample traffic state information sequence corresponding to the sample ECN threshold configuration information;

[0146] performing data cleaning on the total data set and obtaining a training data set of the network node device based on the total data set after data cleaning.

[0147] In some embodiments of the present application, the threshold filtering condition includes the following conditions:

[0148] Condition 1: a sum of PFC threshold values of all uplink port input direction virtual channels of the network node device is greater than an ECN threshold value of a downlink port output direction virtual channel of the network node device;

[0149] Condition 2: The Service buffer threshold of the output direction virtual channel of the network node device is greater than the sum of the PFC threshold value of all input direction virtual channels of the network node device and the Headroom buffer threshold of the buffer space of all input direction virtual channels of the network node device.

[0150] Condition 3: The value of the ECN-PFC Buffer Gap between the ECN threshold and the PFC threshold of each virtual channel of the network node device is such that the traffic in the virtual channel does not trigger the PFC threshold within the effective time of the ECN threshold.

[0151] In addition, based on the same inventive concept, one or more embodiments of the present application also provide a storage medium for storing a computer program. In a specific embodiment, the storage medium can be a U disk, an optical disk, a hard disk, etc. The computer program stored in the storage medium can achieve the following when executed by a processor:

[0152] receiving a current traffic state information sequence of a network node device; wherein the traffic state information sequence is a time sequence composed of traffic state information of the network node device at each time point within a specified time length ending at the current time point, and the traffic state information at each time point includes traffic state information of each virtual channel of the network node device at the time point;

[0153] based on a traffic model file of the network node device, identifying whether the traffic state information sequence is an existing traffic model in the traffic model file; wherein the traffic model file stores a plurality of existing traffic models, and each traffic model is a traffic state information sequence;

[0154] when the traffic state information sequence is not an existing traffic model in the traffic model file, inputting the traffic state information sequence into an ECN threshold prediction model corresponding to the network node device to perform ECN threshold configuration information prediction, and determining the ECN threshold configuration information predicted by the ECN threshold prediction model as the current ECN threshold configuration information of the network node device; wherein the ECN threshold configuration information includes the ECN threshold value of each virtual channel of the network node device; the ECN threshold prediction model corresponding to the network node device is trained by taking a sample traffic state information sequence of the network node device as input and taking a sample ECN threshold configuration information corresponding to the sample traffic state information sequence as output label, and the sample ECN threshold configuration information is an ECN threshold configuration information that can take both mouse flow and elephant flow into account.

[0155] In some embodiments of the present application, the computer program stored in the storage medium, when executed by the processor, can implement:

[0156] The traffic state information sequence is stored in the traffic model file as the traffic model, and the ECN threshold configuration information predicted by the ECN threshold prediction model is stored in the traffic model file as the ECN threshold configuration information corresponding to the traffic model.

[0157] In some embodiments of the present application, the computer program stored in the memory 702, and the processor 701 executes the computer program to further implement:

[0158] When the traffic state information sequence is a traffic model already in the traffic model file, the ECN threshold configuration information corresponding to the traffic state information sequence in the traffic model file is determined as the current ECN threshold configuration information of the network node device.

[0159] In some embodiments of the present application, the computer program stored in the storage medium, when executed by the processor, can implement:

[0160] The sample traffic state information sequence of the current sample in the training data set of the network node device is input into the LSTM model to obtain the predicted ECN threshold configuration information output by the LSTM model; the training data set includes a plurality of samples, and each sample includes a sample traffic state information sequence of the network node device and sample ECN threshold configuration information corresponding to the sample traffic state information sequence;

[0161] Based on the target loss function, the loss value of the predicted ECN threshold configuration information compared with the sample ECN threshold configuration information of the current sample is calculated;

[0162] The weights of the LSTM model are updated based on the loss value;

[0163] The above steps are iteratively performed until the weights of the LSTM model minimize the target loss function;

[0164] The LSTM model whose weights minimize the target loss function is determined as the ECN threshold prediction model of the network node device.

[0165] In some embodiments of the present application, the computer program stored in the storage medium, when executed by the processor, can implement:

[0166] The historical ECN threshold configuration information of the network node device is filtered based on the threshold filtering condition to obtain target historical ECN threshold configuration information that can take into account both mouse flow and elephant flow;

[0167] determine the target historical ECN threshold configuration information as sample ECN threshold configuration information, and determine a historical traffic state information sequence corresponding to the target historical ECN threshold configuration information as a sample traffic state information sequence corresponding to the sample ECN threshold configuration information;

[0168] obtain a total data set of the network node device based on the sample ECN threshold configuration information and the sample traffic state information sequence corresponding to the sample ECN threshold configuration information;

[0169] perform data cleaning on the total data set, and obtain a training data set of the node device based on the total data set after data cleaning.

[0170] In some embodiments of the present application, the threshold screening condition includes the following conditions:

[0171] Condition 1: the sum of PFC threshold values of all input direction virtual channels of the network node device is greater than the ECN threshold value of the output direction virtual channel of the network node device;

[0172] Condition 2: the Service buffer threshold of the output direction virtual channel of the network node device is greater than the sum of the PFC threshold value of all input direction virtual channels of the network node device and the Headroom buffer threshold of the buffer space of all input direction virtual channels of the network node device;

[0173] Condition 3: the value of the ECN-PFC Buffer Gap between the ECN threshold and the PFC threshold of each virtual channel of the network node device is such that the traffic in the virtual channel does not trigger the PFC threshold within the effective time of the ECN threshold.

[0174] The above describes specific embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in a different order than the order in which they are recited and still achieve the desired results. In addition, the processes depicted in the figures do not necessarily require the particular order shown, or sequential order, to achieve the desired results. In certain implementations, multitasking and parallel processing can be advantageous.

[0175] In summary, the above only describes the preferred embodiments of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0176] The systems, apparatuses, modules, or units illustrated in the above examples can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0177] Computer readable media includes permanent and non-permanent, removable and non-removable media, which can be implemented by any method or technology to store information. The information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible by a computing device. According to the definition herein, computer readable media does not include transitory media such as modulated data signals and carriers.

[0178] It should also be noted that the terms "comprising", "including", or any other variant thereof are intended to cover non-exclusive inclusion, so that processes, methods, articles or devices including a series of elements not only include those elements, but also include other elements not explicitly listed or inherent to such processes, methods, articles or devices. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or device including the element.

[0179] The various embodiments in the present application are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, the system embodiment is basically similar to the method embodiment, so the description is relatively simple, and the relevant parts can be referred to the part of the method embodiment.

Claims

1. A method for generating ECN threshold configuration information, characterized in that, The method includes: Receive the current traffic status information sequence of the network node device; wherein, the traffic status information sequence is a time sequence composed of the traffic status information of the network node device at each time point within a specified time length up to the current time point, and the traffic status information at each time point includes: the traffic status information of each virtual channel of the network node device at that time point; Based on the traffic model file of the network node device, it is identified whether the traffic state information sequence is an existing traffic model in the traffic model file; wherein, the traffic model file stores multiple existing traffic models, and each traffic model is a traffic state information sequence. When the traffic state information sequence is not an existing traffic model in the traffic model file, the traffic state information sequence is input into the ECN threshold prediction model corresponding to the network node device to predict ECN threshold configuration information. The ECN threshold configuration information predicted by the ECN threshold prediction model is determined as the current ECN threshold configuration information of the network node device. The ECN threshold configuration information includes the ECN threshold value for each virtual channel of the network node device. The ECN threshold prediction model corresponding to the network node device is trained using the sample traffic state information sequence of the network node device as input and the sample ECN threshold configuration information corresponding to the sample traffic state information sequence as the output label. The sample ECN threshold configuration information is ECN threshold configuration information that can accommodate both mouse flow and elephant flow. The ECN threshold prediction model is trained using the following method: The sample traffic state information sequence of the current sample in the training dataset of the network node device is input into the LSTM model to obtain the predicted ECN threshold configuration information output by the LSTM model; the training dataset contains multiple samples, each sample including: the sample traffic state information sequence of the network node device and the sample ECN threshold configuration information corresponding to the sample traffic state information sequence; Based on the target loss function, calculate the loss value of the predicted ECN threshold configuration information compared to the sample ECN threshold configuration information of the current sample; The weights of the LSTM model are updated based on the loss value; The iteration continues until the weights of the LSTM model minimize the target loss function; The LSTM model whose weights minimize the target loss function is determined as the ECN threshold prediction model for the network node device.

2. The method according to claim 1, characterized in that, The method further includes: The traffic state information sequence is used as the traffic model, and the ECN threshold configuration information predicted by the ECN threshold prediction model is used as the ECN threshold configuration information corresponding to the traffic model and stored in the traffic model file.

3. The method according to claim 1, characterized in that, The method further includes: When the traffic status information sequence is an existing traffic model in the traffic model file, the ECN threshold configuration information corresponding to the traffic status information sequence in the traffic model file is determined as the current ECN threshold configuration information of the network node device.

4. The method according to claim 1, characterized in that, The training dataset was obtained through the following method: The historical ECN threshold configuration information of the network node device is filtered based on threshold filtering conditions to obtain target historical ECN threshold configuration information that can take into account both mouse flow and elephant flow. The target historical ECN threshold configuration information is determined as the sample ECN threshold configuration information, and the historical traffic status information sequence corresponding to the target historical ECN threshold configuration information is determined as the sample traffic status information sequence corresponding to the sample ECN threshold configuration information. The total dataset of the network node device is obtained based on the sample ECN threshold configuration information and the sample traffic status information sequence corresponding to the sample ECN threshold configuration information; The total dataset is cleaned, and the training dataset for the node device is obtained based on the cleaned total dataset.

5. The method according to claim 4, characterized in that, The threshold filtering criteria include the following conditions: Condition 1: The sum of the PFC threshold values ​​of all uplink port input direction virtual channels of the network node device is greater than the ECN threshold value of the downlink port output direction virtual channel of the network node device. Condition 2: The Service cache threshold of the downlink port output virtual channel of the network node device is greater than the sum of the PFC threshold of all uplink port input virtual channels of the network node device and the Headroom cache threshold of the cache space of all uplink port input virtual channels of the network node device. Condition 3: The value of the buffer gap between the ECN threshold and the PFC threshold of each virtual channel of the network node device is such that the traffic in the virtual channel will not trigger the PFC threshold during the effective time of the ECN threshold.

6. A congestion control system, characterized in that, The system includes: network node devices, collectors, and an ECN threshold configuration automatic generation component; The network node device is used to periodically push the traffic status information of the network node device at the current time point to the collector; The collector is used to combine the traffic status information of the network node device at the current time point with the traffic status information of the network node device at other time points within a specified time length before the current time point to obtain the current traffic status information sequence of the network node device, and send the traffic status information sequence to the ECN threshold configuration automatic generation component. The ECN threshold configuration automatic generation component is used to detect whether the traffic state information sequence is an existing traffic model in the traffic model file of the network node device. If the traffic state information sequence is not an existing traffic model in the traffic model file, the traffic state information sequence is input into the ECN threshold prediction model corresponding to the network node device to predict ECN threshold configuration information. The ECN threshold configuration information predicted by the ECN threshold prediction model is determined as the current ECN threshold configuration information of the network node device and returned to the network node device. The traffic model file stores multiple existing traffic models, each of which is a traffic state information sequence. The ECN threshold prediction model corresponding to the network node device uses the sample traffic state information sequence of the network node device as input and the sample ECN threshold configuration information corresponding to the sample traffic state information sequence as output labels to train an LSTM model. The sample ECN threshold configuration information is designed to accommodate both mouse and elephant traffic flows. The ECN threshold prediction model is trained using the following method: the sample traffic state information sequence of the current sample in the training dataset of the network node device is input into an LSTM model to obtain the predicted ECN threshold configuration information output by the LSTM model. The training dataset contains multiple samples, each including: the sample traffic state information sequence of the network node device and the corresponding sample ECN threshold configuration information. Based on the objective loss function, the loss value of the predicted ECN threshold configuration information compared to the sample ECN threshold configuration information of the current sample is calculated. The weights of the LSTM model are updated based on the loss value. This process is iterated until the weights of the LSTM model minimize the objective loss function. The LSTM model whose weights minimize the objective loss function is determined as the ECN threshold prediction model for the network node device. The network node device is used to configure the ECN threshold of its virtual channel based on the current ECN threshold configuration information of the network node device, so as to perform congestion control on its virtual channel.

7. The system according to claim 6, characterized in that, Specifically, the network node device uses Telemetry push mode to push the traffic status information of the network node device at the current time point to the collector.

8. The system according to claim 6, characterized in that, The ECN threshold configuration automatic generation component is also used to store the traffic state information sequence as a traffic model and the ECN threshold configuration information predicted by the ECN threshold prediction model as the ECN threshold configuration information corresponding to the traffic model in the traffic model file.

9. The system according to claim 6, characterized in that, The ECN threshold configuration automatic generation component is also used for: When the traffic status information sequence is an existing traffic model in the traffic model file, the ECN threshold configuration information corresponding to the traffic status information sequence in the traffic model file is determined as the current ECN threshold configuration information of the network node device and returned to the network node device.

Images