A quality of service (QOS) optimization method, device and readable storage medium

Abstract

The application provides a quality of service (QOS) optimization method and device and a readable storage medium. Specifically, QOS parameter information of a link is acquired, and it is determined whether the QOS parameter information meets preset routing requirements. If it is monitored that the QOS parameter information does not meet the preset routing requirements, service data information of a link port is acquired. An optimal QOS routing strategy is determined according to the service data information. QOS optimization processing is performed according to the optimal QOS routing strategy. The QOS optimization method based on machine learning has the key point of data collection, classification and routing selection. The optimal routing is obtained by analyzing, training and calculating through a machine learning algorithm. The application scheme is suitable for QOS optimization of network data, effectively utilizes network resources, improves the performance of QOS, and realizes fast routing.

Description

Technical Field

[0001] This application relates to the field of wireless network technology, and in particular to a method, apparatus and readable storage medium for optimizing Quality of Service (QoS). Background Technology

[0002] QoS (Quality of Service) refers to a network's ability to provide better transmission services for a given data stream. It is a mechanism used to address network latency and congestion. Currently, in traditional networks, the Internet Engineering Task Force (IETF) has defined a series of QoS architectures to guarantee QoS, achieving QoS optimization through adjustments to routing policies.

[0003] For example, existing methods deploy machine learning algorithms in each router, which then makes routing decisions by acquiring relevant network traffic characteristics. However, in real-world networks, routers do not have sufficient computing resources to meet the requirements for machine learning to make routing decisions.

[0004] Therefore, how to achieve fast routing becomes a problem that needs to be solved. Summary of the Invention

[0005] The technical problem to be solved by this application is to provide a service quality (QoS) optimization method, apparatus and readable storage medium to address the above-mentioned shortcomings of the prior art.

[0006] Firstly, this application provides a method for optimizing Quality of Service (QOS), wherein the method...

[0007] The law includes:

[0008] S1. Obtain the QoS parameter information of the link and determine whether the QoS parameter information meets the preset routing requirements;

[0009] S2. If the QoS parameter information is detected to not meet the preset routing requirements, then obtain the service data information of the link port;

[0010] S3. Determine the optimal QoS routing strategy based on the aforementioned service data information;

[0011] S4. Perform QoS optimization processing according to the optimal QoS routing strategy.

[0012] In some embodiments, S3 includes:

[0013] S31. Perform data feature processing based on the business data information to obtain multiple feature data;

[0014] S32. Extract a predetermined number of feature data from the plurality of feature data as target features;

[0015] S33. Classify the target features using the CHS classification algorithm to obtain classification results, which include basic categories and subcategories;

[0016] S34. Determine the average latency of each link, and obtain the weight value of different paths of each data stream based on the average latency;

[0017] S35. Determine the optimal QoS routing strategy based on the classification results and the weight values ​​of different paths for each data stream.

[0018] In some embodiments, S35 includes:

[0019] Using the A3C algorithm, multiple agents are executed asynchronously using a multi-threaded approach, and online policy learning is performed based on the weight values ​​of different paths in each data stream, outputting dynamic link weight update values.

[0020] The optimal QoS routing strategy is determined using the A3CRA algorithm and the dynamic link weight update values.

[0021] In some embodiments, S4 includes:

[0022] Based on the current network status and the needs of different service types, and in conjunction with the optimal QoS routing strategy, the routing path of the target traffic is dynamically switched.

[0023] In some embodiments, QoS parameter information includes link latency, bandwidth, and packet loss rate.

[0024] In some embodiments, in S2, if the average measurement error is detected to be greater than a preset threshold, it is determined that the QoS parameter information does not meet the preset routing requirements. The average measurement error includes the error between the measured values ​​of link latency, bandwidth, and packet loss rate and the initial set values.

[0025] In some embodiments, the business data information includes the timestamp of the business data, the destination IP address, the source IP address, the protocol, and the packet length.

[0026] Secondly, this application provides a Quality of Service (QOS) optimization apparatus, the apparatus comprising:

[0027] The QoS parameter acquisition module is configured to acquire the QoS parameter information of the link and determine whether the QoS parameter information meets the preset routing requirements.

[0028] The service data acquisition module is configured to acquire the service data information of the link port if the QoS parameter information is detected to not meet the preset routing requirements.

[0029] The routing strategy determination module is configured to determine the optimal QoS routing strategy based on the service data information.

[0030] The QoS optimization processing module is configured to perform QoS optimization processing based on the optimal QoS routing policy.

[0031] Thirdly, this application provides a Quality of Service (QoS) optimization apparatus, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to implement the QoS optimization method described in the first aspect above.

[0032] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the Quality of Service (QoS) optimization method described in the first aspect.

[0033] This application provides a QoS optimization method, apparatus, and readable storage medium. Specifically, it acquires QoS parameter information of a link, determines whether the QoS parameter information meets preset routing requirements, and if the QoS parameter information does not meet the preset routing requirements, it acquires service data information of the link port, determines the optimal QoS routing strategy based on the service data information, and performs QoS optimization processing based on the optimal QoS routing strategy. This application provides a QoS optimization method that utilizes machine learning methods for imbalanced data classification and detailed classification, calculates a comprehensive weight value based on latency weights and machine learning methods, and connects data collection, classification, and final routing selection together. The key point of this machine learning-based QoS optimization method is that it analyzes, trains, and calculates the optimal route through machine learning algorithms from data collection to classification and routing selection. This application's solution is applicable to QoS optimization of network data, effectively utilizes network resources, improves QoS performance, and achieves fast routing. Attached Figure Description

[0034] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0035] Figure 1 A flowchart illustrating a Quality of Service (QOS) optimization method provided in this application embodiment;

[0036] Figure 2 This is a schematic diagram illustrating the CHS classification in an embodiment of this application.

[0037] Figure 3 This is a schematic diagram of the M / M / 1 queuing model in the embodiments of this application.

[0038] Figure 4 This is a schematic diagram of the A3C algorithm model construction in the embodiments of this application.

[0039] Figure 5 A schematic diagram of a Quality of Service (QOS) optimization device provided in this application embodiment;

[0040] Figure 6 This is a schematic diagram of a Quality of Service (QOS) optimization device provided in an embodiment of this application.

[0041] The accompanying drawings have illustrated specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to specific embodiments. Detailed Implementation

[0042] To enable those skilled in the art to better understand the technical solution of this application, the embodiments of this application will be further described in detail below with reference to the accompanying drawings.

[0043] It is understood that the specific embodiments and accompanying drawings described herein are merely for explaining this application and are not intended to limit this application.

[0044] It is understood that, without conflict, the various embodiments and features in the embodiments of this application can be combined with each other.

[0045] It is understood that, for ease of description, only the parts relevant to this application are shown in the accompanying drawings, while parts unrelated to this application are not shown in the drawings.

[0046] It is understood that each unit or module involved in the embodiments of this application may correspond to only one entity structure, or may be composed of multiple entity structures, or multiple units or modules may be integrated into one entity structure.

[0047] It is understood that the terms "first," "second," etc., used in the embodiments of this application are used to distinguish different objects or to distinguish different treatments of the same object, rather than to describe a specific order of objects.

[0048] It is understood that, without conflict, the functions and steps marked in the flowcharts and block diagrams of this application may occur in a different order than those marked in the accompanying drawings.

[0049] It is understood that the flowcharts and block diagrams of this application illustrate the possible architecture, functions, and operations of systems, apparatuses, devices, and methods according to various embodiments of this application. Each block in a flowchart or block diagram may represent a unit, module, program segment, or code, containing executable instructions for implementing the specified function. Furthermore, each block or combination of blocks in the block diagrams and flowcharts may be implemented using a hardware-based system to implement the specified function, or using a combination of hardware and computer instructions.

[0050] It is understood that the units and modules involved in the embodiments of this application can be implemented by software or by hardware. For example, the units and modules can be located in the processor.

[0051] Existing QoS optimization methods have the following drawbacks:

[0052] 1) Existing technologies are not perfect in classifying machine learning algorithms, and cannot classify traffic that is unbalanced over time and location more accurately.

[0053] 2) Existing technologies cannot detect the content of encrypted data packets in encrypted traffic, and therefore cannot perform traffic identification.

[0054] 3) In existing technologies, the measured delay is calculated by averaging multiple measurements, resulting in an estimated end-to-end delay. Furthermore, existing path calculation algorithms continuously search for the action with the maximum value of the action-value function in each iteration, failing to output the value function of each consecutive action. Therefore, they cannot handle continuous action spaces.

[0055] 4) Existing technologies only perform classification and routing calculations separately, without combining them for optimization.

[0056] This invention provides QoS guarantees for various network services by introducing discriminative feature information, enabling accurate classification of network traffic. It overcomes the limitation of the DPI method, which requires probing packet content, by combining it with machine learning methods for data classification. Path calculation uses latency as the base weight and employs machine learning to calculate the optimal route.

[0057] This invention presents a QoS optimization method based on machine learning. It collects, classifies, and selects routes from data, and then analyzes, trains, and calculates the optimal route through machine learning algorithms. This method effectively utilizes network resources, improves QoS performance, and achieves fast routing.

[0058] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0059] This application provides a method for optimizing Quality of Service (QoS). The workflow of this method can be implemented by electronic devices, such as computers and handheld smart terminals. For ease of explanation, the embodiments of this application will be described with the computer as the subject of the method execution.

[0060] Figure 1 A schematic diagram of the Quality of Service (QOS) optimization method provided in the embodiments of this application is shown below. Figure 1 As shown, this application provides a service quality (QOS) optimization method, which includes steps S1-S4, as follows:

[0061] S1. Obtain the QoS parameter information of the link and determine whether the QoS parameter information meets the preset routing requirements;

[0062] In some embodiments, QoS parameter information includes link latency, bandwidth, and packet loss rate. Specifically, the iperf command is used to generate different types of traffic, initial state values ​​are set for various performance parameters of the link, UDP packet traffic is generated in the OS3E topology network for testing, and the link's QoS parameter information is obtained.

[0063] S2. If the QoS parameter information is detected to not meet the preset routing requirements, then obtain the service data information of the link port;

[0064] In some embodiments, in S2, if the average measurement error is detected to be greater than a preset threshold, it is determined that the QoS parameter information does not meet the preset routing requirements. The average measurement error includes the error between the measured values ​​of link latency, bandwidth, and packet loss rate and the initial set values.

[0065] Specifically, the measured values ​​of link latency, bandwidth, and packet loss rate at a certain moment are compared with the initial settings to obtain the average measurement error. This allows us to determine whether the measured QoS parameters meet the routing requirements.

[0066] In some embodiments, the business data information includes the timestamp of the business data, the destination IP address, the source IP address, the protocol, and the packet length.

[0067] Specifically, by statistically analyzing the traffic flow rate, source address, and destination address, the system obtains the total number of data packets sent and received at each port, port flow rate, number of bytes received and forwarded, and other data layer network status information. If QoS cannot be guaranteed and service has deteriorated, a new path needs to be replanned. This involves recalculating a path that guarantees QoS requirements, ensuring dynamic switching and online routing functionality to maintain QoS for the target traffic.

[0068] Data collection was performed using Wireshark 18 on the specified experimental network. To ensure data reliability, traffic was collected at three time points: 7:00 AM, 12:00 PM, and 7:00 PM. All data streams were required to reach 300GB of traffic, with each stream lasting approximately 10 minutes. Specific data parameters are shown in Table 1 below.

[0069]

[0070] Table 1

[0071] The final collected raw data consists of network packets containing five-tuple information, including: timestamp, destination IP address, source IP address, protocol, and packet length. To ensure the consistency of the collected data packet information, only the transport layer information of the traffic is captured. Furthermore, only specific applications are opened during data collection; all other processes or programs are closed. After collection, IP subnetting and protocol filtering methods are used to remove unreliable data from the network flow, thereby ensuring the reliability and purity of the collected data as much as possible.

[0072] S3. Determine the optimal QoS routing strategy based on the aforementioned service data information;

[0073] In some embodiments, S3 includes:

[0074] S31. Perform data feature processing based on the business data information to obtain multiple feature data;

[0075] S32. Extract a predetermined number of feature data from the plurality of feature data as target features;

[0076] S33. Classify the target features using the CHS classification algorithm to obtain classification results, which include basic categories and subcategories;

[0077] S34. Determine the average latency of each link, and obtain the weight value of different paths of each data stream based on the average latency;

[0078] S35. Determine the optimal QoS routing strategy based on the classification results and the weight values ​​of different paths for each data stream.

[0079] Specifically, 40 network traffic features were used to characterize the collected data. Based on the direction of network flow, it can be divided into uplink, downlink, and data link, as shown in Table 2 below:

[0080]

[0081] Table 2

[0082] Optionally, after feature extraction, the extracted feature data can be normalized and discretized.

[0083] In some embodiments, different features are assigned a uniform number of discrete intervals to ensure that each feature has the same numerical range during classification. The complete improved Chi2 algorithm includes the following steps: data sorting; searching for label breakpoints; Chi2 merging; if the number of intervals is obtained, the final interval is output; otherwise, Chi2 merging is returned.

[0084] Discretized data is mostly integers, thus transforming the original multi-bit floating-point numbers into easily stored integers. Using C language's numeric storage methods as a benchmark, integer storage requires 16 bits, while double-precision floating-point numbers require 64 bits. This means that the storage after discretization is only one-quarter of the original storage. This significantly saves memory, especially for applications like network stream classification that generate large amounts of data per second.

[0085] In some embodiments, information gain is used to rank the 40 features, and the top 10 features are selected as candidate features. Specific information is shown in Table 3 below:

[0086]

[0087] Table 3

[0088] Then, a linear forward search algorithm was used to find the optimal feature subset for each base classifier. Furthermore, to verify the correlation between the proposed features and the QOS metric, the Pearson Correlation Coefficient (PCC) was used for correlation testing, as shown in the table below. The results show that the top 10 selected features are all strongly correlated with the QOS metric.

[0089] In this application, the CHS classification algorithm is an ensemble learning algorithm composed of a set of binary classifiers. The CHS algorithm includes two key concepts: classifier ranking and fine-grained classification. Furthermore, to ensure that each classifier has different features, different features are selected for each base classifier to maximize accuracy. Too many categories lead to an excessively long chain structure, resulting in severe error propagation. Therefore, shortening the classification structure can reduce the generated error. Thus, combining a chain structure with a hierarchical structure can achieve the above objectives while enabling fine-grained classification. For example, Figure 2 This is a schematic diagram illustrating the CHS classification in an embodiment of this application, as shown below. Figure 2 As shown, the traffic for ASD, ACD, and AUD is grouped into a single category: non-live video.

[0090] To more quickly determine the difficulty of class classification, the classifiers are ranked using aggregation degree. Aggregation degree is defined as the ratio of intra-class to inter-class distances. For class C... i The inter-class distance is S i For an N-gram classification problem:

[0091]

[0092] In the formula, T i Category C i The sample size, x j Traffic category C i A stream within. B i Category C i The center of aggregation is d Δ (•) is a distance or similarity function.

[0093] Inter-class distance D i The calculation is as follows:

[0094]

[0095] Then category C i Degree of aggregation R i for:

[0096]

[0097] Among them, R i The smaller the value, the larger the inter-class distance and the smaller the intra-class distance. In the formula, d... Δ (·) The degree of aggregation is calculated using Euclidean distance.

[0098] The above method is quick to classify and can intuitively reflect the difficulty of the classification categories.

[0099] In this application, during model training, the classifiers are trained sequentially. Furthermore, due to the unique structure of CHS, previously trained data needs to be removed before training the next classifier, meaning the training process becomes increasingly faster. For the hierarchical structure, the training method is the same as traditional methods, and data removal is not required. The process is as follows:

[0100] 1. Use a hierarchical structure to classify subcategories into basic categories based on natural relationships or prior knowledge;

[0101] 2. Sort the classifiers for the basic categories;

[0102] 3. for i = 1↑N do

[0103] 4. Select features for classifier C_i;

[0104] 5. Train the classifier C_i;

[0105] 6. If the classifier needs fine-grained classification, then

[0106] 7. Train the fine-grained classifier;

[0107] 8. Remove samples of category i;

[0108] 9. Use the trained model to classify network streams;

[0109] In this application, the weight value is calculated using average latency as the metric.

[0110] Specifically, the routing algorithm takes the end-to-end average latency as a design premise. Therefore, using end-to-end latency as an indicator for routing and flow control, the end-to-end network is first modeled using a queuing model, the end-to-end average latency is analyzed, and then the routing algorithm is designed. This application primarily considers queuing latency, employing the independence approximation theory to analyze the end-to-end average latency. The end-to-end network is decomposed into multiple links, and the data packet transmission on each link follows an M / M / 1 queuing model.

[0111] Figure 3 This is a schematic diagram of the M / M / 1 queuing model in the embodiments of this application, as shown below. Figure 3 As shown, the M / M / 1 queuing model consists of a queue and a server, corresponding to the router buffer and router node in the network. A node in the network may have multiple interfaces. Assume each interface corresponds to a queue and a server, and the transmission of data packets between interfaces follows the M / M / 1 queuing model. Assume the arrival interval of data packets follows a Poisson distribution with an arrival rate of λ; that is, an average of λ data packets arrive per second. The average interval between two data packets is T, which is determined by the following formula:

[0112]

[0113] Where λ is the arrival rate of data packets, which follows a Poisson distribution.

[0114] When a data packet is available on the server, subsequent data packets are queued in a queue. The service time for each data packet is S, which is determined by the following formula:

[0115]

[0116] in, B is the average length of the data packet, and B is the link bandwidth.

[0117] Without considering processing and propagation delays, data packets on link P ij Average latency It can be represented as:

[0118]

[0119] Among them, B i,j For link P i,j bandwidth, λ is the average length of the data packet. i,j For link P i,j The overall arrival rate of the data packets.

[0120] So, the data packet is at path P s,d End-to-end average delay It can be represented as:

[0121]

[0122] Among them, B k For path P s,d The bandwidth of the k-th link is: λ is the average length of the data packet. k Let be the total arrival rate of data packets on the k-th link.

[0123] To ensure that the latency of data packets is minimized, the algorithm can find a path with the minimum end-to-end average latency for each data stream according to the above formula, and derive the weight value of different paths for each data stream based on the latency.

[0124] In some embodiments, S35 includes: asynchronously executing multiple agents using a multi-threaded method via the A3C algorithm, and performing online policy learning based on the weight values ​​of different paths of each data stream to output dynamic link weight update values; and determining the optimal QoS routing policy using the dynamic link weight update values ​​via the A3CRA algorithm.

[0125] In this step, QoS route calculation mainly consists of two parts. First, the A3C algorithm, using a multi-threaded approach, asynchronously executes multiple agents, combining the latency weights from the previous step for online policy learning, and then outputs dynamic link weight update values. Then, the A3CRA algorithm uses the link weight values ​​calculated by the A3C algorithm to calculate the optimal path.

[0126] Specifically, Figure 4 This is a schematic diagram of the A3C algorithm model construction in the embodiments of this application, as shown below. Figure 4 As shown, the Actor-Critic algorithm consists of Actors and Critics, and is a reinforcement learning algorithm that combines value-centered and policy (action probability)-centered approaches. Actors iteratively make corresponding actions based on the policy, while Critics evaluate the quality of the actions based on a value function. The agent acquires actions through the Actor network. After executing the action, the environment provides the agent with a new state and a reward value R. The Critic network evaluates the reward value R and passes the result back to the Actor network for parameter updates. This entire process is a complete workflow of the agent exploring and training simultaneously.

[0127] In this context, the policy value function of the Critic part is the V of policy π. π (s), the action value function should also satisfy:

[0128] V π (s)=E π (R+γV π (s′))

[0129] Q π (s,a)=R+γV π (s′)

[0130] The Actor part uses an advantage function to represent the degree of goodness of the agent choosing action a in state s. Thus, the formula for the A advantage function is:

[0131] A π (s,a)=Q π (s,a)-V π (s)=R+γV π (s′)-V π (s)

[0132] The goodness of a strategy can be represented by a long-term discount reward, defined as follows:

[0133]

[0134] Calculating the above equation is very difficult; therefore, a simpler method is used to calculate the gradient. That is:

[0135]

[0136] To maximize the discount reward, the discount reward function can be negative, and the loss function can be defined as follows:

[0137] L(π)=-J(π)

[0138] The loss for each sample can only express the estimation bias of the agent in a certain state. In order to represent the total estimation bias, the average loss of each sample is taken, and L(π) is expanded into a summation form to obtain the following sample loss.

[0139]

[0140] The Critic component uses a value function to evaluate the quality of the agent's policy decisions, defining the loss function using mean squared error. That is:

[0141] L(w)=(Q π (s,a)-V π (s)) 2

[0142] Then accumulate the gradient of the Critic network

[0143]

[0144] The A3C algorithm still uses the Actor-Critic framework, and the A3CA Actor network still employs a policy-based learning method. The Actor algorithm mentions an advantage function, and the A3C algorithm uses N-step sampling to improve training efficiency. Therefore, the advantage function can be defined as:

[0145] A(s,a)=R+γR t+1 +…γ n-1 R t+n-1 +γ n V π (s′)-V π (s)

[0146] Compared to the Actor-Critic algorithm, the A3C algorithm adds an entropy term coefficient e to the policy network loss function. Therefore, the gradient update of the Actor in the A3C algorithm is as follows:

[0147]

[0148] The gradient update of the critic network part of the A3C algorithm adopts the same gradient update strategy as the critic part of the Actor-Critic algorithm. That is:

[0149]

[0150] S4. Perform QoS optimization processing according to the optimal QoS routing strategy.

[0151] In some embodiments, S4 includes: dynamically switching the routing path of the target traffic based on the current network status and the needs of different service type flows, in conjunction with the optimal QoS routing policy.

[0152] This application provides a QoS optimization method. It utilizes machine learning methods for imbalanced data classification and detailed classification, and calculates a comprehensive weight value based on latency weights and machine learning methods, connecting data collection, classification, and final route selection. The key to this machine learning-based QoS optimization method lies in analyzing, training, and calculating the optimal route through machine learning algorithms from data collection and classification to route selection. This solution is applicable to QoS optimization of network data, effectively utilizing network resources, improving QoS performance, and achieving fast routing.

[0153] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0154] Figure 5 A schematic diagram of the Quality of Service (QOS) optimization device provided in the embodiments of this application is shown below. Figure 5 As shown, this application provides a Quality of Service (QOS) optimization device, the device comprising:

[0155] The QoS parameter acquisition module 11 is configured to acquire the QoS parameter information of the link and determine whether the QoS parameter information meets the preset routing requirements.

[0156] The service data acquisition module 12 is configured to acquire the service data information of the link port if the QoS parameter information is detected to not meet the preset routing requirements.

[0157] The routing strategy determination module 13 is configured to determine the optimal QoS routing strategy based on the service data information.

[0158] The QoS optimization processing module 14 is configured to perform QoS optimization processing according to the optimal QoS routing policy.

[0159] Regarding the limitations on the QoS optimization device, please refer to the limitations on the QoS optimization method in the above embodiments of this application, which will not be repeated here.

[0160] Figure 6 Another schematic diagram of the QoS optimization device provided in the embodiments of this application is shown below. Figure 6 As shown, in some embodiments, this application provides a Quality of Service (QoS) optimization apparatus, including a memory 22 and a processor 21. The memory stores a computer program, and the processor is configured to run the computer program to execute the QoS optimization methods in the above embodiments of this application.

[0161] The memory is connected to the processor. The memory can be flash memory, read-only memory or other types of memory. The processor can be a central processing unit or a microcontroller.

[0162] In some embodiments, this application provides a Quality of Service (QoS) optimization apparatus, including a memory and a processor. The memory stores a computer program, and the processor is configured to run the computer program to execute the QoS optimization methods described in the above embodiments of this application.

[0163] The memory is connected to the processor. The memory can be flash memory, read-only memory or other types of memory. The processor can be a central processing unit or a microcontroller.

[0164] In some embodiments, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the Quality of Service (QoS) optimization method described in the above embodiments of this application.

[0165] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program modules, or other data). Computer-readable storage media include, but are not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory or other memory technologies, CD-ROM (Compact Disc Read-Only Memory), DVD or other optical disc storage, cartridges, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer.

[0166] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of this application, and this application is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and substance of this application, and these modifications and improvements are also considered to be within the scope of protection of this application.

Claims

1. A method for optimizing Quality of Service (QOS), characterized in that, The method includes: S1. Obtain the QoS parameter information of the link and determine whether the QoS parameter information meets the preset routing requirements; S2. If the QoS parameter information is detected to not meet the preset routing requirements, then obtain the service data information of the link port; S3. Determine the optimal QoS routing strategy based on the aforementioned service data information; S4. Perform QoS optimization processing according to the optimal QoS routing strategy; S3 includes: S31. Perform data feature processing based on the business data information to obtain multiple feature data; S32. Extract a predetermined number of feature data from the plurality of feature data as target features; S33. Classify the target features using the CHS classification algorithm to obtain classification results, which include basic categories and subcategories; S34. Determine the average latency of each link, and obtain the weight value of different paths of each data stream based on the average latency; S35. Determine the optimal QoS routing strategy based on the classification results and the weight values ​​of different paths of each data stream; S35 includes: Using the A3C algorithm, multiple agents are executed asynchronously using a multi-threaded approach, and online policy learning is performed based on the weight values ​​of different paths in each data stream, outputting dynamic link weight update values. The optimal QoS routing strategy is determined using the A3CRA algorithm and the dynamic link weight update values.

2. The service quality (QOS) optimization method according to claim 1, characterized in that, S4 includes: Based on the current network status and the needs of different service types, and in conjunction with the optimal QoS routing strategy, the routing path of the target traffic is dynamically switched.

3. The service quality (QOS) optimization method according to claim 1 or 2, characterized in that, QoS parameter information includes link latency, bandwidth, and packet loss rate.

4. The service quality (QOS) optimization method according to claim 1 or 2, characterized in that, In S2, if the average measurement error is detected to be greater than the preset threshold, it is determined that the QoS parameter information does not meet the preset routing requirements. The average measurement error includes the error between the measured values ​​of link delay, bandwidth, and packet loss rate and the initial set values.

5. The service quality (QOS) optimization method according to claim 1 or 2, characterized in that, Business data information includes the timestamp of the business data, destination IP address, source IP address, protocol, and packet length.

6. A service quality (QOS) optimization device, characterized in that, The device includes: The QoS parameter acquisition module is configured to acquire the QoS parameter information of the link and determine whether the QoS parameter information meets the preset routing requirements. The service data acquisition module is configured to acquire the service data information of the link port if the QoS parameter information is detected to not meet the preset routing requirements. The routing strategy determination module is configured to determine the optimal QoS routing strategy based on the service data information. The QoS optimization processing module is configured to perform QoS optimization processing based on the optimal QoS routing policy. The routing policy determination module is specifically configured as follows: Based on the business data information, data feature processing is performed to obtain multiple feature data; A predetermined number of feature data are extracted from the plurality of feature data as target features; The target features are classified using the CHS classification algorithm to obtain classification results, which include basic categories and subcategories. Determine the average latency of each link, and obtain the weight value of different paths for each data stream based on the average latency; Based on the classification results and the weight values ​​of different paths in each data stream, the optimal QoS routing strategy is determined, including: using the A3C algorithm, multiple agents are executed asynchronously using a multi-threaded method, and online policy learning is performed based on the weight values ​​of different paths in each data stream to output dynamic link weight update values; using the A3CRA algorithm, the optimal QoS routing strategy is determined using the dynamic link weight update values.

7. A service quality (QOS) optimization device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to implement the Quality of Service (QOS) optimization method as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the Quality of Service (QOS) optimization method as described in any one of claims 1-5.

Images