WO2019080474A1 - METHOD FOR DETECTING PASSIVE OPTICAL NETWORK BUSINESS FLOW AND SYSTEM AND INFORMATION MEDIUM - Google Patents

Abstract

Disclosed are a passive optical network business flow sensing method and system and a storage medium. The method comprises: by a master control layer, extracting business flow characteristic parameters according to an obtained training sample set, performing pre-processing on the business flow characteristic parameters, and determining a business flow characteristic set; performing Bayes training according to the characteristic set, updating parameter information of a Bayes classifier, and sending the parameter information to an agent layer; by the agent layer, periodically collecting business flow characteristic parameters and updating the training sample set, and establishing a corresponding Bayes classifier according to the parameter information; determining, according to the updated training sample set, the updated business flow characteristic set, and determining, according to the updated characteristic set and the Bayes classifier, a classification identification result; and performing business optimization adjustment according to the classification identification result.

Description

Method and system for detecting traffic flow of passive optical network, storage medium

Cross-reference to related applications

The present application is based on a Chinese patent application filed on Jan. 25, 2017, the entire disclosure of which is hereby incorporated by reference.

Technical field

The present invention relates to the field of service flow sensing technologies, and in particular, to a method and system for detecting traffic flow of a passive optical network, and a storage medium.

Background technique

Passive Optical Network (PON) is the main form of optical access network. The PON system is mainly composed of an OLT and an ONU, and is connected to a service node through an optical line terminal (OLT), and is connected to a user through an optical network unit (ONU).

As the services carried by the passive optical network become more complex, in order to obtain better quality of service (QoS) guarantees, relevant network behaviors are implemented for service identification and classification, and the premise of further improving end-to-end QoS of services is further improved. basis. When analyzing the performance of a service, it is often necessary to learn the traffic characteristics of a single service, and the statistical characteristics of the concurrent flows carried by the network, which are used to guide the formulation and implementation of the traffic engineering strategy, and the service flow sensing method has emerged. Traffic flow sensing is a higher-level traffic monitoring method. It classifies and identifies data packets according to different service flow definitions, and performs corresponding resource optimization scheduling to improve the optical access network's ability to support multiple services. In the service flow sensing technology of passive optical networks, the service classification and recognition algorithm based on traffic flow features plays an increasingly important role, and the adopted algorithm model directly determines the accuracy and efficiency of traffic flow perception.

The Bayes classification model is a neural network algorithm that can be used for classification and recognition. Bayesian classification is a classification model based on statistical methods. Bayes' theorem is the theoretical basis of Bayesian learning method. Based on the Bayesian theorem, the Bayesian classification model reduces the computational cost by the conditional independence hypothesis and predicts the class of unknown data samples belonging to the highest posterior probability. The above advantages make the Bayesian model have great application potential in business flow perception.

As the services carried by passive optical networks become more diverse and complex, fast and efficient traffic flow sensing algorithm models are particularly important. However, the existing Bayesian classification method is difficult to directly apply to devices in passive optical networks; the main reasons are as follows:

1) The PON system adopts the master-slave architecture of the OLT/ONU, that is, the ONU is controlled by a complex OLT to control multiple functions. The Bayesian classifier requires a large number of sample sets for training before traffic flow perception, and complex Bayesian training increases the complexity of the ONU.

2) Multiple ONUs independently perform Bayesian classification. In the absence of unified control, it is difficult to ensure the consistency of the PON system's service perception results.

The information disclosed in this Background section is only intended to increase the understanding of the general background of the application, and should not be construed as an admission or in any form.

Summary of the invention

The purpose of the present application is to provide a method and system for detecting traffic flow of a passive optical network, and a storage medium, thereby overcoming the defect that the existing passive optical network service flow has low perceived efficiency.

A method for sensing traffic flow of a passive optical network according to an embodiment of the present disclosure includes: the main control layer extracts a service flow characteristic parameter according to the acquired training sample set, where the service flow characteristic parameter includes: a data packet length, and a data packet arrival. The interval, the service duration, and the load degree of the ONU node; the main control layer determines the feature set of the service flow according to the service flow characteristic parameter; the main control layer performs Bayesian training according to the feature set, and updates the Bayesian classifier Parameter information, and sending the parameter information to the proxy layer; the proxy layer periodically collects new service flow feature parameters, and establishes a corresponding Bayesian classifier according to the parameter information; the proxy layer is based on the new service flow feature The parameter determines the updated feature set of the service flow, and determines the classification and recognition result according to the updated feature set and the Bayesian classifier; the agent layer performs service optimization adjustment according to the classification and recognition result.

In a possible implementation manner, the determining, by the main control layer, the feature set of the service flow according to the service flow feature parameter, performing normalization processing according to the service flow feature parameter, and determining a feature set of the service flow:

Where U(i) represents the feature set of service flow i; P _SIZE (i) is the packet length, P _INTERVAL (i) is the packet arrival interval, P _DUR (i) is the service duration, P _LOAD (i) The load level of the ONU node; P _{SIZE_MAX} is the maximum packet length, P _{INTERVAL_MAX} is the maximum arrival interval, P _{DUR_MAX} is the maximum service duration, and P _{LOAD_MAX} is the maximum load degree of the ONU node.

In a possible implementation manner, after the proxy layer periodically collects new service flow feature parameters, the method further includes: the proxy layer sends the new service flow feature parameter to the main control layer; and the main control layer is configured according to the new The traffic flow feature parameter updates the training sample set.

In a possible implementation, the determining, by the proxy layer, the updated feature set of the service flow according to the new service flow feature parameter includes: the proxy layer normalizing the new service flow feature parameter, and determining that the service flow is updated. Feature set.

Based on the same inventive concept, a passive optical network service flow sensing system according to an embodiment of the present disclosure includes: an optical line terminal and an optical network unit; and the optical line terminal is configured to extract a service flow feature according to the acquired training sample set. a parameter, the service flow characteristic parameter includes: a data packet length, a data packet arrival interval, a service duration, and a load degree of the ONU node; determining a feature set of the service flow according to the service flow feature parameter; and performing a feature according to the feature set The Yes training, updating the parameter information of the Bayesian classifier, and sending the parameter information to the proxy layer; the optical network unit is configured to periodically collect new service flow feature parameters, and establish a phase according to the parameter information. Corresponding Bayesian classifier; determining, according to the updated training sample set, a feature set after the service flow is updated, and determining a classification and recognition result according to the updated feature set and the Bayesian classifier; The classification and recognition results are used to optimize the business.

In a possible implementation, the optical line terminal is specifically configured to perform normalization processing according to the service flow characteristic parameter to determine a feature set of the service flow:

Where U(i) represents the feature set of service flow i; P _SIZE (i) is the packet length, P _INTERVAL (i) is the packet arrival interval, P _DUR (i) is the service duration, P _LOAD (i) The load level of the ONU node; P _{SIZE_MAX} is the maximum packet length, P _{INTERVAL_MAX} is the maximum arrival interval, P _{DUR_MAX} is the maximum service duration, and P _{LOAD_MAX} is the maximum load degree of the ONU node.

In a possible implementation manner, after the optical network unit periodically collects new service flow feature parameters, the optical network unit is further configured to: send a new service flow feature parameter to the optical line terminal;

The optical line terminal updates the training sample set according to the new service flow feature parameter.

In a possible implementation, the optical network unit is specifically configured to: normalize the new service flow feature parameters, and determine the updated feature set of the service flow.

The embodiment of the present invention provides a computer readable storage medium, where the computer readable storage medium stores a configuration program, and when the configuration program is executed by the processor, the passive light according to any one of the above embodiments is implemented. The steps of the method of network traffic flow perception.

The method and system for sensing traffic flow of a passive optical network provided by the embodiment of the present application divides the function of Bayesian classification and identification into a main control layer and a proxy layer; the main control layer is located at the OLT, and is mainly responsible for complex Bayesian training and The parameter information of the Bayesian classifier is determined; the main control layer performs unified Bayesian training from the global service flow of the PON system, and the formed Bayesian classifier has global consistency. Using the layered Bayesian model, on the one hand, the Bayes proxy layer can reduce the complexity of the ONU. On the other hand, the Bayes master module at the OLT controls the Bayes proxy module in all ONUs in a unified manner, ensuring the PON pair service from a global perspective. The flow performs perceptual computational accuracy and consistency. At the same time, the service load degree of the ONU is taken as one of the characteristic parameters, and the impact of the ONU node load on the service flow perception is fully considered, so that the sensing result is more accurate. Feedback Bayesian classifier update mode: The main control layer can update the Bayesian classifier according to the actual operating conditions perceived by the traffic flow classification, and further improve the accuracy of the operation.

Other features and advantages of the present application will be set forth in the description which follows. The objectives and other advantages of the present invention can be realized and obtained by the structure of the invention.

DRAWINGS

The accompanying drawings are used to provide a further understanding of the invention, In the drawing:

1 is a flowchart of a method for sensing traffic flow of a passive optical network according to an embodiment of the present application;

2 is a schematic structural diagram of a layered Bayesian model in an embodiment of the present application;

FIG. 3 is a structural diagram of a system for detecting traffic flow of a passive optical network according to an embodiment of the present application.

Detailed ways

The specific embodiments of the present application are described in detail below with reference to the accompanying drawings, but it should be understood that the scope of the present invention is not limited by the specific embodiments.

The technical solutions in the embodiments of the present application are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present application. It is a part of the embodiments of the present application, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present application without departing from the inventive scope are the scope of the present application. The term "comprising" or variations such as "comprises" or "comprises", etc., are to be understood to include the recited elements or components, and Other components or other components are not excluded.

The word "exemplary" is used exclusively herein to mean "serving as an example, embodiment, or illustrative." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or preferred.

In addition, numerous specific details are set forth in the Detailed Description of the <RTIgt; Those skilled in the art will appreciate that the present application may be practiced without some specific details. In some instances, methods, means, and components that are well-known to those skilled in the art are not described in detail in order to facilitate the disclosure.

According to an embodiment of the present application, a method for traffic flow sensing of a passive optical network is provided, which designs a Bayesian classification model as a two-layer architecture: a master layer and a proxy layer. The main control layer is only responsible for the training and updating of the Bayesian classification model, and distributes the well-trained Bayesian classification model parameters to the agent layer for unified configuration; under the control of the main control layer, the agent layer only has direct business flow. Function of Classification Identification FIG. 1 is a flowchart of the method, and specifically includes steps 101-106:

Step 101: The main control layer extracts a service flow characteristic parameter according to the obtained training sample set, where the service flow characteristic parameter includes: a data packet length, a data packet arrival interval, a service duration, and a load degree of the ONU node.

For the master-slave architecture of the PON system, the layered Bayesian model is used in the embodiment of the present application to complete the service flow perception. The layered Bayesian model is divided into a main control layer and a proxy layer. The main control layer can be set at the optical line terminal (OLT), the proxy layer is set at the optical network unit (ONU), and each ONU is provided with a separate The agent module, the structure diagram of the layered Bayesian model is shown in Figure 2.

Specifically, the main control layer is composed of a Bayesian main control module and a training module. The training module obtains the training sample set, and the main control module performs unified Bayesian training on the input training sample set until a complete Bayesian classifier is formed, thereby determining the parameters of the Bayesian classifier that need to be sent to the agent layer. Information; the Bayesian training method used is the same as the existing Bayesian training method. After forming the Bayesian classifier, the Bayes master module distributes the parameter information of the Bayesian classifier to all Bayesian agent modules of the agent layer to ensure the consistency of each agent module in the service flow classification perception. The agent layer consists of a number of Bayesian agent modules. Each agent module obtains the same Bayesian classifier parameter information from the main control layer, and then can use the same classifier to perform service flow classification sensing.

The Bayesian classification principle is briefly introduced as follows: with attribute variables X1, X2, ..., Xn, C is a class variable, and D is a sample set. According to the Bayesian formula, the formula (1) can be known.

最大的C值即为属性值集合{X'1,…X'n}所属的类。 Obtain an estimate of P(C), P(X1|C),..., P(Xn|C) through training set D for a given set of attribute values {X'1,...X'n} such that

The largest C value is the class to which the attribute value set {X'1,...X'n} belongs.

In the embodiment of the present application, in the traffic flow perception based on the Bayesian model, the traffic flow feature is input information for performing Bayesian classification sensing. Therefore, the service flow characteristic parameters need to be extracted first, and the service flow characteristic parameters mainly include: data packet length, data packet arrival interval, service duration, and load degree of the ONU node. In order to fully consider the impact of the traffic load degree of the ONU node on the service flow, the load level is one of the characteristic parameters, and the load degree may be a ratio of the service data volume to the total capacity.

Step 102: The main control layer determines a feature set of the service flow according to the service flow characteristic parameter.

In the embodiment of the present application, before the Bayesian training is performed by using the service flow feature parameter, the feature parameter is pre-processed, and the feature set of the service flow is determined according to the service flow feature parameter. Specifically, the pre-processing process is a normalization process to avoid over-fitting. Specifically, the feature set U(i) of the service flow can be calculated by using the following formula (2):

Where U(i) represents the feature set of service flow i; P _SIZE (i) is the packet length, P _INTERVAL (i) is the packet arrival interval, P _DUR (i) is the service duration, P _LOAD (i) The load level of the ONU node; P _{SIZE_MAX} is the maximum packet length, P _{INTERVAL_MAX} is the maximum arrival interval, P _{DUR_MAX} is the maximum service duration, and P _{LOAD_MAX} is the maximum load degree of the ONU node.

Step 103: The main control layer performs Bayesian training according to the feature set, updates the parameter information of the Bayesian classifier, and sends the parameter information to the proxy layer.

In the embodiment of the present application, the Bayesian training is performed by the main control layer, the parameter information of the Bayesian classifier is determined, and the parameter information is sent to the proxy layer in the form of a broadcast, so that all the Bayes in the proxy layer can be guaranteed. The proxy module (ie ONU) samples the same Bayesian classifier to maintain consistency of business-aware results from a global perspective.

Step 104: The proxy layer periodically collects new service flow feature parameters, and establishes a corresponding Bayesian classifier according to the parameter information.

In the embodiment of the present application, the node in the proxy layer (ie, the ONU node) can collect the records that are perceived by the service flow classification, and extract the characteristic parameters of the received bidirectional data stream for each access service flow, that is, the feature parameters can be collected. A new service flow characteristic parameter, the service flow characteristic parameter is used to input the value Bayesian classifier; at the same time, the Bayesian classifier is re-determined according to the parameter information of the Bayesian classifier delivered by the main control layer (ie, the OLT) .

Step 105: The proxy layer determines the updated feature set of the service flow according to the new service flow characteristic parameter, and determines the classification and recognition result according to the updated feature set and the Bayesian classifier.

Step 106: The agent layer performs business optimization adjustment according to the classification identification result.

In the embodiment of the present application, the proxy layer function is implemented by a "Bayes proxy module" running in the ONU device. According to the Bayesian classifier information parameters broadcast by the OLT, the Bayesian classifier is implemented and fixed in hardware by configuring the FPGA. The Bayes proxy module is implemented in hardware to improve the running speed and improve the real-time traffic flow perception of the PON system. Sex. Under the condition that the Bayesian classifiers are consistent, the "Bayes proxy modules" in each ONU work independently to perform traffic flow perception.

Specifically, the "Bayes proxy module" extracts each new service flow feature parameter and performs normalization processing to obtain the updated feature set of the service flow, and the normalization method can also adopt the formula (2). The feature parameters are normalized according to formula (2) to avoid overfitting, thereby obtaining a feature set U(i) describing the traffic flow. The updated feature set is input into the Bayesian classifier to obtain the classification result of the service flow, that is, the priority of the service flow; then the ONU adjusts the service data packet queue in the buffer area according to the classification identification result, and performs service optimization. Scheduling (for example, prioritizing or prioritizing bandwidth allocation, etc.).

The passive optical network service flow sensing method provided by the embodiment of the present application divides the Bayesian classification and recognition function into a main control layer and a proxy layer; the main control layer is located at the OLT, and is mainly responsible for complex Bayesian training and determining the shell The parameter information of the Yesi classifier; the Bayesian master layer performs unified Bayesian training from the global traffic flow of the PON system, and the formed Bayesian classifier has global consistency. Using the layered Bayesian model, on the one hand, the Bayes proxy layer can reduce the complexity of the ONU. On the other hand, the Bayes master module at the OLT controls the Bayes proxy module in all ONUs in a unified manner, ensuring the PON pair service from a global perspective. The flow performs perceptual computational accuracy and consistency. At the same time, the service load degree of the ONU is taken as one of the characteristic parameters, and the impact of the ONU node load on the service flow perception is fully considered, so that the sensing result is more accurate.

On the basis of the foregoing embodiment, after the step proxy layer periodically collects new service flow feature parameters, the method further includes: the proxy layer sends the new service flow feature parameter to the main control layer; and the main control layer is based on the new service. The flow feature parameter updates the training sample set.

In the embodiment of the present application, a feedback Bayesian classifier training update mode is adopted: the new service flow feature parameter and the corresponding classification recognition result are periodically fed back to the main control layer, so that the main control layer can continuously update the training. Sample set. The Bayes main control module of the main control layer can periodically perform Bayesian training according to the new training sample set to form a new Bayesian classifier; meanwhile, the Bayes proxy module in the ONU is updated under the control of the main control module. Bayesian classifier. Feedback Bayesian classifier update mode: The main control layer can update the Bayesian classifier according to the actual operating conditions perceived by the traffic flow classification, and further improve the accuracy of the operation.

The method flow of the passive optical network service flow sensing is described in detail above. The method can also be implemented by the corresponding system. The structure and function of the system are described in detail below.

A passive optical network service flow sensing system is provided in the embodiment of the present application. Referring to FIG. 3, the system includes: an OLT and an ONU.

The OLT is configured to extract a service flow characteristic parameter according to the obtained training sample set, where the service flow characteristic parameter includes: a data packet length, a data packet arrival interval, a service duration, and a load degree of the ONU node; determining a service flow according to the service flow characteristic parameter. The feature set; Bayesian training according to the feature set, updating the parameter information of the Bayesian classifier, and sending the parameter information to the proxy layer.

The ONU is configured to periodically collect new service flow feature parameters, and establish a corresponding Bayesian classifier according to the parameter information; determine the updated feature set of the service flow according to the updated training sample set, and according to the updated feature set And the Bayesian classifier determines the classification recognition result; performs business optimization adjustment according to the classification recognition result.

In a possible implementation manner, the optical line terminal is specifically configured to perform normalization according to the service flow characteristic parameter:

Where U(i) represents the feature set of service flow i; P _SIZE (i) is the packet length, P _INTERVAL (i) is the packet arrival interval, P _DUR (i) is the service duration, P _LOAD (i) The load level of the ONU node; P _{SIZE_MAX} is the maximum packet length, P _{INTERVAL_MAX} is the maximum arrival interval, P _{DUR_MAX} is the maximum service duration, and P _{LOAD_MAX} is the maximum load degree of the ONU node.

In a possible implementation, after the optical network unit periodically collects new service flow feature parameters, the optical network unit is further configured to: send a new service flow feature parameter to the optical line terminal; The new traffic flow feature parameter updates the training sample set.

In a possible implementation, the optical network unit is specifically configured to perform normalization processing on the new service flow feature parameters to determine the updated feature set of the service flow.

The embodiment of the present invention provides a computer readable storage medium, where the computer readable storage medium stores a configuration program, and when the configuration program is executed by the processor, the passive light according to any one of the above embodiments is implemented. The steps of the method of network traffic flow perception.

The passive optical network service flow sensing system provided by the embodiment of the present application divides the Bayesian classification and recognition function into a main control layer and a proxy layer; the main control layer is located at the OLT, and is mainly responsible for complex Bayesian training and determining the shell. The parameter information of the Less classifier; the main control layer performs unified Bayesian training from the global traffic flow of the PON system, and the formed Bayesian classifier has global consistency. Using the layered Bayesian model, on the one hand, the Bayes proxy layer can reduce the complexity of the ONU. On the other hand, the Bayes master module at the OLT controls the Bayes proxy module in all ONUs in a unified manner, ensuring the PON pair service from a global perspective. The flow performs perceptual computational accuracy and consistency. At the same time, the service load degree of the ONU is taken as one of the characteristic parameters, and the impact of the ONU node load on the service flow perception is fully considered, so that the sensing result is more accurate. Feedback Bayesian classifier update mode: The main control layer can update the Bayesian classifier according to the actual operating conditions perceived by the traffic flow classification, and further improve the accuracy of the operation.

The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, ie may be located A place, or it can be distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the embodiment. Those of ordinary skill in the art can understand and implement without deliberate labor.

Through the description of the above embodiments, those skilled in the art can clearly understand that the various embodiments can be implemented by means of software plus a necessary general hardware platform, and of course, by hardware. Based on such understanding, the above-described technical solutions may be embodied in the form of software products in essence or in the form of software products, which may be stored in a computer readable storage medium such as ROM/RAM, magnetic Discs, optical discs, etc., include instructions for causing a computer device (which may be a personal computer, server, or network device, etc.) to perform the methods described in various embodiments or portions of the embodiments.

The foregoing description of the specific exemplary embodiments of the present application is for purposes of illustration and illustration. The description is not intended to limit the invention to the precise forms disclosed. The embodiments were chosen and described in order to explain the particular embodiments of the invention, Choose and change. The scope of the application is intended to be defined by the claims and their equivalents.

Industrial applicability

According to the embodiment of the present application, the function of Bayesian classification and recognition is divided into a main control layer and a proxy layer; the main control layer is located at the OLT, and is mainly responsible for complex Bayesian training and determining parameter information of the Bayesian classifier; The layer performs unified Bayesian training from the global traffic flow of the PON system, and the formed Bayesian classifier has global consistency.

Using the layered Bayesian model, on the one hand, the Bayes proxy layer can reduce the complexity of the ONU. On the other hand, the Bayes master module at the OLT controls the Bayes proxy module in all ONUs in a unified manner, ensuring the PON pair service from a global perspective. The flow performs perceptual computational accuracy and consistency. At the same time, the service load degree of the ONU is taken as one of the characteristic parameters, and the impact of the ONU node load on the service flow perception is fully considered, so that the sensing result is more accurate. Feedback Bayesian classifier update mode: The main control layer can update the Bayesian classifier according to the actual operating conditions perceived by the traffic flow classification, and further improve the accuracy of the operation.

Claims (9)

A method for traffic flow sensing of a passive optical network, comprising: The main control layer extracts service flow characteristic parameters according to the obtained training sample set, where the service flow characteristic parameters include: a data packet length, a data packet arrival interval, a service duration, and a load degree of the ONU node; The main control layer determines a feature set of the service flow according to the service flow characteristic parameter; The main control layer performs Bayesian training according to the feature set, updates parameter information of the Bayesian classifier, and sends the parameter information to the proxy layer; The proxy layer periodically collects new service flow feature parameters, and establishes a corresponding Bayesian classifier according to the parameter information; The agent layer determines the updated feature set of the service flow according to the new service flow characteristic parameter, and determines the classification and recognition result according to the updated feature set and the Bayesian classifier; The agent layer performs business optimization adjustment according to the classification and recognition result. The method according to claim 1, wherein the determining, by the main control layer, the feature set of the service flow according to the service flow characteristic parameter comprises: Performing normalization processing according to the service flow characteristic parameter to determine a feature set of the service flow: Where U(i) represents the feature set of service flow i; P _SIZE (i) is the packet length, P _INTERVAL (i) is the packet arrival interval, P _DUR (i) is the service duration, P _LOAD (i) The load level of the ONU node; P _{SIZE_MAX} is the maximum packet length, P _{INTERVAL_MAX} is the maximum arrival interval, P _{DUR_MAX} is the maximum service duration, and P _{LOAD_MAX} is the maximum load degree of the ONU node. The method of claim 1, wherein after the proxy layer periodically collects new service flow feature parameters, the method further includes: The proxy layer sends the new service flow feature parameters to the main control layer; The main control layer updates the training sample set according to the new service flow characteristic parameter. The method according to claim 1, wherein the determining, by the proxy layer, the updated feature set of the service flow according to the new service flow characteristic parameter comprises: The agent layer normalizes the new service flow feature parameters to determine the feature set after the service flow is updated. A passive optical network service flow sensing system, comprising: an optical line terminal and an optical network unit; The optical line terminal is configured to extract a service flow feature parameter according to the acquired training sample set, where the service flow feature parameter includes: a data packet length, a data packet arrival interval, a service duration, and a load degree of the ONU node; The stream feature parameter determines a feature set of the service flow; performs Bayesian training according to the feature set, updates parameter information of the Bayesian classifier, and sends the parameter information to the proxy layer; The optical network unit is configured to periodically collect new service flow feature parameters, and establish a corresponding Bayesian classifier according to the parameter information; and determine a service flow updated feature set according to the updated training sample set. And determining, according to the updated feature set and the Bayesian classifier, a classification recognition result; performing service optimization adjustment according to the classification recognition result. The system according to claim 5, wherein the optical line terminal is specifically configured to: perform normalization processing according to the service flow characteristic parameter to determine a feature set of the service flow:

Where U(i) represents the feature set of service flow i; P _SIZE (i) is the packet length, P _INTERVAL (i) is the packet arrival interval, P _DUR (i) is the service duration, P _LOAD (i) The load level of the ONU node; P _{SIZE_MAX} is the maximum packet length, P _{INTERVAL_MAX} is the maximum arrival interval, P _{DUR_MAX} is the maximum service duration, and P _{LOAD_MAX} is the maximum load degree of the ONU node. The system of claim 5, wherein the optical network unit is further configured to: after periodically collecting new service flow feature parameters: Sending a new service flow characteristic parameter to the optical line terminal; The optical line terminal updates the training sample set according to the new service flow feature parameter. The system of claim 5, wherein the optical network unit is specifically configured to: The new service flow feature parameters are normalized to determine the updated feature set of the service flow. A computer readable storage medium having stored thereon a configuration program, the configuration program being executed by a processor to implement the passive optical network traffic flow sensing according to any one of claims 1 to 4. The steps of the method.

Images