Traffic data rollback method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of network models, in particular to a flow data rollback method and device, computer equipment and a storage medium. The method comprises the steps of obtaining first performance data in an operation process of a target network model in a network model updating process; determining a data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data under the condition that the first performance data does not meet the model operation condition; and according to the data rollback strategy, rolling back part of traffic data or all traffic data processed by the target network model to the historical network model. According to the method and the device, the flexibility of rolling back the traffic data to the historical network model is improved, and the proportion of inputting the traffic data to the target network model and the historical network model in the rolling-back strategy is dynamically adjusted according to the actual condition of the target network model, so that the accurate and flexible adjustment of different abnormal conditions of the target network model is realized, and the user experience is improved. And continuous and stable operation of the target network model is ensured.

Description

Technical Field

[0001] The present application relates to the field of network model technology, and in particular to a method, device, computer equipment and storage medium for rolling back traffic data. Background Art

[0002] In modern complex network environments, in order to optimize computing resources, network model training is usually placed on the server in the data center, and the server pushes the trained new model to the network device to achieve the update of the network model, that is, the update from the old model to the new model.

[0003] In order to prevent the updated new model from having anomalies during operation and thus affecting network services, a traffic data rollback mechanism is currently set up. The rollback mechanism can roll back the traffic data that the new model needs to process to the old model when anomalies occur in the new model. However, the existing technology has low flexibility in rolling back all traffic data processed by the new model to the old model. Summary of the invention

[0004] Based on this, it is necessary to provide a traffic data rollback method, device, computer equipment and storage medium that can improve the flexibility of rolling back traffic data to an old model in order to address the above technical problems.

[0005] In a first aspect, the present application provides a method for rolling back traffic data. The method comprises:

[0006] During the network model updating process, obtaining first performance data during the operation of the target network model;

[0007] In a case where the first performance data does not meet the model operation condition, determining a data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data;

[0008] According to the data rollback strategy, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0009] In one embodiment, rolling back part of the traffic data or all of the traffic data processed by the target network model to the historical network model according to the data rollback strategy includes:

[0010] According to the data rollback strategy, adjusting the actual allocation coefficient corresponding to the traffic data processed by the target network model to obtain a target allocation coefficient;

[0011] According to the target allocation coefficient, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0012] In one embodiment, when the performance level is a mild abnormal level, the target allocation coefficient includes a plurality of first allocation coefficients that decrease in sequence, and each of the first allocation coefficients is greater than 0 and less than the actual allocation coefficient, and rolling back part of the traffic data or all of the traffic data processed by the target network model to the historical network model according to the target allocation coefficient includes:

[0013] According to the gradient order of the first allocation coefficients from large to small, a portion of the traffic data processed by the target network model is gradient rolled back to the historical network model.

[0014] In one embodiment, the method further comprises:

[0015] During the process of gradient rolling back part of the traffic data processed by the target network model to the historical network model, if it is detected that the second performance data during the operation of the target network model meets the model operation conditions, it is prohibited to continue to gradient roll back part of the traffic data processed by the target network model to the historical network model.

[0016] In one embodiment, the data rollback strategy includes first coefficient adjustment values ​​corresponding to each of the first allocation coefficients, and adjusting the actual allocation coefficient corresponding to the traffic data processed by the target network model according to the data rollback strategy to obtain the target allocation coefficient includes:

[0017] The actual allocation coefficient is respectively subjected to difference calculation with each of the first coefficient adjustment values ​​to obtain the plurality of first allocation coefficients.

[0018] In one embodiment, when the performance level is a moderate abnormal level, the target allocation coefficient is greater than 0 and less than the actual allocation coefficient, and rolling back part of the traffic data or all of the traffic data processed by the target network model to the historical network model according to the target allocation coefficient includes:

[0019] According to the target allocation coefficient, part of the traffic data processed by the target network model is rolled back to the historical network model.

[0020] In one embodiment, when the performance level is a severe abnormal level, the target allocation coefficient is equal to 0, and rolling back part of the traffic data or all of the traffic data processed by the target network model to the historical network model according to the target allocation coefficient includes:

[0021] According to the target allocation coefficient, all traffic data processed by the target network model are rolled back to the historical network model.

[0022] In one embodiment, the data rollback strategy includes a second coefficient adjustment value, and adjusting the actual allocation coefficient corresponding to the traffic data processed by the target network model according to the data rollback strategy to obtain the target allocation coefficient includes:

[0023] The actual allocation coefficient is subjected to a difference operation with the second coefficient adjustment value to obtain the target allocation coefficient.

[0024] In one embodiment, determining the data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data includes:

[0025] Determine a quantitative value of a data difference between the first performance data and a preset performance threshold;

[0026] Determining the performance level corresponding to the first performance data according to the data difference quantization value;

[0027] The data rollback strategy corresponding to the target network model is determined according to the candidate level corresponding to each candidate rollback strategy and the performance level.

[0028] In one embodiment, the method further comprises:

[0029] When the first performance data meets the model operation condition, all traffic data of the historical network model processor is migrated to the target network model.

[0030] In a second aspect, the present application also provides a flow data rollback device. The device comprises:

[0031] An acquisition module, used for acquiring first performance data during the operation of the target network model during the network model update process;

[0032] a determination module, configured to determine, when the first performance data does not meet the model operation condition, a data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data;

[0033] The rollback module is used to roll back part of the traffic data or all of the traffic data processed by the target network model to the historical network model according to the data rollback strategy.

[0034] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0035] During the network model updating process, obtaining first performance data during the operation of the target network model;

[0036] In a case where the first performance data does not meet the model operation condition, determining a data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data;

[0037] According to the data rollback strategy, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0038] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0039] During the network model updating process, obtaining first performance data during the operation of the target network model;

[0040] In a case where the first performance data does not meet the model operation condition, determining a data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data;

[0041] According to the data rollback strategy, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0042] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0043] During the network model updating process, obtaining first performance data during the operation of the target network model;

[0044] In a case where the first performance data does not meet the model operation condition, determining a data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data;

[0045] According to the data rollback strategy, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0046] The above-mentioned traffic data rollback method, device, computer equipment and storage medium obtain the first performance data during the operation of the target network model, and then, when the first performance data does not meet the model operation conditions, determine the data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data, and according to the data rollback strategy, roll back part of the traffic data or all of the traffic data processed by the target network model to the historical network model. According to the above content, it can be seen that in the process of rolling back the traffic data to the historical network model, the present application does not directly roll back all the traffic data to the historical network model, but determines the data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data, and then, according to the data rollback strategy, rolls back part of the traffic data or all of the traffic data processed by the target network model to the historical network model; thereby, the flexibility of rolling back the traffic data to the historical network model is improved, and the ratio of the traffic data input to the target network model and the historical network model in the rollback strategy is dynamically adjusted according to the actual situation of the target network model, so as to achieve accurate and flexible adjustment of different abnormal situations of the target network model, and ensure the continuous and stable operation of the target network model. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 An application environment diagram of a flow data rollback method provided in an embodiment of the present application;

[0048] Figure 2 A schematic diagram of a flow chart of a first method for rolling back flow data provided in an embodiment of the present application;

[0049] Figure 3 A flow chart of a second method for rolling back flow data provided in an embodiment of the present application;

[0050] Figure 4 A schematic diagram of a flow chart of a third method for rolling back flow data provided in an embodiment of the present application;

[0051] Figure 5 A schematic diagram of a fourth method for rolling back flow data provided in an embodiment of the present application;

[0052] Figure 6 A structural block diagram of a flow data rollback device provided in an embodiment of the present application;

[0053] Figure 7 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0055] It is understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. In the description of the present application, the description of the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily have to be directed to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples without contradicting each other.

[0056] The flow data rollback method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. By obtaining the first performance data during the operation of the target network model, and then, when the first performance data does not meet the model operation conditions, the data rollback strategy corresponding to the target network model is determined according to the performance level corresponding to the first performance data, and according to the data rollback strategy, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.

[0057] In one embodiment, Figure 2 As shown, a flow data rollback method is provided, which is applied to Figure 1 The terminal 102 in the example is used as an example to illustrate, and the following steps are included:

[0058] S201, during the process of updating the network model, obtaining first performance data during the operation of the target network model.

[0059] Among them, the first performance data refers to the data used to characterize the operation status of the target network model; specifically, the first performance parameters may include but are not limited to: data parsing efficiency of the target network model, processing delay of the target network model, recall rate of the target network model, etc. The first performance parameters include many parameter categories, and the parameter types included in the first performance parameters are not limited here.

[0060] In one embodiment of the present application, when it is necessary to obtain the first performance data during the operation of the target network model, the stage corresponding to the current round of operation of the target network model can be verified. If the current round of operation of the target network model is a preliminary verification stage, the pre-set sample traffic data can be obtained, and the sample traffic data can be input into the target network model for operation, and the operation status of the target network model can be monitored in real time according to the SDN (Software-Defined Networking) controller to obtain the first performance data during the operation of the target network model. Alternatively, select part of the traffic data processed by the historical network model and input it into the target network model for operation, and monitor the operation status of the target network model in real time according to the SDN controller to obtain the first performance data during the operation of the target network model.

[0061] If the target network model in this round of operation is in the gradual traffic switching stage, the amount of data input to the target network model in this round is determined based on the corresponding ratio of the traffic data processed in the target network model during the previous round of traffic switching to the traffic data processed in the historical network model. Then, the traffic data corresponding to the data amount is input into the target network model for operation, and the operating status of the target network model is monitored in real time according to the SDN controller to obtain the first performance data during the operation of the target network model.

[0062] It should be noted that when it is detected that there are steps that require model updates, such as regular updates, model performance optimization requirements, and the release of new model functions, the SDN controller starts the model update operation; and obtains the latest version of the target network model from the model repository, and detects the integrity and availability of the version; selects a suitable network device and a backup environment corresponding to the production environment of the historical network model, deploys the target network model in the backup environment, and triggers the preliminary verification phase for the target network model; the backup environment and the production environment can be deployed on the same network device or on different network devices. If deployed on different network devices (usually in a side-by-side manner), the SDN controller dynamically adjusts the traffic path and directs the traffic to the backup environment for testing to ensure that the traffic can be seamlessly switched between the production environment and the backup environment to maintain path consistency.

[0063] In another embodiment of the present application, an abnormality detection alarm system can also be set up to implement abnormal alarm processing for the target network model during real-time detection of the first performance data of the target network model, if it is detected that the difference between the first performance data and the pre-set standard performance data exceeds a threshold.

[0064] S202: When the first performance data does not meet the model operation conditions, determine a data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data.

[0065] It should be noted that the model operating conditions include a performance data threshold. If the first performance data is less than the performance data threshold, it is determined that the first performance data does not meet the model operating conditions. On the contrary, if the first performance data is greater than or equal to the performance data threshold, it is determined that the first performance data meets the model operating conditions.

[0066] Furthermore, in the case where the first performance data does not meet the model operation conditions, in order to prevent the model anomaly of the target network model from affecting the subsequent model update process, it is necessary to determine the data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data, so as to complete the data rollback operation according to the data rollback strategy. In the case where the first performance data meets the model operation conditions, there is no need to roll back the target network model.

[0067] In one embodiment of the present application, when determining the performance level corresponding to the first performance data, different performance data ranges corresponding to different candidate levels can be determined in advance, and then, when the first performance data does not meet the model operating conditions, the performance data range to which the first performance data belongs is judged, and then, the candidate level corresponding to the performance data range to which the first performance data belongs is used as the performance level corresponding to the first performance data.

[0068] In another embodiment of the present application, when it is necessary to determine the data rollback policy corresponding to the target network model, the correspondence between the candidate levels and the candidate rollback policies can be determined in advance, and then, after determining the performance level corresponding to the first performance data, a reference level that is the same as the performance level is determined from the candidate levels, and then, the candidate rollback policy corresponding to the reference level in the correspondence is used as the data rollback policy corresponding to the target network model.

[0069] S203: According to the data rollback strategy, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0070] In one embodiment of the present application, the data rollback strategy records the ratio coefficient of the target network model processing traffic data and the historical network model processing traffic data; then, according to the ratio coefficient, the data volume of the target network model processing traffic data and the data volume of the historical network model processing traffic data are adjusted respectively, so as to achieve the rollback of part or all of the traffic data processed by the target network model to the historical network model.

[0071] Specifically, according to the data rollback strategy, the actual allocation coefficient corresponding to the traffic data processed by the target network model is adjusted to obtain the target allocation coefficient; then, according to the target allocation coefficient, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0072] The above-mentioned traffic data rollback method obtains the first performance data during the operation of the target network model, and then, when the first performance data does not meet the model operation conditions, determines the data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data, and according to the data rollback strategy, rolls back part of the traffic data or all of the traffic data processed by the target network model to the historical network model. According to the above content, it can be seen that in the process of rolling back the traffic data to the historical network model, the present application does not directly roll back all the traffic data to the historical network model, but determines the data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data, and then, according to the data rollback strategy, rolls back part of the traffic data or all of the traffic data processed by the target network model to the historical network model; thereby, the flexibility of rolling back the traffic data to the historical network model is improved, and the ratio of the traffic data input to the target network model and the historical network model in the rollback strategy is dynamically adjusted according to the actual situation of the target network model, so as to achieve accurate and flexible adjustment of different abnormal situations of the target network model, and ensure the continuous and stable operation of the target network model.

[0073] In one embodiment, if Figure 3 As shown, when it is necessary to roll back part of the traffic data or all of the traffic data processed by the target network model to the historical network model according to the data rollback strategy, the following contents may be specifically included:

[0074] S301, according to the data rollback strategy, adjust the actual allocation coefficient corresponding to the traffic data processed by the target network model to obtain the target allocation coefficient.

[0075] It should be noted that in order to flexibly select part of the traffic data or all of the traffic data to roll back to the historical network model according to different processing situations of the target network model, different target allocation coefficients can be determined according to different performance levels corresponding to the data rollback strategy.

[0076] As one example, the performance levels may include a mild abnormality level, a moderate abnormality level, and a severe abnormality level.

[0077] Specifically, when the performance level is a mild abnormality level, the data rollback strategy includes multiple first coefficient adjustment values. Therefore, when it is necessary to adjust the actual allocation coefficient corresponding to the traffic data processed by the target network model to obtain the target allocation coefficient, the actual allocation coefficient can be subtracted from each first coefficient adjustment value to obtain multiple first allocation coefficients; therefore, it can be understood that the data rollback strategy includes first coefficient adjustment values ​​corresponding to each first allocation coefficient.

[0078] For example, multiple first coefficient adjustment values ​​can be 10%, 20%, 30% and 40%; if the actual allocation coefficient is 70%, then after performing difference calculation on the actual allocation coefficient and each first coefficient adjustment value, a target allocation coefficient of 60% corresponding to the first coefficient adjustment value of 10% can be obtained; a target allocation coefficient of 50% corresponding to the first coefficient adjustment value of 20%; a target allocation coefficient of 40% corresponding to the first coefficient adjustment value of 30%; and a target allocation coefficient of 30% corresponding to the first coefficient adjustment value of 40%. Therefore, it can be determined that when the performance level is a mild abnormality level, there are four target allocation coefficients, namely, 30%, 40%, 50% and 60%.

[0079] When the performance level is a moderate abnormality level and a severe abnormality level, the data rollback strategy includes a second coefficient adjustment value. Therefore, when it is necessary to adjust the actual allocation coefficient corresponding to the traffic data processed by the target network model to obtain the target allocation coefficient, the actual allocation coefficient can be subtracted from the second coefficient adjustment value to obtain the target allocation coefficient.

[0080] For example, if the second coefficient adjustment value is 50% and the actual allocation coefficient is 70%, the actual allocation coefficient is subtracted from the second coefficient adjustment value to obtain a target allocation coefficient of 20%.

[0081] For example, if the second coefficient adjustment value is 50%, and the actual allocation coefficient is 50%, then the actual allocation coefficient is subtracted from the second coefficient adjustment value to obtain a target allocation coefficient of 0%.

[0082] S302: Roll back part of the traffic data or all of the traffic data processed by the target network model to the historical network model according to the target allocation coefficient.

[0083] In one embodiment of the present application, when the performance level is a mild abnormal level, a progressive rollback can be used to gradually adjust the traffic data distribution ratio flowing to the target network model and the historical network model, so as to achieve model correction with minimal impact on the target network model.

[0084] Specifically, when the performance level is a mild abnormality level, the target allocation coefficient includes multiple first allocation coefficients that decrease in sequence, and each first allocation coefficient is greater than 0 and less than the actual allocation coefficient. Therefore, when it is necessary to roll back part of the traffic data or all of the traffic data processed by the target network model to the historical network model, the part of the traffic data processed by the target network model can be gradient rolled back to the historical network model in the gradient order of the first allocation coefficients from large to small.

[0085] As an example, if the first allocation coefficients include: 60%, 50%, 40% and 30% in sequence, then in the first round of model updating, 60% of the traffic data processed by the historical network model is introduced into the target network model, in the second round of model updating, 50% of the traffic data processed by the historical network model is introduced into the target network model, in the third round of model updating, 40% of the traffic data processed by the historical network model is introduced into the target network model, and in the fourth round of model updating, 30% of the traffic data processed by the historical network model is introduced into the target network model.

[0086] It is further explained that in the process of rolling back the gradient of part of the traffic data processed by the target network model to the historical network model, if it is detected that the second performance data during the operation of the target network model meets the model operation conditions, it is prohibited to continue to roll back the gradient of part of the traffic data processed by the target network model to the historical network model.

[0087] As an example, if the first allocation coefficient includes: sixty percent, fifty percent, forty percent and thirty percent in sequence, then in the first round of model updating, sixty percent of the traffic data processed by the historical network model is introduced into the target network model, and the second performance data of the target network model during operation at the current moment is detected. If the model operation conditions are not met, the second round of model updating is performed. In the second round of model updating, fifty percent of the traffic data processed by the historical network model is introduced into the target network model, and the second performance data of the target network model during operation at the current moment is detected. If the model operation conditions are met, it is prohibited to continue to gradient roll back part of the traffic data processed by the target network model to the historical network model, and then the traffic data flowing into the target network model continues to be gradually increased.

[0088] In another embodiment of the present application, when the performance level is a moderate abnormal level, a partial rollback can be used to reduce the traffic data processed by the target network model and increase the traffic data processed by the historical network model, thereby achieving model correction with minimal impact on the target network model.

[0089] Specifically, when the performance level is a moderate abnormal level, the target allocation coefficient is greater than 0 and less than the actual allocation coefficient. When it is necessary to roll back part of the traffic data or all of the traffic data processed by the target network model to the historical network model, part of the traffic data processed by the target network model can be rolled back to the historical network model according to the target allocation coefficient.

[0090] Wherein, when the performance level is a moderate abnormal level, the second coefficient adjustment value is smaller than the actual allocation coefficient.

[0091] In another embodiment of the present application, when the performance level is a severe abnormal level, a quick rollback can be used to immediately switch all traffic data processed by the target network model back to the historical network model to implement model correction of the target network model.

[0092] Specifically, when the performance level is a severe abnormal level, the target allocation coefficient is equal to 0. When it is necessary to roll back part of the traffic data or all of the traffic data processed by the target network model to the historical network model, all of the traffic data processed by the target network model can be rolled back to the historical network model according to the target allocation coefficient.

[0093] Among them, when the performance level is a severe abnormal level, the second coefficient adjustment value is equal to the actual allocation coefficient.

[0094] It should be noted that, when the first performance data meets the model operation conditions, all traffic data of the historical network model processor are migrated to the target network model.

[0095] The above-mentioned traffic data rollback method obtains the target allocation coefficient by adjusting the actual allocation coefficient corresponding to the traffic data processed by the target network model, and then, according to the target allocation coefficient, rolls back part of the traffic data or all of the traffic data processed by the target network model to the historical network model, thereby improving the flexibility of rolling back the traffic data to the historical network model, and dynamically adjusting the ratio of traffic data input to the target network model and the historical network model in the rollback strategy according to the actual situation of the target network model, thereby achieving accurate and flexible adjustment of different abnormal situations of the target network model and ensuring the continuous and stable operation of the target network model.

[0096] In one embodiment, if Figure 4As shown, when it is necessary to determine the data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data, the following contents may be included:

[0097] S401, determining a quantized value of a data difference between first performance data and a preset performance threshold.

[0098] In one embodiment of the present application, when it is necessary to determine the data difference quantization value, a difference operation can be performed on the first performance data and the preset performance threshold, and the obtained operation result is the data difference quantization value between the first performance data and the preset performance threshold.

[0099] S402: Determine a performance level corresponding to the first performance data according to the data difference quantization value.

[0100] In one embodiment of the present application, when determining the performance level corresponding to the first performance data, different data difference quantization value ranges corresponding to different candidate levels can be determined in advance, and then, when the first performance data does not meet the model operating conditions, the data difference quantization value between the first performance data and a preset performance threshold is determined, and the performance data range to which the data difference quantization value belongs is determined, and then, the candidate level corresponding to the performance data range to which the data difference quantization value belongs is used as the performance level corresponding to the first performance data.

[0101] S403: Determine a data rollback strategy corresponding to the target network model according to the candidate level and performance level corresponding to each candidate rollback strategy.

[0102] It should be noted that the candidate rollback strategy with the same candidate level and performance level may be used as the data rollback strategy corresponding to the target network model.

[0103] In one embodiment of the present application, when it is necessary to determine the data rollback policy corresponding to the target network model, the correspondence between the candidate levels and the candidate rollback policies can be determined in advance, and then, after determining the performance level corresponding to the first performance data, a reference level that is the same as the performance level is determined from the candidate levels, and then, the candidate rollback policy corresponding to the reference level in the correspondence is used as the data rollback policy corresponding to the target network model.

[0104] The above-mentioned traffic data rollback method determines the data difference quantification value between the first performance data and the preset performance threshold, and then determines the data rollback strategy corresponding to the target network model. The flexibility of rolling back the traffic data to the historical network model is improved, and the ratio of traffic data input to the target network model and the historical network model in the rollback strategy is dynamically adjusted according to the actual situation of the target network model, so as to achieve accurate and flexible adjustment of different abnormal situations of the target network model and ensure the continuous and stable operation of the target network model.

[0105] In one embodiment, if Figure 5 As shown, when part of the traffic data or all of the traffic data processed by the target network model needs to be rolled back to the historical network model, the following contents may be included:

[0106] S501, during the process of updating the network model, obtaining first performance data during the operation of the target network model.

[0107] S502: When the first performance data does not meet the model operation condition, determine a quantitative value of the data difference between the first performance data and a preset performance threshold.

[0108] S503: Determine a performance level corresponding to the first performance data according to the data difference quantization value.

[0109] S504: Determine a data rollback strategy corresponding to the target network model according to the candidate level and performance level corresponding to each candidate rollback strategy.

[0110] S505: According to the data rollback strategy, the actual allocation coefficient corresponding to the traffic data processed by the target network model is adjusted to obtain the target allocation coefficient.

[0111] S506: According to the target allocation coefficient, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0112] The above-mentioned traffic data rollback method obtains the first performance data during the operation of the target network model, and then, when the first performance data does not meet the model operation conditions, determines the data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data, and according to the data rollback strategy, rolls back part of the traffic data or all of the traffic data processed by the target network model to the historical network model. According to the above content, it can be seen that in the process of rolling back the traffic data to the historical network model, the present application does not directly roll back all the traffic data to the historical network model, but determines the data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data, and then, according to the data rollback strategy, rolls back part of the traffic data or all of the traffic data processed by the target network model to the historical network model; thereby, the flexibility of rolling back the traffic data to the historical network model is improved, and the ratio of the traffic data input to the target network model and the historical network model in the rollback strategy is dynamically adjusted according to the actual situation of the target network model, so as to achieve accurate and flexible adjustment of different abnormal situations of the target network model, and ensure the continuous and stable operation of the target network model.

[0113] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0114] Based on the same inventive concept, the embodiment of the present application also provides a flow data rollback device for implementing the flow data rollback method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the one or more flow data rollback device embodiments provided below can refer to the limitations of the flow data rollback method above, and will not be repeated here.

[0115] In one embodiment, Figure 6 As shown, a flow data rollback device is provided, comprising: an acquisition module 10, a determination module 20 and a rollback module 30, wherein:

[0116] The acquisition module 10 is used to acquire the first performance data during the operation of the target network model during the network model update process.

[0117] The determination module 20 is used to determine a data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data when the first performance data does not meet the model operation conditions.

[0118] The rollback module 30 is used to roll back part of the traffic data or all of the traffic data processed by the target network model to the historical network model according to the data rollback strategy.

[0119] In one embodiment, according to the data rollback strategy, the actual allocation coefficient corresponding to the traffic data processed by the target network model is adjusted to obtain the target allocation coefficient;

[0120] According to the target allocation coefficient, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0121] In one embodiment, according to the gradient order of the first allocation coefficients from large to small, a portion of the traffic data processed by the target network model is gradient rolled back to the historical network model.

[0122] In one embodiment, during the process of gradient rolling back part of the traffic data processed by the target network model to the historical network model, if it is detected that the second performance data during the operation of the target network model meets the model operation conditions, it is prohibited to continue to gradient roll back part of the traffic data processed by the target network model to the historical network model.

[0123] In one embodiment, a difference operation is performed between the actual allocation coefficient and each first coefficient adjustment value to obtain a plurality of first allocation coefficients.

[0124] In one embodiment, part of the traffic data processed by the target network model is rolled back to the historical network model according to the target allocation coefficient.

[0125] In one embodiment, all traffic data processed by the target network model is rolled back to the historical network model according to the target allocation coefficient.

[0126] In one embodiment, a difference operation is performed between the actual allocation coefficient and the second coefficient adjustment value to obtain the target allocation coefficient.

[0127] In one embodiment, determining a quantitative value of a data difference between the first performance data and a preset performance threshold;

[0128] Determining a performance level corresponding to the first performance data according to the data difference quantization value;

[0129] According to the candidate level and performance level corresponding to each candidate rollback strategy, a data rollback strategy corresponding to the target network model is determined.

[0130] In one embodiment, when the first performance data meets the model operation conditions, all traffic data of the historical network model processor is migrated to the target network model.

[0131] The above-mentioned traffic data rollback device obtains the first performance data during the operation of the target network model, and then, when the first performance data does not meet the model operation conditions, determines the data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data, and rolls back part of the traffic data or all of the traffic data processed by the target network model to the historical network model according to the data rollback strategy. According to the above content, it can be seen that in the process of rolling back the traffic data to the historical network model, the present application does not directly roll back all the traffic data to the historical network model, but determines the data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data, and then rolls back part of the traffic data or all of the traffic data processed by the target network model to the historical network model according to the data rollback strategy; thereby, the flexibility of rolling back the traffic data to the historical network model is improved, and the ratio of the traffic data input to the target network model and the historical network model in the rollback strategy is dynamically adjusted according to the actual situation of the target network model, so as to achieve accurate and flexible adjustment of different abnormal situations of the target network model, and ensure the continuous and stable operation of the target network model.

[0132] Each module in the above-mentioned traffic data rollback device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each of the above modules.

[0133] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 7As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for rolling back traffic data is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0134] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0135] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0136] During the network model updating process, obtaining first performance data during the operation of the target network model;

[0137] When the first performance data does not meet the model operation condition, determining a data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data;

[0138] According to the data rollback strategy, part or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0139] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0140] According to the data rollback strategy, the actual allocation coefficient corresponding to the traffic data processed by the target network model is adjusted to obtain the target allocation coefficient;

[0141] According to the target allocation coefficient, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0142] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0143] According to the gradient order of the first allocation coefficients from large to small, a portion of the traffic data processed by the target network model is gradient rolled back to the historical network model.

[0144] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0145] In the process of rolling back the gradient of part of the traffic data processed by the target network model to the historical network model, if it is detected that the second performance data during the operation of the target network model meets the model operation conditions, it is prohibited to continue rolling back the gradient of part of the traffic data processed by the target network model to the historical network model.

[0146] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0147] The actual allocation coefficient is respectively subjected to difference calculation with each first coefficient adjustment value to obtain a plurality of first allocation coefficients.

[0148] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0149] According to the target allocation coefficient, part of the traffic data processed by the target network model is rolled back to the historical network model.

[0150] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0151] According to the target allocation coefficient, all traffic data processed by the target network model is rolled back to the historical network model.

[0152] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0153] The actual allocation coefficient is calculated by difference with the second coefficient adjustment value to obtain the target allocation coefficient.

[0154] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0155] Determine a quantitative value of a data difference between the first performance data and a preset performance threshold;

[0156] Determining a performance level corresponding to the first performance data according to the data difference quantization value;

[0157] According to the candidate level and performance level corresponding to each candidate rollback strategy, a data rollback strategy corresponding to the target network model is determined.

[0158] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0159] When the first performance data meets the model operation conditions, all traffic data of the historical network model processor is migrated to the target network model.

[0160] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0161] During the network model updating process, obtaining first performance data during the operation of the target network model;

[0162] When the first performance data does not meet the model operation condition, determining a data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data;

[0163] According to the data rollback strategy, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0164] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0165] According to the data rollback strategy, the actual allocation coefficient corresponding to the traffic data processed by the target network model is adjusted to obtain the target allocation coefficient;

[0166] According to the target allocation coefficient, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0167] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0168] According to the gradient order of the first allocation coefficients from large to small, part of the traffic data processed by the target network model is gradient rolled back to the historical network model.

[0169] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0170] In the process of rolling back the gradient of part of the traffic data processed by the target network model to the historical network model, if it is detected that the second performance data during the operation of the target network model meets the model operation conditions, it is prohibited to continue rolling back the gradient of part of the traffic data processed by the target network model to the historical network model.

[0171] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0172] The actual allocation coefficient is respectively subjected to difference calculation with each first coefficient adjustment value to obtain a plurality of first allocation coefficients.

[0173] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0174] According to the target allocation coefficient, part of the traffic data processed by the target network model is rolled back to the historical network model.

[0175] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0176] According to the target allocation coefficient, all traffic data processed by the target network model is rolled back to the historical network model.

[0177] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0178] The actual allocation coefficient is calculated by difference with the second coefficient adjustment value to obtain the target allocation coefficient.

[0179] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0180] Determine a quantitative value of a data difference between the first performance data and a preset performance threshold;

[0181] Determining a performance level corresponding to the first performance data according to the data difference quantization value;

[0182] According to the candidate level and performance level corresponding to each candidate rollback strategy, a data rollback strategy corresponding to the target network model is determined.

[0183] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0184] When the first performance data meets the model operation conditions, all traffic data of the historical network model processor is migrated to the target network model.

[0185] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0186] During the network model updating process, obtaining first performance data during the operation of the target network model;

[0187] When the first performance data does not meet the model operation condition, determining a data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data;

[0188] According to the data rollback strategy, part or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0189] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0190] According to the data rollback strategy, the actual allocation coefficient corresponding to the traffic data processed by the target network model is adjusted to obtain the target allocation coefficient;

[0191] According to the target allocation coefficient, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

[0192] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0193] According to the gradient order of the first allocation coefficients from large to small, part of the traffic data processed by the target network model is gradient rolled back to the historical network model.

[0194] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0195] In the process of rolling back the gradient of part of the traffic data processed by the target network model to the historical network model, if it is detected that the second performance data during the operation of the target network model meets the model operation conditions, it is prohibited to continue rolling back the gradient of part of the traffic data processed by the target network model to the historical network model.

[0196] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0197] The actual allocation coefficient is respectively subjected to difference calculation with each first coefficient adjustment value to obtain a plurality of first allocation coefficients.

[0198] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0199] According to the target allocation coefficient, part of the traffic data processed by the target network model is rolled back to the historical network model.

[0200] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0201] According to the target allocation coefficient, all traffic data processed by the target network model is rolled back to the historical network model.

[0202] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0203] The actual allocation coefficient is calculated by difference with the second coefficient adjustment value to obtain the target allocation coefficient.

[0204] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0205] Determine a quantitative value of a data difference between the first performance data and a preset performance threshold;

[0206] Determining a performance level corresponding to the first performance data according to the data difference quantization value;

[0207] According to the candidate level and performance level corresponding to each candidate rollback strategy, a data rollback strategy corresponding to the target network model is determined.

[0208] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0209] When the first performance data meets the model operation conditions, all traffic data of the historical network model processor is migrated to the target network model.

[0210] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0211] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0212] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0213] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for rolling back flow data, characterized in that: The method comprises: During the network model updating process, obtaining first performance data during the operation of the target network model; In a case where the first performance data does not meet the model operation condition, determining a data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data; According to the data rollback strategy, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

2. The method according to claim 1, characterized in that The step of rolling back part of the traffic data or all of the traffic data processed by the target network model to the historical network model according to the data rollback strategy includes: According to the data rollback strategy, adjusting the actual allocation coefficient corresponding to the traffic data processed by the target network model to obtain a target allocation coefficient; According to the target allocation coefficient, part of the traffic data or all of the traffic data processed by the target network model is rolled back to the historical network model.

3. The method according to claim 2, characterized in that In the case where the performance level is a mild abnormal level, the target allocation coefficient includes a plurality of first allocation coefficients that decrease in sequence, and each of the first allocation coefficients is greater than 0 and less than the actual allocation coefficient, and according to the target allocation coefficient, rolling back part of the traffic data or all of the traffic data processed by the target network model to the historical network model includes: According to the gradient order of the first allocation coefficients from large to small, a portion of the traffic data processed by the target network model is gradient rolled back to the historical network model.

4. The method according to claim 3, characterized in that The method further comprises: During the process of gradient rolling back part of the traffic data processed by the target network model to the historical network model, if it is detected that the second performance data during the operation of the target network model meets the model operation conditions, it is prohibited to continue to gradient roll back part of the traffic data processed by the target network model to the historical network model.

5. The method according to claim 3 or 4, characterized in that: The data rollback strategy includes first coefficient adjustment values ​​corresponding to each of the first allocation coefficients. According to the data rollback strategy, the actual allocation coefficient corresponding to the traffic data processed by the target network model is adjusted to obtain the target allocation coefficient, including: The actual allocation coefficient is respectively subjected to difference calculation with each of the first coefficient adjustment values ​​to obtain the plurality of first allocation coefficients.

6. The method according to claim 2, characterized in that When the performance level is a moderate abnormal level, the target allocation coefficient is greater than 0 and less than the actual allocation coefficient, and according to the target allocation coefficient, rolling back part of the traffic data or all of the traffic data processed by the target network model to the historical network model includes: According to the target allocation coefficient, part of the traffic data processed by the target network model is rolled back to the historical network model.

7. The method according to claim 2, characterized in that: When the performance level is a severe abnormal level, the target allocation coefficient is equal to 0, and according to the target allocation coefficient, rolling back part of the traffic data or all of the traffic data processed by the target network model to the historical network model includes: According to the target allocation coefficient, all traffic data processed by the target network model are rolled back to the historical network model.

8. The method according to claim 6 or 7, characterized in that: The data rollback strategy includes a second coefficient adjustment value, and adjusting the actual allocation coefficient corresponding to the traffic data processed by the target network model according to the data rollback strategy to obtain the target allocation coefficient includes: The actual allocation coefficient is subjected to a difference operation with the second coefficient adjustment value to obtain the target allocation coefficient.

9. The method according to claim 1, characterized in that: The determining, according to the performance level corresponding to the first performance data, a data rollback strategy corresponding to the target network model includes: Determine a quantitative value of a data difference between the first performance data and a preset performance threshold; Determining the performance level corresponding to the first performance data according to the data difference quantization value; The data rollback strategy corresponding to the target network model is determined according to the candidate level corresponding to each candidate rollback strategy and the performance level.

10. The method according to any one of claims 1 to 9, characterized in that The method further comprises: When the first performance data meets the model operation condition, all traffic data of the historical network model processor is migrated to the target network model.

11. A flow data rollback device, characterized in that: The device comprises: An acquisition module, used for acquiring first performance data during the operation of the target network model during the network model update process; a determination module, configured to determine, when the first performance data does not meet the model operation condition, a data rollback strategy corresponding to the target network model according to the performance level corresponding to the first performance data; The rollback module is used to roll back part of the traffic data or all of the traffic data processed by the target network model to the historical network model according to the data rollback strategy.

12. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 10 are implemented.

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

14. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 10 are implemented.

Images