Data migration method and device, computer equipment, storage medium and program product

Abstract

The invention relates to a data migration method and device, computer equipment, a storage medium and a program product. The method comprises the following steps: acquiring configuration data to be migrated to a target system in a source system; based on a preset data migration model, performing enhancement processing on the encoded data features of the configuration data, and decoding the encoded data features after enhancement processing to obtain target data; and migrating the target data to the target system. By adopting the method, the migration efficiency can be improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a data migration method, apparatus, computer equipment, storage medium and program product. Background Art

[0002] In the computer field, data migration refers to the transfer of data from one computer storage system to another computer storage system. It is widely used in scenarios such as the implementation, upgrade or integration of computer storage systems.

[0003] In the related art, when performing data migration, the data in the source system is usually converted into a data version compatible with the target system according to the configuration mapping relationship between the source system and the target system, and then migrated to the target system to complete the data migration.

[0004] However, the related technology has the problem of low data migration efficiency. Summary of the invention

[0005] Based on this, it is necessary to provide a data migration method, apparatus, computer equipment, storage medium and program product that can improve migration efficiency in response to the above technical problems.

[0006] In a first aspect, the present application provides a data migration method, the method comprising:

[0007] Obtain configuration data from the source system to be migrated to the target system;

[0008] Based on a preset data migration model, the encoded data features of the configuration data are enhanced, and the enhanced encoded data features are decoded to obtain target data;

[0009] Migrate the target data to the target system.

[0010] In one embodiment, the data migration model includes a feature encoding model and a feature decoding model; enhancing the encoded data features of the configuration data based on the preset data migration model, and decoding the enhanced encoded data features to obtain target data, including:

[0011] The mapping relationship between the source system and the target system is extracted through the feature coding model, and the coding data features are enhanced based on the mapping relationship to obtain the enhanced coding data features;

[0012] The enhanced encoded data features are decoded through the feature decoding model to obtain the target data.

[0013] In one embodiment, the feature coding model includes a system mapping model, a feature extraction model and a feature fusion model; the mapping relationship between the source system and the target system is extracted through the feature coding model, and the coded data features are enhanced based on the mapping relationship to obtain the enhanced coded data features, including:

[0014] Extract mapping features of configuration data through the system mapping model to obtain the mapping relationship between the source system and the target system;

[0015] Perform feature dimension-upgrading processing on the configuration data according to the feature extraction model to obtain the encoding data features;

[0016] The mapping relationship and the encoded data features are input into a feature fusion model, and the encoded data features are enhanced based on the mapping relationship to obtain the enhanced encoded data features.

[0017] In one embodiment, the data migration model includes: a system mapping model, a feature extraction model, a feature fusion model and a feature decoding model; the training process of the data migration model includes:

[0018] According to the first migration data set, an initial system mapping model in the initial data migration model is trained to obtain a candidate data migration model; the candidate data model includes a system mapping model;

[0019] According to the second migration data set, the initial feature extraction model, the initial feature fusion model and the initial feature decoding model in the candidate data migration model are trained to obtain a data migration model.

[0020] In one embodiment, the second migration data set includes a historical configuration data set; based on the second migration data set, an initial feature extraction model, an initial feature fusion model, and an initial feature decoding model in the candidate data migration model are trained to obtain a data migration model, including:

[0021] Inputting the unmigrated data in the historical configuration data set into the candidate data migration model to obtain the transformed data of the unmigrated data;

[0022] Based on the completeness quantification value, accuracy quantification value and migration efficiency quantification value between the converted data and the real migration data corresponding to the unmigrated data, an evaluation value of the candidate data migration model is obtained;

[0023] Based on the evaluation values, the initial feature extraction model, the initial feature fusion model and the initial feature decoding model are trained to obtain a data migration model.

[0024] In one embodiment, the method further comprises:

[0025] Monitor the data migration status between source and target systems;

[0026] In case of target data migration exception, roll back the target data to the configuration data and collect migration exception log information of the target data;

[0027] Based on the migration exception log information, the model parameters in the data migration model are corrected.

[0028] In a second aspect, the present application further provides a data migration device, including:

[0029] A data acquisition module is used to acquire configuration data to be migrated from a source system to a target system;

[0030] A data enhancement module, used to enhance the encoded data features of the configuration data based on a preset data migration model, and to decode the enhanced encoded data features to obtain target data;

[0031] The data migration module is used to migrate the target data to the target system.

[0032] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method in any one of the embodiments of the first aspect are implemented.

[0033] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method in any one of the embodiments of the first aspect above.

[0034] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the method in any one of the embodiments of the first aspect above.

[0035] The above-mentioned data migration method, device, computer equipment, storage medium and program product, based on the configuration data to be migrated from the source system to the target system, enhances the coded data features of the configuration data based on the preset data migration model, and decodes the coded data features after the enhancement process to obtain the target data, and migrates the target data to the target system. In this method, based on the acquisition of the configuration data, by calling the data migration model, the steps of extracting the coded data features of the configuration data, enhancing the extracted coded data features, and decoding the coded data features after the enhancement process are automatically executed to realize the fully automatic conversion of the configuration data from the source system to the target system, and migrate the converted target data to the target system, thereby improving the degree of automation of the data migration process, avoiding the complexity and errors caused by manual operations, and improving the efficiency of data migration. In addition, the entire data migration process is carried out according to the preset data migration model, without relying on specific tools and services. Due to the high flexibility and scalability of the data migration model, it supports adapting to the ever-changing business needs and technical architecture of the enterprise, without the need to frequently change the migration plan, and reducing the cost of technology updates for the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0037] FIG1 is an application environment diagram of a data migration method in one embodiment;

[0038] FIG2 is a schematic diagram of a flow chart of a data migration method in one embodiment;

[0039] FIG3 is a schematic diagram of a flow chart of a data conversion step in one embodiment;

[0040] FIG4 is a schematic diagram of the architecture of a data migration model in one embodiment;

[0041] FIG5 is a schematic diagram of a flow chart of a model training step in one embodiment;

[0042] FIG6 is a schematic diagram of a flow chart of a model training step in another embodiment;

[0043] FIG7 is a schematic diagram of a flow chart of a model training step in another embodiment;

[0044] FIG8 is a schematic diagram of a flow chart of a data migration method in another embodiment;

[0045] FIG9 is a schematic diagram of a flow chart of a model training step in another embodiment;

[0046] FIG10 is a block diagram of a data conversion device in one embodiment;

[0047] FIG. 11 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0048] 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.

[0049] The technical background of this application is first described below.

[0050] In the computer field, data migration refers to the process of selecting, preparing, extracting, and transforming data and permanently transferring it from one computer storage system to another. The data migration process includes verifying the integrity of the migrated data and decommissioning the original old data storage. In practical applications, data migration is a key consideration for computer system implementation, upgrade, or integration.

[0051] In the related art, when performing data migration, there are at least two migration methods: one is to use manual migration, relying on migration experience to configure and convert the data of the source system. In the scenario of processing large-scale and complex data structures, data migration is often slow or even fails due to data loss or incorrect format; the other is to use specific tools to migrate data for specified types of systems, which limits the data migration scenario. On this basis, whenever migration of different types of systems is required, corresponding migration tools need to be developed, which greatly increases the cost of data migration. In summary, the data migration method in the related art has many disadvantages in the scenario of data migration between large and complex computer storage systems, which affects the efficiency and accuracy of data migration.

[0052] Based on this, the embodiments of the present application provide a data migration method, apparatus, computer equipment, storage medium and program product. On the basis of acquiring the configuration data to be migrated from the source system to the target system, the encoded data features of the configuration data are enhanced based on a preset data migration model, and the encoded data features after the enhanced processing are decoded to obtain the target data, and the target data is migrated to the target system, thereby realizing fully automatic conversion of the configuration data from the source system to the target system, and migrating the converted target data to the target system, thereby improving the degree of automation of the data migration process, avoiding the complexity and errors caused by manual operations, and improving the efficiency of data migration.

[0053] The data migration method provided in the embodiment of the present application can be applied to the application environment shown in Figure 1. It includes a source system 102 and a target system 104, wherein the source system 102 and the target system 104 can be heterogeneous cloud platforms, each heterogeneous cloud platform includes a variety of different types of processors and computing resources, including but not limited to a central processing unit (CPU), a graphics processing unit (GPU), a field programmable gate array (FPGA), etc.

[0054] In actual applications, heterogeneous cloud platforms can allocate computing loads to the most suitable processors for executing tasks according to the characteristics and requirements of the tasks. Multiple processors can execute tasks in parallel, improving the performance and energy efficiency of the entire heterogeneous cloud platform system, and thus being widely used in various enterprises. In addition, with the in-depth development of cloud computing technology, enterprises have an increasing demand for data migration between heterogeneous cloud platforms and the frequency of data migration. Therefore, data migration is not only a technical challenge, but also the key for enterprises to maintain business continuity and their own service advantages.

[0055] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0056] In an exemplary embodiment, as shown in FIG2 , a data migration method is provided. The embodiment of the present application takes the method applied to the target system as an example for illustration. It is understandable that the method can also be applied to the source system, and can also be applied to the migration system including the source system and the target system, and is implemented through the interaction between the source system and the target system. In the embodiment of the present application, the method includes the following steps:

[0057] S201, obtaining configuration data to be migrated from a source system to a target system.

[0058] The source system and the target system refer to computer storage systems, which can be two heterogeneous platforms, or two heterogeneous cloud platforms, or a computer system and a cloud platform, etc. The configuration data is the data migrated from the source system to the target system, which can be part of the data of the source system or all of the data of the source system.

[0059] When obtaining configuration data, it is possible to respond to a data migration request from the source system and detect whether the data migration request includes a configuration data identifier. If the configuration data identifier is included, the data corresponding to the configuration data identifier is queried from the database of the source system as the configuration data to be migrated. If the configuration data identifier is not included, all data in the source system are compressed by default as configuration data.

[0060] In another configuration data acquisition scenario, the source system may also carry the configuration data in the data migration request. In this case, the target system directly parses the data migration request to obtain the configuration data to be migrated.

[0061] It should be noted that the source system and the target system include a variety of different processors, and the data formats, data sizes, and data types supported by different systems are different. Based on this, in order to ensure the data availability after the configuration data is migrated, in the process of migrating data from the source system to the target system, it is necessary to first perform a series of conversions on the configuration data of the source system according to the data storage rules of the target system, and then apply the converted data to the target system.

[0062] S202, enhancing the encoded data features of the configuration data based on a preset data migration model, and decoding the enhanced encoded data features to obtain target data.

[0063] Among them, the data migration model can be a deep learning model based on artificial neural networks, including a complex network structure of multiple layers of neurons, which supports automatic learning of features and processing modes from large amounts of data and can process high-dimensional and complex data such as images, voices, and texts. The data migration model can also be a reinforcement learning model, which continuously interacts with the environment and adjusts the data processing strategy based on the reward feedback obtained.

[0064] In the embodiment of the present application, the data migration model is used to convert the configuration data in the source system into the target data supported by the target system while keeping the content of the configuration data unchanged. During use, the configuration data is input into the data migration model, and the data migration model performs a series of operations such as analysis, encoding, and decoding on the configuration data to obtain the target data output by the data migration model.

[0065] Optionally, the data migration model parses the configuration data, obtains the mapping relationship between the configuration data and the target platform, and extracts the encoding features of the configuration data. The encoding features are then enhanced according to the mapping relationship to obtain enhanced features of the configuration data, and then the enhanced encoded data features are decoded to obtain the target data.

[0066] S203, migrating the target data to the target system.

[0067] After obtaining the target data output by the data migration model, the target data can be directly migrated to the target system, or the target data can be verified again for data integrity and data authenticity, and the verified target data can be migrated to the target system to complete the entire data migration.

[0068] In the embodiment of the present application, on the basis of obtaining the configuration data, by calling the data migration model, the steps of extracting the encoded data features of the configuration data, enhancing the extracted encoded data features, and decoding the encoded data features after the enhancement processing are automatically executed, so as to realize the fully automatic conversion of the configuration data from the source system to the target system, and migrate the converted target data to the target system, thereby improving the automation degree of the data migration process, avoiding the complexity and errors caused by manual operation, and improving the efficiency of data migration. In addition, the entire data migration process is based on a preset data migration model, and does not rely on specific tools and services. Due to the high flexibility and scalability of the data migration model, it supports adapting to the ever-changing business needs and technical architecture of the enterprise, without the need to frequently change the migration plan, and reducing the cost of technology updates for the enterprise.

[0069] As can be seen from the above embodiments, the data migration model is used to convert configuration data into target data supported by the target system, and the data migration model may be a deep learning model, a reinforcement learning model, etc. Next, the architecture of the data migration model and a data processing step of the data migration model are described.

[0070] In an exemplary embodiment, as shown in FIG3 , the data migration model includes a feature encoding model and a feature decoding model; enhancing the encoded data features of the configuration data based on the preset data migration model, and decoding the enhanced encoded data features to obtain target data, including the following steps:

[0071] S301, extracting the mapping relationship between the source system and the target system through the feature coding model, and enhancing the coding data features based on the mapping relationship to obtain the enhanced coding data features.

[0072] The configuration data is input into the feature encoding model in the data migration model. The feature encoding model extracts the configuration features and encoding data features of the configuration data from the system mapping dimension and the data processing dimension respectively, and predicts the mapping relationship between the source system and the target system based on the configuration features. The mapping relationship is then integrated with the encoding data features to improve the effectiveness of the encoding data features and provide reliable encoding data features for subsequent decoding.

[0073] S302, decoding the enhanced encoded data features through a feature decoding model to obtain target data.

[0074] In the data migration model, the feature encoding model and the feature decoding model are connected in series. The feature decoding model takes the enhanced encoded data features output by the feature encoding model as input, and converts the enhanced encoded data features according to a preset data conversion method to obtain the target data.

[0075] Optionally, the data conversion method of the feature decoding model is format adjustment, and the enhanced encoded data features are decoded according to the format of the target system to obtain target data.

[0076] Optionally, the data conversion method of the feature decoding model is data compression, and the enhanced encoded data features are decoded according to a preset compression ratio to obtain target data.

[0077] Optionally, the data conversion method of the feature decoding model is field mapping. Based on the field name, the corresponding target encoded data features are matched from the enhanced encoded data features, and the target encoded data features are summarized to obtain the target data.

[0078] In an embodiment of the present application, a mapping relationship between a source system and a target system is extracted through a feature encoding model, and the encoded data features are enhanced based on the mapping relationship to obtain the enhanced encoded data features, and then the enhanced encoded data features are decoded through a feature decoding model to obtain target data, and through the encoding and decoding symmetric architecture of the data migration model, end-to-end conversion of configuration data and target data is achieved, thereby improving data processing efficiency.

[0079] The encoding implementation of the feature encoding model in the aforementioned embodiment can be performed using multiple different functional network sub-models. In an exemplary embodiment, as shown in FIG4 , the feature encoding model includes a system mapping model, a feature extraction model, and a feature fusion model. In FIG4 , the system mapping model and the feature extraction model are connected in parallel, the input of the system mapping model and the feature extraction model is configuration data, and the output of the system mapping model (the mapping relationship between the source system and the target system) and the output of the feature extraction model (the encoded data features) are used together as the input of the feature fusion model, and the output result of the feature fusion model (the encoded data features after enhancement processing) is the input of the feature decoding model.

[0080] In the data migration model shown in FIG4 , the mapping relationship between the source system and the target system is extracted through the feature encoding model, and the encoded data features are enhanced based on the mapping relationship to obtain the enhanced encoded data features, including the following steps:

[0081] The mapping features of the configuration data are extracted through the system mapping model to obtain the mapping relationship between the source system and the target system; the configuration data is processed for feature dimension upgrading according to the feature extraction model to obtain the encoding data features; the mapping relationship and the encoding data features are input into the feature fusion model, and the encoding data features are enhanced based on the mapping relationship to obtain the enhanced encoding data features.

[0082] The system mapping model is used to extract system-related mapping features in the configuration data, and then the mapping relationship that matches the mapping features of the configuration data is queried from the correspondence between multiple groups of mapping features and mapping relationships stored internally in the system mapping model as the mapping relationship between the source system and the target system.

[0083] The feature extraction model includes a plurality of convolutional layers connected in series. The plurality of convolutional layers connected in series of the feature extraction model are used to extract features of the configuration data to obtain encoded data features.

[0084] The mapping relationship and the encoded data features are input into the feature fusion model to enhance the encoded data features according to the mapping relationship between the source system and the target system to obtain the enhanced encoded data features.

[0085] In an embodiment of the present application, configuration data is processed in parallel from two dimensions through a system mapping model and a feature extraction model to obtain a mapping relationship between a source system and a target system and encoding data features; the mapping relationship and encoding data features are fused through a feature fusion model, so that the enhanced encoded data features, while retaining key data information, increase system conversion rule features, thereby improving the effectiveness of the encoded data features and further improving the accuracy of data conversion.

[0086] The above embodiment describes the application process of the data migration model. Next, the training steps of the data migration model are described. In an exemplary embodiment, as shown in FIG5 , the data migration model includes: a system mapping model, a feature extraction model, a feature fusion model, and a feature decoding model; the training process of the data migration model includes the following steps:

[0087] S501 , training an initial system mapping model in an initial data migration model according to a first migration data set to obtain a candidate data migration model; the candidate data model includes a system mapping model.

[0088] The first migration data set includes historical configuration data before migration and historical target data after migration between multiple groups of systems in different scenarios. In the embodiment of the present application, the initial system mapping model is a neural network model, and the historical configuration data and historical target data are input into the initial system mapping model. The initial system mapping model captures the potential rules and structures of the data, and learns the system mapping relationship between the system of the historical configuration data and the system corresponding to the historical target data.

[0089] Optionally, the first migration data set includes a first historical data set D between the two target systems. a The second historical data set D between other systems b The initial system mapping model is trained using a two-stage transfer learning approach. represents the initial system mapping model, represents the model parameters, Represents historical configuration data, Represents historical target data and explains the training process of the initial system mapping model:

[0090] In the first stage, based on the second historical dataset D b The initial system mapping model is pre-trained, and the loss function of the model is expressed as:

[0091]

[0092] In the above formula, The second historical data set D b The number of samples, represents the quantitative difference function between A and B, such as the absolute value function, is a loss function used to measure the difference between the historical prediction data and the historical target data of the model. The initial system mapping model is trained according to the above loss function until the loss function value is less than the preset loss threshold, and a pre-trained system mapping model is obtained.

[0093] In the second stage, the parameters of the initial system mapping model obtained by pre-training are adjusted using the first historical data set Da. The learning goal is to adapt the model to the characteristics and distribution of the target data set and improve the performance of the model on the target task. At this time, the loss function of the model is expressed as:

[0094]

[0095] in, is the number of samples of the first historical data set Da.

[0096] In the second stage of parameter adjustment, in order to avoid excessive damage to the pre-training parameters, a smaller learning rate can be used for training. At the same time, early stopping can be used to prevent the model from overfitting.

[0097] In actual training, the initial system mapping model is a sub-model of the initial data migration model. Training the initial system mapping model is equivalent to training the initial system mapping model in the initial data migration model. Further, when the initial system mapping model training is completed, the initial data migration model is updated to the candidate data migration model. In other words, the initial data migration model includes the initial system mapping model, the initial feature extraction model, the initial feature fusion model and the initial feature decoding model, and the candidate data migration model includes the system mapping model, the initial feature extraction model, the initial feature fusion model and the initial feature decoding model.

[0098] S502: According to the second migration data set, an initial feature extraction model, an initial feature fusion model and an initial feature decoding model in the candidate data migration model are trained to obtain a data migration model.

[0099] Similar to the content of the first migration data set, the second migration data set includes historical configuration data before migration and historical target data after migration between multiple groups of systems in different scenarios. However, the difference between the second migration data set and the first migration data set is that the second migration data set can be a partial data set in the first migration data set, such as the first historical data set D in the first migration data set in the above step. a Of course, it can also be a data set that is completely different from the first migration data set, and the amount of data in the second migration data set is smaller than that of the first migration data set.

[0100] When training the candidate data migration model, the second migration data set is input into the candidate data migration model, and the format conversion logic between the historical configuration data and the historical target data is learned through the candidate data migration model to obtain the data migration model.

[0101] In an embodiment of the present application, the initial system mapping model in the initial data migration model is trained according to the first migration data set to obtain a candidate data migration model; according to the second migration data set, the models other than the system mapping model in the candidate data migration model are trained to obtain a data migration model. The initial data migration model is trained in two stages using two types of data sets, so that the training data set of the data migration model involves more training samples, thereby improving the training performance of the data migration model.

[0102] In an exemplary embodiment, as shown in FIG6 , the second migration data set includes a historical configuration data set; according to the second migration data set, the initial feature extraction model, the initial feature fusion model and the initial feature decoding model in the candidate data migration model are trained to obtain the data migration model, including the following steps:

[0103] S601: inputting the unmigrated data in the historical configuration data set into the candidate data migration model to obtain the conversion data of the unmigrated data.

[0104] The historical configuration data set includes unmigrated data and real migration data corresponding to the unmigrated data.

[0105] The unmigrated data is input into the initial feature extraction model and the system mapping model of the candidate migration model. The feature extraction results of the initial feature extraction model and the system mapping model are fused through the initial feature fusion model. The fused features are decoded through the initial feature decoding model to obtain the converted data of the unmigrated data.

[0106] Taking SourceConfig as an example to represent unmigrated data, the data processing steps of the candidate data migration model can be expressed as follows:

[0107]

[0108] Among them, "Encoder()" represents the encoding function of the initial feature extraction model, and "EncodedFeatures" represents the encoded data features output by the initial feature extraction model.

[0109] Next, based on the initial feature fusion model, the prediction results (system mapping relationship) of the system mapping model are used to enhance the coded data features output by the initial feature extraction model. The enhanced coded data features are:

[0110]

[0111] Then, the task of the initial feature decoding model (Decoder) is to decode the fused enhanced encoded data features according to the conversion rules and the knowledge learned by the model, and convert them into target data that meets the requirements of the target system, so as to achieve accurate conversion and adaptation of the data and obtain the converted data of the unmigrated data:

[0112]

[0113] “ " represents the enhanced processed coded data features, " "Represents a series of parameters in the data transfer model in addition to the system mapping model parameters. During the model training process, these parameters are continuously adjusted and improved through continuous data reception, learning and optimization to reach the optimal state, thereby ensuring the accuracy and effectiveness of the conversion.

[0114] S602, obtaining an evaluation value of the candidate data migration model based on the completeness quantization value, the accuracy quantization value and the migration efficiency quantization value between the converted data and the real migration data corresponding to the non-migrated data.

[0115] The converted data of the unmigrated data and the actual migrated data are compared to obtain the integrity quantification value, accuracy quantification value and migration efficiency quantification value between the two. Then, according to the preset weights, the weighted sum of the integrity quantification value, accuracy quantification value and migration efficiency quantification value is used as the evaluation value of the candidate data migration model.

[0116] The integrity quantization value represents the loss or damage of configuration data during the model configuration data conversion process. The less configuration data is lost or damaged, the closer the data volume of the converted data is to the actual migration data, and the higher the integrity quantization value is. The more configuration data is lost or damaged, the greater the difference between the converted data and the actual migration data is, and the lower the integrity quantization value is.

[0117] The accuracy quantification value represents the consistency of the key fields of the configuration data and the real migration data during the model's configuration data conversion process. The more consistent the configuration data is with the real migration data, the more elements of the configuration data are accurately mapped or converted, and the higher the accuracy quantification value. For example, a certain reward value is given for each accurately mapped field or correctly converted configuration parameter, and an additional reward is given when the overall mapping or conversion accuracy reaches a certain threshold.

[0118] The migration efficiency quantitative value represents the efficiency of completing the mapping and conversion tasks while ensuring the quality. It can be represented by the time required to complete the mapping or conversion, the memory resources used, the computing resources used, etc. The shorter the time, the fewer computing resources used, and the higher the migration efficiency quantitative value; the longer the time, the more computing resources used, and the lower the migration efficiency quantitative value.

[0119] Next, the evaluation value R of the candidate data migration model can be expressed as:

[0120]

[0121] In the above formula, , and Respectively represent the accuracy quantification value, integrity quantification value and migration efficiency quantification value, with a value range of 0 to 1; , and The weights representing the accuracy quantification value, the integrity quantification value, and the migration efficiency quantification value may be 0.5, 0.3, and 0.2, respectively, and may be adjusted according to actual application scenarios and requirements.

[0122] S603: Based on the evaluation value, the initial feature extraction model, the initial feature fusion model and the initial feature decoding model are trained to obtain a data migration model.

[0123] In an embodiment of the present application, an evaluation value of the candidate data migration model is obtained based on the integrity quantification value, accuracy quantification value and migration efficiency quantification value between the converted data and the real migration data corresponding to the non-migrated data, and the initial feature extraction model, the initial feature fusion model and the initial feature decoding model are trained according to the evaluation value to ensure the accuracy, completeness and efficiency of data mapping and conversion of the data migration model in actual application scenarios.

[0124] In an exemplary embodiment, as shown in FIG7 , the method further includes the following steps:

[0125] S701, monitor the data migration status between the source system and the target system.

[0126] The data migration status between the source system and the target system includes two situations: successful migration and abnormal migration. The abnormal migration situation can be further divided into various abnormal situations such as format conversion abnormality and data compression abnormality.

[0127] After converting the configuration data into the target data, continue to monitor the status of the target data on the target system. If the data migration is successful, use the configuration data and the target data as a set of positive samples and fill them into the data set of the data migration model to further optimize the parameters of the data migration model.

[0128] S702: In case of target data migration exception, the target data is rolled back to the configuration data, and migration exception log information of the target data is collected.

[0129] If the target data migration is abnormal, the target data will be rolled back to the configuration data to avoid the propagation of erroneous data on the target system. At the same time, considering that there are many migration abnormalities, the migration abnormality log information of the target data can be further collected to determine the factors of the target data migration abnormality, such as format conversion abnormality, data compression abnormality, etc.

[0130] S703: Modify the model parameters in the data migration model based on the migration exception log information.

[0131] Migration exception log information includes configuration data, target data, migration exception type, exception degree, and data migration model actions, etc.

[0132] Optionally, the configuration data and the target data are used as a group of negative samples and filled into the data set of the data migration model. The data migration model is called to learn the negative sample information, adjust its own model parameters, and realize parameter correction.

[0133] Optionally, the migration exception log information is sent to a pre-trained lightweight fault-tolerant processing model, and the migration exception log information is parsed by the fault-tolerant processing model to obtain model parameters output by the fault-tolerant processing model. The model parameters output by the fault-tolerant processing model are then used to replace the current model parameters in the data migration model to achieve parameter correction of the data migration model.

[0134] Among them, the training process of the fault-tolerant processing model includes the following steps: ① Select a lightweight model architecture. According to the task requirements and data characteristics, select a suitable lightweight model architecture, such as MobileNet, ShuffleNet, etc., which has fewer parameters and computational complexity while ensuring certain performance. ② Data preprocessing and loading. Further preprocess the constructed fault-tolerant data set, such as standardization and normalization, to make it meet the input requirements of the fault-tolerant processing model, and use a data loader, such as PyTorch's DataLoader or TensorFlow's tf.data, to load the data into the training process in small batches. ③ Model initialization. According to the selected lightweight model, use random initialization or pre-trained model initialization parameters. If there is a relevant pre-trained model and the data is similar to the domain of the pre-trained model, the parameters of the pre-trained model can be used for initialization to speed up the training convergence. ④ Define the loss function and optimizer. According to the nature of the fault-tolerant processing task, select a suitable loss function, such as cross entropy loss, mean square error loss, etc. At the same time, select an optimizer suitable for lightweight models, such as stochastic gradient descent (SGD), Adagrad, Adadelta, etc., and set appropriate parameters such as learning rate and momentum. ⑤ Training cycle. Start the training cycle. In each training batch, input the data into the model, calculate the loss, calculate the gradient through back propagation, and use the optimizer to update the parameters of the model. At the same time, the intermediate results of the model can be saved at certain intervals to facilitate the subsequent selection of the optimal model. ⑥ Model evaluation and adjustment. After each training cycle, use the validation set to evaluate the model and calculate relevant evaluation indicators, such as accuracy and recall. According to the evaluation results, adjust hyperparameters such as learning rate and training batch size, or fine-tune the model architecture to obtain a fault-tolerant processing model.

[0135] In the embodiment of the present application, in the case of abnormal migration of target data, the target data is rolled back to the configuration data, and the migration abnormality log information of the target data is collected. Based on the migration abnormality log information, the model parameters in the data migration model are corrected, and the parameters and strategies of the model are adjusted according to the feedback signal using the reinforcement learning algorithm to optimize the performance of the data migration model, reduce the probability of errors, and improve the accuracy and efficiency of migration. In addition, as the data migration task continues and data accumulates, the performance of the model is continuously evaluated, and the model is further optimized and updated according to the new data and feedback information to adapt to the ever-changing business needs and data characteristics.

[0136] In one embodiment, as shown in FIG8 , taking the data migration model including a system mapping model, a feature extraction model, an attention mechanism model and a feature decoding model as an example, the data migration method includes the following steps:

[0137] S801: Train the initial data migration model according to the migration data set to obtain a data migration model.

[0138] S802, input the configuration data into the data migration model to obtain the target data.

[0139] S803: During the target data migration process, determine whether the target system reports an error.

[0140] S804: If an error is reported, the target data is rolled back to the last stable state, and migration exception information is collected.

[0141] The migration exception information is input into the data migration model to instruct the data migration model to optimize the model parameters by using reinforcement learning.

[0142] S805: If no error is reported, the target data is successfully migrated to the target system.

[0143] The training process of the data migration model, as shown in FIG9 , includes the following steps:

[0144] S901, collecting a first data set for migration of relevant system data.

[0145] S902: pre-train and fine-tune the initial system mapping model in the data migration model according to the first data set to obtain a system mapping model.

[0146] S903, collecting a second data set for target system data migration.

[0147] S904: Input the configuration data of the second data set into the system mapping model to obtain a mapping relationship.

[0148] S905, inputting the configuration data of the second data set into a feature extraction module to obtain encoding data features.

[0149] S906, the mapping relationship and the encoded data features are integrated through the attention mechanism model.

[0150] S907, decoding the fused features through a feature decoding model to obtain target data.

[0151] S908, collecting a third data set of target system data migration anomalies.

[0152] S909: Train the initial fault-tolerant processing model according to the third data set.

[0153] S910: Use the trained fault-tolerant processing model to score the migration exception log information of the data migration model, and optimize the parameters of the data migration model according to the score value.

[0154] In the embodiment of the present application, the attention mechanism is introduced and combined with the fault-tolerant model to the reward mechanism feedback method, so that the data migration model can better refer to the results of the system mapping model, and realize mutual promotion and optimization between models. Specifically, in the pre-training stage, the output results obtained by the system mapping model and the encoding features of the feature extraction model are fused using the attention mechanism, making full use of the historical cloud platform data resources, which can enhance the features extracted by the configuration conversion model, making the decoding process more referential to the results of the system mapping model, and improving the performance and adaptability of the model.

[0155] In another feasible scenario, the entire configuration data migration process is modeled as a reinforcement learning problem, and the data migration model can also be obtained through reinforcement learning training with large-scale unlabeled data. In this scenario, the feature extraction module, attention mechanism module, and feature decoding module are used as a configuration conversion model, and the system mapping model and configuration conversion model are used as an intelligent agent. Rewards or penalties are given according to the effect and goal of data migration. Through continuous learning and exploration, the entire data migration model can adaptively adjust its strategy to achieve better fusion and migration effects.

[0156] In the reinforcement learning process, the state space, action space, task objectives and reward functions are defined.

[0157] (1) State space.

[0158] The state space defines the environmental information that the data transfer model can perceive during the learning process.

[0159] For the system mapping model, the state space includes the following key elements: ① The feature vector of the source configuration data: it contains various attributes of the source configuration data, such as data type, data size, and data structure; ② The proportion and accuracy of completed mapping: it reflects the progress and quality of the current mapping work.

[0160] For example, assuming that the source configuration data is a set of database tables with different fields and data values, the status may include a description of the characteristics of each field (such as field name, data type, length), the ratio of the number of fields that have been successfully mapped to the total number of fields, and accuracy assessment indicators of the mapped fields (such as matching degree, consistency, etc.).

[0161] For the configuration conversion model, the state space includes the following key elements: ① Feature information extracted by the encoder: This is the key feature representation obtained after being processed by the encoder. ② Quality assessment of the intermediate conversion results: including the similarity with the target configuration, data integrity, etc.

[0162] Taking the conversion of cloud platform configuration data as an example, the state may include the numerical distribution of the feature vector output by the encoder and the degree of difference between the current conversion result and the configuration required by the target platform in key parameters.

[0163] (2) Action space.

[0164] The action space consists of the feasible actions that the data model can take in a specific state.

[0165] For the system mapping model, the action space includes: feature selection and matching strategy: deciding which features to select from the source configuration data to map with the target, and which matching algorithm to use; adjusting mapping parameters: for example, modifying the similarity threshold, adjusting the priority of mapping rules, etc.

[0166] For example, when processing database field mapping, the model can choose direct matching based on field names, or complex matching strategies based on data types and semantics, and can also dynamically adjust the matching similarity threshold to control the strictness of the mapping.

[0167] For the configuration conversion model, the action space includes: Select conversion rules and algorithms: select conversion methods suitable for current data from the predefined rule set, such as deciding whether to perform further optimization operations: such as data compression, format adjustment, etc. When converting the configuration data of the source cloud platform to the format of the target cloud platform, the model can choose direct numerical conversion rules, or more complex conversion rules based on logical judgment, and can also decide whether to compress the converted data to reduce storage space.

[0168] (3) Mission objectives:

[0169] The main task of the data conversion model is to achieve accurate and efficient configuration data migration, so that the source configuration data can be accurately converted into the format and parameters that meet the requirements of the target cloud platform. Ensure that the results of mapping and conversion are functionally equivalent to the source configuration data, that is, the same business logic and performance can be achieved on the target platform. At the same time, it can minimize manual intervention and realize the automated configuration data migration process.

[0170] The secondary task objectives of the data conversion model include: optimizing resource utilization: reducing the consumption of computing resources, memory, and storage during the mapping and conversion process; improving the generalization ability of the model: enabling it to adapt to configuration data migration tasks of different types and scales; enhancing the interpretability of the model: facilitating the understanding of the decision-making process and results of the model for debugging and optimization.

[0171] (4) Reward function.

[0172] The reward function is used to measure the quality of the actions taken by the model and provide feedback signals for the model's learning. For automatic mapping models and configuration conversion models, the reward function can comprehensively consider the accuracy of mapping and conversion, data integrity, and conversion efficiency.

[0173] After defining the above content, training the data migration model includes the following steps:

[0174] (1) Data processing.

[0175] The data set is processed by first adding various random noises to various types of data to perform data enhancement, and then normalizing the data to obtain the data for the training model.

[0176] (2) Select an optimizer.

[0177] Select an optimizer suitable for this model, such as Adam, SGD, etc. The optimizer is used to update the model parameters according to the gradient of the loss function to minimize the loss function value. According to the characteristics of the model and the scale of the training data, select a suitable optimizer and its hyperparameters such as learning rate and momentum.

[0178] (3) Initialization phase.

[0179] The parameters of the system mapping model and the configuration conversion model, including the weights and biases in the neural network, are randomly initialized. This operation aims to give the model a random starting state to avoid being restricted by specific initial conditions, thereby increasing the possibility of exploring the global optimal solution. At the same time, setting the initial strategy is also a key step. The setting of the strategy can be based on randomly selecting actions so that the model is not affected by preconceived ideas in the initial stage and fully explores various possible action choices. As another implementation method, the initial strategy can also be established based on prior knowledge. If there is existing experience or theoretical basis in the relevant field, it can provide the model with a certain guiding initial action direction.

[0180] (4) Exploration stage

[0181] In the exploration phase, the model selects actions and executes them based on the current state and the established initial strategy. In this phase, the model actively tries different actions and extensively explores the state space and action space. Through a large number of random attempts, the model is able to accumulate rich and diverse empirical data. This unbiased exploration helps the model fully understand the possible results of different actions in various states, thereby building a preliminary understanding of the problem and providing a broad data foundation for subsequent learning and optimization.

[0182] (5) Feedback and learning stage.

[0183] After the model executes an action, it immediately enters the feedback and learning phase. In this phase, the reward value is first calculated based on the execution result of the action. As a key feedback signal, the reward value quantitatively indicates the quality of the action. The reinforcement learning Q-learning algorithm is used to update the parameters and strategies of the model to maximize the future cumulative rewards. Q-learning can guide action selection by estimating the value (Q value) of the state-action pair to optimize the long-term reward accumulation. By learning and analyzing the reward signal, the model is guided to gradually adjust its decision-making strategy so that it can learn to choose better actions in the long-term interaction process.

[0184] (6) Model evaluation and tuning.

[0185] To ensure continuous improvement of model performance, it is crucial to regularly evaluate the model on the validation set. As a data set independent of the training set, the validation set can objectively reflect the generalization ability and performance of the model on new data. During the evaluation process, pay close attention to performance indicators such as accuracy, recall, and F1 value. Based on the evaluation results, make detailed adjustments to the model's hyperparameters (such as learning rate, number of network layers, etc.), weight coefficients, or strategies. If the model performance does not improve within a certain period of time, it is necessary to think deeply and take positive measures, such as redesigning the reward function to provide more accurate and effective feedback, adjusting the model structure (increasing or decreasing the number of neurons, changing the network connection method, etc.), or choosing a more advanced and appropriate reinforcement learning algorithm.

[0186] (7) Convergence and optimization phase.

[0187] When the model has been iterated many times and its performance has stabilized or reached the preset performance index after a certain number of iterations, the model can be considered to have converged. At the same time, the early stopping method can be used to prevent the model from overfitting.

[0188] Finally, a data transformation model including a system mapping model and a configuration transformation model is obtained.

[0189] In addition, the embodiment of the present application adds the new data obtained by subsequent monitoring of the fault-tolerant processing model to the state space and action space. These new features of the state space can provide more information about the current state of the system, enrich the description of the state, and enable the model to have a more comprehensive understanding of the system. For example, according to the situation reflected by the new data, it may be found that some actions that were not originally included in the action space are necessary, or the existing actions are adjusted and improved to more effectively respond to changes in the system state. In turn, it is possible to make more full use of the newly generated data, adapt to the ever-changing environment and needs, and maintain good performance and effects. In summary, the embodiment of the present application efficiently integrates the system mapping model, feature extraction module, attention mechanism model, feature decoding model and fault-tolerant processing model to achieve fully automated analysis, mapping, conversion and fault-tolerant processing, avoiding the complexity and errors caused by manual operation, making the migration process more efficient and predictable, and significantly improving the migration efficiency and accuracy.

[0190] In summary, the data migration method provided in the embodiment of the present application has shown significant beneficial effects in many aspects, and provides an efficient, reliable, economical and flexible solution for enterprises to migrate data between heterogeneous systems, which helps enterprises improve business continuity and market competitiveness.

[0191] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned 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 does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can 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 to be 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.

[0192] Based on the same inventive concept, the embodiment of the present application also provides a data migration device for implementing the data migration 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 one or more data migration device embodiments provided below can refer to the limitations on the data migration method above, and will not be repeated here.

[0193] In an exemplary embodiment, as shown in FIG10 , a data migration device is provided, including: a data acquisition module 1001, a data enhancement module 1002 and a data migration module 1003, wherein:

[0194] The data acquisition module 1001 is used to acquire configuration data to be migrated from the source system to the target system;

[0195] The data enhancement module 1002 is used to enhance the coded data features of the configuration data based on a preset data migration model, and decode the enhanced coded data features to obtain target data;

[0196] The data migration module 1003 is used to migrate the target data to the target system.

[0197] In an exemplary embodiment, the data enhancement module 1002 includes a data encoding unit and a data decoding unit, wherein:

[0198] A data encoding unit, used to extract a mapping relationship between a source system and a target system through a feature encoding model, and enhance the encoded data features based on the mapping relationship to obtain the enhanced encoded data features;

[0199] The data decoding unit is used to decode the enhanced encoded data features through a feature decoding model to obtain target data.

[0200] In an exemplary embodiment, the data encoding unit includes a relationship acquisition subunit, a feature dimension enhancement subunit and a feature fusion subunit, wherein:

[0201] The relationship acquisition subunit is used to extract mapping features of the configuration data through the system mapping model to obtain the mapping relationship between the source system and the target system;

[0202] A feature dimension-increasing subunit is used to perform feature dimension-increasing processing on the configuration data according to the feature extraction model to obtain the encoding data features;

[0203] The feature fusion subunit is used to input the mapping relationship and the encoded data features into the feature fusion model, and enhance the encoded data features based on the mapping relationship to obtain the enhanced encoded data features.

[0204] In an exemplary embodiment, the data migration device includes: a first training module and a second training module, wherein:

[0205] A first training module is used to train an initial system mapping model in the initial data migration model according to the first migration data set to obtain a candidate data migration model; the candidate data model includes a system mapping model;

[0206] The second training module is used to train the initial feature extraction model, the initial feature fusion model and the initial feature decoding model in the candidate data migration model according to the second migration data set to obtain the data migration model.

[0207] In an exemplary embodiment, the second training module includes a conversion unit, an evaluation unit and a training unit, wherein:

[0208] A conversion unit, used for inputting the unmigrated data in the historical configuration data set into the candidate data migration model to obtain conversion data of the unmigrated data;

[0209] An evaluation unit, configured to obtain an evaluation value of the candidate data migration model based on a completeness quantization value, an accuracy quantization value, and a migration efficiency quantization value between the converted data and the real migration data corresponding to the unmigrated data;

[0210] The training unit is used to train the initial feature extraction model, the initial feature fusion model and the initial feature decoding model based on the evaluation value to obtain a data migration model.

[0211] In an exemplary embodiment, the data migration device includes: a status monitoring module, a log collection module and a model correction module, wherein:

[0212] Status monitoring module, used to monitor the data migration status between the source system and the target system;

[0213] The log collection module is used to roll back the target data to the configuration data and collect the migration exception log information of the target data when the target data migration is abnormal;

[0214] The model correction module is used to correct the model parameters in the data migration model based on the migration exception log information.

[0215] Each module in the above data migration device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module 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 module.

[0216] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG11. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via 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, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store migration data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a data migration method is implemented.

[0217] Those skilled in the art will understand that the structure shown in FIG. 11 is merely a block diagram of a partial 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 shown in the figure, or combine certain components, or have a different arrangement of components.

[0218] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0219] In an exemplary embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0220] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0221] 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 regulations.

[0222] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method 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 the present application can include at least one of non-volatile memory 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), magnetic 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, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0223] 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 application.

[0224] The above-described 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 data migration method, characterized in that: The method comprises: Obtain configuration data from the source system to be migrated to the target system; Based on a preset data migration model, the encoded data features of the configuration data are enhanced, and the enhanced encoded data features are decoded to obtain target data; The target data is migrated to the target system.

2. The method according to claim 1, characterized in that The data migration model includes a feature encoding model and a feature decoding model; the encoding data features of the configuration data are enhanced based on the preset data migration model, and the enhanced encoding data features are decoded to obtain target data, including: Extracting a mapping relationship between the source system and the target system through the feature coding model, and enhancing the coding data features based on the mapping relationship to obtain enhanced coding data features; The enhanced encoded data features are decoded by the feature decoding model to obtain the target data.

3. The method according to claim 2, characterized in that The feature coding model includes a system mapping model, a feature extraction model and a feature fusion model; extracting the mapping relationship between the source system and the target system through the feature coding model, and enhancing the coded data features based on the mapping relationship to obtain the enhanced coded data features, including: Extracting mapping features from the configuration data using the system mapping model to obtain a mapping relationship between the source system and the target system; Performing feature dimension-upgrading processing on the configuration data according to the feature extraction model to obtain the encoding data features; The mapping relationship and the coding data feature are input into the feature fusion model, and the coding data feature is enhanced based on the mapping relationship to obtain the enhanced coding data feature.

4. The method according to any one of claims 1 to 3, characterized in that: The data migration model includes: a system mapping model, a feature extraction model, a feature fusion model and a feature decoding model; the training process of the data migration model includes: According to the first migration data set, an initial system mapping model in the initial data migration model is trained to obtain a candidate data migration model; the candidate data model includes the system mapping model; According to the second migration data set, the initial feature extraction model, the initial feature fusion model and the initial feature decoding model in the candidate data migration model are trained to obtain the data migration model.

5. The method according to claim 4, characterized in that The second migration data set includes a historical configuration data set; the initial feature extraction model, the initial feature fusion model and the initial feature decoding model in the candidate data migration model are trained according to the second migration data set to obtain the data migration model, including: Inputting the unmigrated data in the historical configuration data set into the candidate data migration model to obtain conversion data of the unmigrated data; Obtaining an evaluation value of the candidate data migration model based on a completeness quantization value, an accuracy quantization value, and a migration efficiency quantization value between the converted data and the real migration data corresponding to the unmigrated data; Based on the evaluation value, the initial feature extraction model, the initial feature fusion model and the initial feature decoding model are trained to obtain the data migration model.

6. The method according to claim 1, characterized in that The method further comprises: Monitoring the data migration status between the source system and the target system; In the event of target data migration exception, the target data is rolled back to the configuration data, and migration exception log information of the target data is collected; Based on the migration exception log information, model parameters in the data migration model are modified.

7. A data migration device, characterized in that: The device comprises: A data acquisition module is used to acquire configuration data to be migrated from a source system to a target system; A data enhancement module, used to enhance the coded data features of the configuration data based on a preset data migration model, and decode the enhanced coded data features to obtain target data; A data migration module is used to migrate the target data to the target system.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. 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 6 are implemented.

10. 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 6 are implemented.

Images