Intelligent data protection method and system

Abstract

The present invention relates to the field of electronic digital data disaster recovery processing technology, and in particular to an intelligent data protection method and system, which first predicts the predicted access frequency of each target data through data attributes and access frequency records, and then further combines the predicted access frequency, preset priority and data volume to obtain the recovery priority of each target data, and then uses the recovery priority of the target data to update the data in the cache data center, so that the cache data center will dynamically cache the latest and most important data. In this way, by utilizing the characteristic that the data recovery speed of the cache data center is higher than that of the disaster recovery backup data center, important data can be quickly restored through the cache data center in priority when a disaster occurs, so that basic or popular business functions can be quickly restored, reducing the losses caused to the enterprise, and perfectly solving the problem that business functions cannot be quickly restored due to the slow data recovery speed in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic digital data disaster recovery processing, and in particular to an intelligent data protection method and system. Background Art

[0002] Data disaster recovery refers to the replication and storage of data in multiple geographical locations or facilities to prevent data loss or unavailability caused by natural disasters, technical failures or human errors. Its core goal is to ensure that critical business data can be quickly restored and business continuity can be maintained when a disaster occurs. Data disaster recovery usually involves the development of detailed disaster recovery plans and strategies, including regular data backup, establishment of backup data centers or cloud services, and implementation of disaster recovery testing. Through these measures, organizations can reduce the risk of data loss and improve system reliability and stability, thereby protecting enterprises from the impact of disasters on data integrity and availability.

[0003] Data recovery in data disaster recovery is a key part of data disaster recovery technology. It refers to recovering data from backup or off-site replicated data sources to resume normal business operations after a disaster event or in a scheduled operation. This process involves identifying and assessing the scope of data loss, and then operating according to a pre-established recovery strategy, such as obtaining data from a remote backup facility or resynchronizing cloud data. Data recovery is not just a simple copy of data files, but also includes verifying the integrity of the data, testing whether the recovered data can meet business needs, and ultimately restoring business operations to normal. Through effective data recovery measures, enterprises can respond quickly to disaster events, minimize the impact of data loss, and ensure business continuity and reliability.

[0004] However, in reality, data is often massive, which means that when performing data recovery, it is necessary to deal with a large amount of data and complex system architecture. The huge volume of data may cause the recovery process to take a long time, and puts extremely high demands on network bandwidth and storage resources, increasing the complexity and cost of recovery operations. In this case, the speed of data recovery in existing technologies is difficult to achieve the expected effect. Even with backup and disaster recovery plans, the recovery of large-scale data may still take hours or even days, which is a severe challenge for enterprises that need to quickly restore business functions. Therefore, how to improve the speed and efficiency of data recovery so that enterprises can quickly restore business functions has become one of the problems that need to be solved in the field of data disaster recovery. Summary of the invention

[0005] Therefore, the present invention provides an intelligent data protection method and system to solve the problem in the prior art that business functions cannot be quickly restored due to the slow data recovery speed.

[0006] The present invention provides an intelligent data protection method, comprising:

[0007] Obtain data attributes and access frequency records of various target data in the main data center;

[0008] Based on data attributes and access frequency records, the predicted access frequency of each target data is obtained;

[0009] Obtain the preset priority and data volume for each target data;

[0010] According to the predicted access frequency, preset priority and data volume, the recovery priority of each target data is obtained;

[0011] The data in the cache data center is updated based on the recovery priority of multiple target data, wherein the cache data center is used to back up the data of the primary data center, the capacity of the cache data center is smaller than that of the disaster recovery backup data center, and the data recovery speed of the cache data center is higher than that of the disaster recovery backup data center.

[0012] The present invention also provides a preferred solution: based on data attributes and access frequency records, the predicted access frequency of each target data is obtained, including:

[0013] Based on the data attributes and access frequency records, an input vector sequence is established, the input vector sequence includes multiple input vectors, and each element of the input vector corresponds to the access frequency of the target data within a period of time and multiple data attributes of the target data;

[0014] Inputting the input vector sequence into a preset recurrent neural network model to obtain an output result output by the preset recurrent neural network model;

[0015] According to the output results, the predicted access frequency is obtained.

[0016] The present invention also provides a preferred solution: the preset recurrent neural network model includes a first recurrent network and a second recurrent network connected in sequence, wherein the first recurrent network includes a first input layer, at least one first recurrent layer and a first output layer connected in sequence, the nodes in the first input layer correspond to each element in the input vector respectively, each node of the first input layer is respectively connected to each node in the adjacent first recurrent layer, the first input layer is used to input an input vector respectively in each cycle, the number of cycles of the first recurrent layer is equal to the number of input vectors in the input vector sequence, each node of the first output layer is respectively connected to each node in the adjacent first recurrent layer, the first output layer is used to output a context vector after the last cycle, and the number of output nodes of the first output layer is equal to the number of nodes in the context vector. The number of elements is the same; the second recurrent network includes a second input layer, at least one second recurrent layer and a second output layer connected in sequence, the nodes in the second input layer correspond to each element in the context vector respectively, each node of the second input layer is connected to each node in the adjacent second recurrent layer respectively, the second input layer is used to input the context vector only in the first loop, the second recurrent layer is used to output in each loop according to the input of the second input layer and the state of the second recurrent layer in the previous loop, the number of nodes in the second output layer is one, the nodes of the second output layer are connected to each node in the adjacent second recurrent layer, the second output layer is used to output an access frequency prediction value in each loop, and multiple access frequency prediction values ​​constitute the output result of the preset recurrent neural network model.

[0017] The present invention also provides a preferred solution: the element corresponding to the access frequency of the target data in the input vector is the first; in the first cyclic network, the first node in the first input layer is used to input the element corresponding to the access frequency of the target data in the input vector, and the first cyclic layer adjacent to the first input layer is used to output in each loop according to the input of the first input layer in this loop, the state of the second cyclic layer in the previous loop, and the input of the first node in the first input layer in the previous loop.

[0018] The present invention also provides a preferred solution: in the first recurrent network, the expression corresponding to the first recurrent layer adjacent to the first input layer is:

[0019]

[0020] Where t represents the number of cycles, h t represents the output of the first loop layer adjacent to the first input layer at the tth loop, f() represents the activation function, X t represents the input vector of the first input layer at the tth cycle, W represents the weight matrix between the first input layer and the first cycle layer, and h t-1represents the output of the first recurrent layer adjacent to the first input layer at the t-1th cycle, U represents the weight matrix of the first recurrent layer adjacent to the first input layer between two adjacent cycles, represents the input of the first node in the first input layer in the t-1th cycle, V represents the weight vector between the first node in the first loop layer in this cycle and the first node in the first input layer in the previous cycle, and b is the bias vector of the first loop layer adjacent to the first input layer.

[0021] The present invention also provides a preferred solution: according to the output result, the predicted access frequency is obtained, including:

[0022] The maximum value is selected from the multiple access frequency prediction values ​​as the predicted access frequency.

[0023] The present invention also provides a preferred solution: according to the predicted access frequency, the preset priority and the data volume, the recovery priority of each target data is obtained, including:

[0024] The recovery priority of the target data is calculated by the following formula:

[0025] P=w1×P p +w2×f p -w3×e a×N ;

[0026] Among them, P represents the recovery priority of the target data, P p Indicates the preset priority of the target data, f p represents the predicted access frequency of the target data, e is a natural constant, N represents the data volume of the target data, a is a preset unit adjustment coefficient, and w1, w2 and w3 are different preset weights respectively.

[0027] The present invention also provides a preferred solution: the cache data center includes multiple backup data, each backup data corresponds to one or more target data of the main data center, and each backup data also corresponds to a remaining storage time value. The remaining storage time value is used to characterize the remaining residence time of the corresponding backup data in the cache data center. The remaining storage time value gradually decreases with the passage of time. When the remaining storage time value is zero, the cache data center deletes the backup data corresponding to the remaining storage time value.

[0028] The present invention also provides a preferred solution: updating the data in the cache data center based on the recovery priority of multiple target data, including:

[0029] Based on the recovery priorities of multiple target data, select the target data with the highest recovery priority as the target data to be backed up, and obtain the recovery priority corresponding to the target data to be backed up;

[0030] Based on the recovery priority corresponding to the target data to be backed up, an initial remaining storage time value of the target data to be backed up is obtained, wherein the initial remaining storage time value is proportional to the recovery priority;

[0031] Check whether there is backup data of the same type as the target data to be backed up in the cache data center;

[0032] If there is backup data of the same type as the target data to be backed up in the cache data center, the backup data of the same type is updated based on the target data to be backed up, and the remaining storage time value corresponding to the backup data of the same type is updated based on the initial remaining storage time;

[0033] If there is no backup data of the same type as the target data to be backed up in the cache data center, check whether the remaining storage space of the cache data center exceeds the data volume of the target data to be backed up;

[0034] If the remaining storage space of the cache data center exceeds the data volume of the target data to be backed up, the target data to be backed up and the corresponding remaining storage time value are saved in the cache data center;

[0035] If the remaining storage space of the cache data center does not exceed the data volume of the target data to be backed up, the backup data with the minimum remaining storage time value in the cache data center is deleted, and the step of checking whether the remaining storage space of the cache data center exceeds the data volume of the target data to be backed up is re-executed.

[0036] The present invention also provides an intelligent data protection system, including a main data center, a disaster recovery backup data center and a cache data center, and also includes:

[0037] Data monitoring module, used to obtain data attributes and access frequency records of various target data in the main data center;

[0038] A frequency prediction module is used to obtain the predicted access frequency of each target data based on data attributes and access frequency records;

[0039] A data preparation module is used to obtain the preset priority and data volume of each target data;

[0040] A priority calculation module is used to obtain the recovery priority of each target data according to the predicted access frequency, preset priority and data volume;

[0041] The data update module is used to update the data in the cache data center based on the recovery priority of multiple target data, wherein the cache data center is used to back up the data of the main data center, the capacity of the cache data center is smaller than the disaster recovery backup data center, and the data recovery speed of the cache data center is higher than that of the disaster recovery backup data center.

[0042] The beneficial effects of adopting the above embodiment are:

[0043] The present invention provides an intelligent data protection method and system, which caches some important data by setting up a cache data center between a disaster recovery backup data center and a main data center. Specifically, it first predicts the predicted access frequency of each target data through data attributes and access frequency records, and then further combines the predicted access frequency, preset priority and data volume to obtain the recovery priority of each target data. Then, the recovery priority of the target data is used to update the data in the cache data center, so that the cache data center will dynamically cache the latest and most important data. In this way, by utilizing the characteristic that the data recovery speed of the cache data center is higher than that of the disaster recovery backup data center, when a disaster occurs, important data can be quickly restored through the cache data center in priority, so that basic or popular business functions can be quickly restored, reducing the losses caused to the enterprise, and perfectly solving the problem that business functions cannot be quickly restored due to the slow data recovery speed in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A method flow chart of an embodiment of an intelligent data protection method provided by the present invention;

[0045] Figure 2 for Figure 1 The method flow chart of step S102 in FIG.

[0046] Figure 3 A schematic diagram of the expansion of the preset recurrent neural network model in the present invention in the time dimension;

[0047] Figure 4 A schematic diagram of the structure of an embodiment of the intelligent data protection system provided by the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] Combination Figure 1 As shown, a specific embodiment of the present invention discloses an intelligent data protection method, including:

[0050] S101, obtaining data attributes and access frequency records of various target data in the main data center;

[0051] S102, obtaining a predicted access frequency of each target data based on data attributes and access frequency records;

[0052] S103, obtaining a preset priority and data volume for each target data;

[0053] S104, obtaining a restoration priority of each target data according to the predicted access frequency, the preset priority and the data volume;

[0054] S105. Update the data in the cache data center based on the recovery priority of multiple target data, wherein the cache data center is used to back up the data of the primary data center, the capacity of the cache data center is smaller than that of the disaster recovery backup data center, and the data recovery speed of the cache data center is higher than that of the disaster recovery backup data center.

[0055] It should be noted that the main data center in the above process refers to the servers and other equipment that are mainly used to provide services in daily life, while the disaster recovery backup data center refers to the servers and other equipment that are mainly used for data disaster recovery and backup. The cache data center is an additional storage device set up in the main data center and the disaster recovery backup data center on the basis of existing technologies. It is the transfer of the idea of ​​"trading space for time" in the computer internal storage system at the macro server level. The cache data center is mainly used to store more important data, so its capacity can be smaller than the disaster recovery backup data center. However, because the cache data center is responsible for rapid data recovery in emergency situations, the recovery speed of the cache data center needs to be faster than that of the disaster recovery backup data center. This can be achieved by using better equipment hardware, using a network with larger bandwidth and less load pressure, or locating the cache data center closer to the main data center.

[0056] In addition, the access frequency record in the above process refers to the record of the access frequency of the target data in the past period of time, such as the number of times a certain target data was accessed every day or every hour in the past.

[0057] Preset priority refers to an artificially set index that represents the importance of a certain target data. For example, in web pages, the priority of data is usually set according to its direct impact on user experience and page performance. High-priority data includes static resources such as CSS style sheets, JavaScript files, and images and fonts in the page, which directly affect page loading speed and user experience. Database query results and data of key pages or functions, such as homepage content or shopping cart information, also have high priority because they are the most concerned parts when users visit for the first time. Lower-priority data includes historical data or logs, such as access logs and operation history records, which have less impact on user experience. Data of secondary pages, such as infrequently visited function pages or infrequently updated content pages, also fall into the category of low priority. Personalized data of a few users, such as settings or content of a specific user, and temporarily generated data such as real-time statistics, have relatively low priority because they have limited impact on the overall user group or have a short life cycle.

[0058] Data volume refers to the amount of memory occupied by the data.

[0059] Data attributes refer to the attribute information related to the data, such as media type, file extension, data format, data logical level, and any other attributes that can be described artificially.

[0060] Compared with the prior art, the present invention first predicts the predicted access frequency of each target data through data attributes and access frequency records, and then further combines the predicted access frequency, preset priority and data volume to obtain the recovery priority of each target data, and then uses the recovery priority of the target data to update the data in the cache data center, so that the cache data center will dynamically cache the latest and most important data. In this way, by utilizing the characteristic that the data recovery speed of the cache data center is higher than that of the disaster recovery backup data center, when a disaster occurs, important data can be quickly restored through the cache data center in priority, so that basic or popular business functions can be quickly restored, reducing the losses caused to the enterprise, and perfectly solving the problem of the inability to quickly restore business functions due to the slow data recovery speed in the prior art.

[0061] Further, combined with Figure 2 As shown, in a preferred embodiment, the above step S102, based on the data attributes and the access frequency records, obtains the predicted access frequency of each target data, specifically including:

[0062] S201, based on data attributes and access frequency records, establish an input vector sequence, the input vector sequence includes multiple input vectors, each element of the input vector corresponds to the access frequency of target data within a period of time and multiple data attributes of the target data;

[0063] S202, inputting the input vector sequence into a preset recurrent neural network model to obtain an output result output by the preset recurrent neural network model;

[0064] S203: Obtain the predicted access frequency according to the output result.

[0065] Since data disaster recovery is mainly to defend against future risks, it is very necessary to predict the future access frequency and evaluate the priority through the predicted access frequency. In practice, the access frequency record can be used to directly predict the future access frequency. For example, the future prediction can be made by fitting the curve composed of the access frequency record. However, in practice, there are various types of data, and the usage functions of different data are different. The access frequency of different data does not change according to a simple linear relationship. For example, for verification data such as user information and user passwords, the access frequency may be relatively regular, while for data such as video resources and image resources, the access frequency of such data may not be so regular due to the influence of the trend of popular content on the Internet. Therefore, the error rate of prediction directly through access frequency records will be high.

[0066] Therefore, this embodiment further proposes a method of prediction through a neural network, which is well-known for its nonlinear modeling ability and can learn complex data relationships and patterns from a large amount of historical data. This enables the neural network to combine multiple factors (i.e., multiple data attributes in this embodiment) to more accurately and reasonably predict the future access frequency of different data types, thereby ensuring that the recovery priority of subsequent calculations is more reasonable, making the data stored in the cache data center more practical.

[0067] Specifically, in a preferred embodiment, the preset recurrent neural network model includes a first recurrent network and a second recurrent network connected in sequence, wherein the first recurrent network includes a first input layer, at least one first recurrent layer and a first output layer connected in sequence, the nodes in the first input layer correspond to each element in the input vector respectively, each node of the first input layer is respectively connected to each node in the adjacent first recurrent layer, the first input layer is used to input an input vector respectively in each cycle, the number of cycles of the first recurrent layer is equal to the number of input vectors in the input vector sequence, each node of the first output layer is respectively connected to each node in the adjacent first recurrent layer, the first output layer is used to output a context vector after the last cycle, and the number of output nodes of the first output layer and the context vector The second recurrent network includes a second input layer, at least one second recurrent layer and a second output layer connected in sequence, the nodes in the second input layer correspond to each element in the context vector respectively, each node of the second input layer is connected to each node in the adjacent second recurrent layer respectively, the second input layer is used to input the context vector only in the first loop, the second recurrent layer is used to output in each loop according to the input of the second input layer and the state of the second recurrent layer in the previous loop, the number of nodes in the second output layer is one, the nodes of the second output layer are connected to each node in the adjacent second recurrent layer, the second output layer is used to output an access frequency prediction value in each loop, and multiple access frequency prediction values ​​constitute the output result of the preset recurrent neural network model.

[0068] The significance of the above neural network model is that the input sequence and output sequence lengths of the entire preset neural network model can be inconsistent, that is, the number of input vectors may not be equal to the number of output access frequency prediction values. For example, the access frequency records of the past 10 days can be used to predict the access frequency of the next day, and the access frequency of any number of days in the next 3, 6, 10, etc. can also be predicted. This greatly ensures the flexibility of the model, allowing it to be flexibly trained and deployed according to actual conditions.

[0069] Furthermore, in a preferred embodiment, the element corresponding to the access frequency of the target data in the input vector is the first. Correspondingly, in the first loop network in this embodiment, the first node in the first input layer is used to input the element corresponding to the access frequency of the target data in the input vector, and the first loop layer adjacent to the first input layer is used to output in each loop according to the input of the first input layer in this loop, the state of the second loop layer in the previous loop, and the input of the first node in the first input layer in the previous loop.

[0070] Correspondingly, in the first recurrent network, the expression corresponding to the first recurrent layer adjacent to the first input layer is:

[0071]

[0072] Where t represents the number of cycles, h t represents the output of the first loop layer adjacent to the first input layer at the tth loop, f() represents the activation function, X t represents the input vector of the first input layer at the tth cycle, W represents the weight matrix between the first input layer and the first cycle layer, and h t-1 represents the output of the first recurrent layer adjacent to the first input layer at the t-1th cycle, U represents the weight matrix of the first recurrent layer adjacent to the first input layer between two adjacent cycles, represents the input of the first node in the first input layer in the t-1th cycle, V represents the weight vector between the first node in the first loop layer in this cycle and the first node in the first input layer in the previous cycle, and b is the bias vector of the first loop layer adjacent to the first input layer.

[0073] The above design is an improvement on the existing recurrent neural network, because the first element in each input vector corresponds to the access frequency, and the other elements correspond to data attributes. For the same type of data, the data attributes are unchanged, that is, in the input vector sequence, only the first element in each input vector will change. Therefore, in this embodiment, the first recurrent layer is designed to consider not only the current input and the previous state of the recurrent layer in each cycle, but also the first element of the previous input, which is equivalent to making the first recurrent network consider the current input vector and the previous input vector at the same time in each prediction, greatly increasing the accuracy of the prediction.

[0074] In reality, because the data attributes of the current input vector and the previous input vector are repeated, and the current input also has the attribute information, the data actually input to this loop layer is only the first element of the previous input vector. The actual improvement to the first loop network structure is only the first element connecting the current loop layer and the previous loop layer, which is reflected in the above formula as V is a vector rather than a matrix. In general, the above design greatly improves the accuracy of the entire preset neural network (especially the first loop network) through minor changes while minimizing the increase in the amount of data in the preset neural network. Figure 3 This is a schematic diagram of the temporal structure expansion of the preset recurrent neural network model in this embodiment. For the convenience of display (not representing the actual structure), the number of nodes in each neural network layer in the figure is relatively small, the number of cycles is also relatively small, and the first cycle layer and the second cycle layer are both one layer.

[0075] Further, in a preferred embodiment, the above step S203, obtaining the predicted access frequency according to the output result, specifically includes:

[0076] The maximum value is selected from the multiple access frequency prediction values ​​as the predicted access frequency.

[0077] The preset recurrent neural network can output multiple access frequency prediction values, and this embodiment uses the maximum value as the predicted access frequency to ensure that potential important data is backed up by the cache data center, achieving the effect of data fault tolerance. At the same time, the above design can also reduce the amount of calculation and improve the execution speed of the entire method.

[0078] Further, in a preferred embodiment, the above step S104, based on the predicted access frequency, the preset priority and the data volume, obtains the restoration priority of each target data, specifically including:

[0079] The recovery priority of the target data is calculated by the following formula:

[0080] P=w1×P p +w2×f p -w3×e a×N ;

[0081] Among them, P represents the recovery priority of the target data, P p Indicates the preset priority of the target data, f p represents the predicted access frequency of the target data, e is a natural constant, N represents the data volume of the target data, a is a preset unit adjustment coefficient, and w1, w2 and w3 are different preset weights respectively.

[0082] The significance of the above formula is that a reasonable recovery priority value is obtained by weighted summation from three perspectives: preset priority, predicted access frequency, and data volume, to ensure the rationality of data backup. It is worth noting that in the above formula, only the item corresponding to the data volume is calculated using an exponential function. In practice, the amount of data affects the data utilization rate of the cache data center, and if the amount of data is too large, the cache data center only stores this type of data, then the entire solution will lose its original meaning. Therefore, in this embodiment, an exponential function is used to increase the impact of the data volume. When the data volume is too large, the third item in the above formula will increase rapidly, thereby ensuring that the data will not be written and monopolize the cache data center to ensure the recoverability of other data.

[0083] It is understandable that, in practice, depending on the specific situation, the restoration priority may also be calculated based on multiple access frequency prediction values, for example, by calculating an average value.

[0084] Furthermore, in a preferred embodiment, the cache data center includes multiple backup data, each backup data corresponds to one or more target data of the main data center, and each backup data also corresponds to a remaining retention time value, which is used to characterize the remaining residence time of the corresponding backup data in the cache data center. The remaining retention time value gradually decreases over time. When the remaining retention time value is zero, the cache data center deletes the backup data corresponding to the remaining retention time value.

[0085] Correspondingly, the above step S105, updating the data in the cache data center based on the recovery priority of the multiple target data, specifically includes:

[0086] Based on the recovery priorities of multiple target data, select the target data with the highest recovery priority as the target data to be backed up, and obtain the recovery priority corresponding to the target data to be backed up;

[0087] Based on the recovery priority corresponding to the target data to be backed up, an initial remaining storage time value of the target data to be backed up is obtained, wherein the initial remaining storage time value is proportional to the recovery priority;

[0088] Check whether there is backup data of the same type as the target data to be backed up in the cache data center;

[0089] If there is backup data of the same type as the target data to be backed up in the cache data center, the backup data of the same type is updated based on the target data to be backed up, and the remaining storage time value corresponding to the backup data of the same type is updated based on the initial remaining storage time;

[0090] If there is no backup data of the same type as the target data to be backed up in the cache data center, check whether the remaining storage space of the cache data center exceeds the data volume of the target data to be backed up;

[0091] If the remaining storage space of the cache data center exceeds the data volume of the target data to be backed up, the target data to be backed up and the corresponding remaining storage time value are saved in the cache data center;

[0092] If the remaining storage space of the cache data center does not exceed the data volume of the target data to be backed up, the backup data with the minimum remaining storage time value in the cache data center is deleted, and the step of checking whether the remaining storage space of the cache data center exceeds the data volume of the target data to be backed up is re-executed.

[0093] The above design ensures the replaceability of data in the cache data center, while ensuring the real-time nature of the backed-up data. If a category of target data obtains a high recovery priority in multiple evaluations, it indicates that the target data of this category is indeed important data. In this case, this type of data will always be backed up in the cache data center with the latest version, and the higher the recovery priority, the longer the remaining storage time value. Conversely, if a type of target data only obtains a higher predicted priority in a certain evaluation, then this type of data will only be retained in the cache data center for a period of time to leave room for the truly important data to be backed up. This ensures the effectiveness of the actual implementation of this embodiment.

[0094] Combination Figure 4 As shown, the present invention also provides an intelligent data protection system, including a main data center, a disaster recovery backup data center and a cache data center, and the system also includes:

[0095] Data monitoring module, used to obtain data attributes and access frequency records of various target data in the main data center;

[0096] A frequency prediction module is used to obtain the predicted access frequency of each target data based on data attributes and access frequency records;

[0097] A data preparation module is used to obtain the preset priority and data volume of each target data;

[0098] A priority calculation module is used to obtain the recovery priority of each target data according to the predicted access frequency, preset priority and data volume;

[0099] The data update module is used to update the data in the cache data center based on the recovery priority of multiple target data, wherein the cache data center is used to back up the data of the main data center, the capacity of the cache data center is smaller than the disaster recovery backup data center, and the data recovery speed of the cache data center is higher than that of the disaster recovery backup data center.

[0100] It should be noted here that the corresponding system provided in the above embodiments can implement the technical solutions described in the above method embodiments. The specific implementation principles of the above modules or units can be found in the corresponding contents in the above method embodiments, which will not be repeated here.

[0101] The present invention provides an intelligent data protection method and system, which caches some important data by setting up a cache data center between a disaster recovery backup data center and a main data center. Specifically, it first predicts the predicted access frequency of each target data through data attributes and access frequency records, and then further combines the predicted access frequency, preset priority and data volume to obtain the recovery priority of each target data. Then, the recovery priority of the target data is used to update the data in the cache data center, so that the cache data center will dynamically cache the latest and most important data. In this way, by utilizing the characteristic that the data recovery speed of the cache data center is higher than that of the disaster recovery backup data center, when a disaster occurs, important data can be quickly restored through the cache data center in priority, so that basic or popular business functions can be quickly restored, reducing the losses caused to the enterprise, and perfectly solving the problem that business functions cannot be quickly restored due to the slow data recovery speed in the prior art.

[0102] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0103] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent data protection method, characterized in that: include: Obtain data attributes and access frequency records of various target data in the main data center; Based on data attributes and access frequency records, the predicted access frequency of each target data is obtained; Obtain the preset priority and data volume for each target data; According to the predicted access frequency, preset priority and data volume, the recovery priority of each target data is obtained; Based on the recovery priority of multiple target data, the data in the cache data center is updated, wherein the cache data center is used to back up the data of the primary data center, the capacity of the cache data center is smaller than that of the disaster recovery backup data center, and the data recovery speed of the cache data center is higher than that of the disaster recovery backup data center; Among them, based on data attributes and access frequency records, the predicted access frequency of each target data is obtained, including: Based on the data attributes and access frequency records, an input vector sequence is established, the input vector sequence includes multiple input vectors, and each element of the input vector corresponds to the access frequency of the target data within a period of time and multiple data attributes of the target data; Inputting the input vector sequence into a preset recurrent neural network model to obtain an output result output by the preset recurrent neural network model; According to the output results, the predicted access frequency is obtained; Among them, the preset recurrent neural network model includes a first recurrent network and a second recurrent network connected in sequence, wherein the first recurrent network includes a first input layer, at least one first recurrent layer and a first output layer connected in sequence, the nodes in the first input layer correspond to each element in the input vector respectively, each node of the first input layer is respectively connected to each node in the adjacent first recurrent layer, the first input layer is used to input an input vector respectively in each cycle, the number of cycles of the first recurrent layer is equal to the number of input vectors in the input vector sequence, each node of the first output layer is respectively connected to each node in the adjacent first recurrent layer, the first output layer is used to output a context vector after the last cycle, and the number of output nodes of the first output layer is equal to the number of elements in the context vector The same; the second recurrent network includes a second input layer, at least one second recurrent layer and a second output layer connected in sequence, the nodes in the second input layer correspond to each element in the context vector respectively, each node of the second input layer is connected to each node in the adjacent second recurrent layer respectively, the second input layer is used to input the context vector only in the first loop, the second recurrent layer is used to output in each loop according to the input of the second input layer and the state of the second recurrent layer in the previous loop, the number of nodes in the second output layer is one, the nodes of the second output layer are connected to each node in the adjacent second recurrent layer, the second output layer is used to output an access frequency prediction value in each loop, and multiple access frequency prediction values ​​constitute the output result of the preset recurrent neural network model output; Among them, according to the output results, the predicted access frequency is obtained, including: Selecting a maximum value from multiple access frequency prediction values ​​as the predicted access frequency; Among them, according to the predicted access frequency, preset priority and data volume, the recovery priority of each target data is obtained, including: The recovery priority of the target data is calculated by the following formula: ； in, Indicates the recovery priority of the target data. Indicates the preset priority of the target data. represents the predicted access frequency of the target data, is a natural constant, Indicates the amount of target data, is the preset unit adjustment factor, , and They have different preset weights.

2. The intelligent data protection method according to claim 1, characterized in that: The element corresponding to the access frequency of the target data in the input vector is the first one; in the first loop network, the first node in the first input layer is used to input the element corresponding to the access frequency of the target data in the input vector, and the first loop layer adjacent to the first input layer is used to output in each loop according to the input of the first input layer in this loop, the state of the second loop layer in the previous loop, and the input of the first node in the first input layer in the previous loop.

3. The intelligent data protection method according to claim 2, characterized in that: In the first recurrent network, the expression corresponding to the first recurrent layer adjacent to the first input layer is: ； in, Indicates the number of cycles, The first recurrent layer adjacent to the first input layer is The output of the next cycle is represents the activation function, Indicates that the first input layer is The input vector for the next loop, represents the weight matrix between the first input layer and the first recurrent layer, The first recurrent layer adjacent to the first input layer is The output of the next cycle is Represents the weight matrix between two adjacent cycles of the first cycle layer adjacent to the first input layer, Indicates The input of the first node in the first input layer in the second cycle, represents the weight vector between the first node in the first loop layer in this loop and the first input layer in the previous loop, is the bias vector of the first recurrent layer adjacent to the first input layer.

4. The intelligent data protection method according to claim 1, characterized in that: The cache data center includes multiple backup data, each of which corresponds to one of multiple target data in the main data center. Each of the backup data also corresponds to a remaining storage time value, which is used to characterize the remaining residence time of the corresponding backup data in the cache data center. The remaining storage time value gradually decreases over time. When the remaining storage time value is zero, the cache data center deletes the backup data corresponding to the remaining storage time value.

5. The intelligent data protection method according to claim 4, characterized in that: Updates data in the cache data center based on the recovery priority of various target data, including: Based on the recovery priorities of multiple target data, select the target data with the highest recovery priority as the target data to be backed up, and obtain the recovery priority corresponding to the target data to be backed up; Based on the recovery priority corresponding to the target data to be backed up, an initial remaining storage time value of the target data to be backed up is obtained, wherein the initial remaining storage time value is proportional to the recovery priority; Check whether there is backup data of the same type as the target data to be backed up in the cache data center; If there is backup data of the same type as the target data to be backed up in the cache data center, the backup data of the same type is updated based on the target data to be backed up, and the remaining storage time value corresponding to the backup data of the same type is updated based on the initial remaining storage time; If there is no backup data of the same type as the target data to be backed up in the cache data center, check whether the remaining storage space of the cache data center exceeds the data volume of the target data to be backed up; If the remaining storage space of the cache data center exceeds the data volume of the target data to be backed up, the target data to be backed up and the corresponding remaining storage time value are saved in the cache data center; If the remaining storage space of the cache data center does not exceed the data volume of the target data to be backed up, the backup data with the minimum remaining storage time value in the cache data center is deleted, and the step of checking whether the remaining storage space of the cache data center exceeds the data volume of the target data to be backed up is re-executed.

6. An intelligent data protection system, characterized in that: It includes the main data center, disaster recovery backup data center and cache data center, as well as: Data monitoring module, used to obtain data attributes and access frequency records of various target data in the main data center; A frequency prediction module is used to obtain the predicted access frequency of each target data based on data attributes and access frequency records; A data preparation module is used to obtain the preset priority and data volume of each target data; A priority calculation module is used to obtain the recovery priority of each target data according to the predicted access frequency, preset priority and data volume; A data update module is used to update the data in the cache data center based on the recovery priority of multiple target data, wherein the cache data center is used to back up the data of the main data center, the capacity of the cache data center is smaller than that of the disaster recovery backup data center, and the data recovery speed of the cache data center is higher than that of the disaster recovery backup data center; Among them, based on data attributes and access frequency records, the predicted access frequency of each target data is obtained, including: Based on the data attributes and access frequency records, an input vector sequence is established, the input vector sequence includes multiple input vectors, and each element of the input vector corresponds to the access frequency of the target data within a period of time and multiple data attributes of the target data; Inputting the input vector sequence into a preset recurrent neural network model to obtain an output result output by the preset recurrent neural network model; According to the output results, the predicted access frequency is obtained; Among them, the preset recurrent neural network model includes a first recurrent network and a second recurrent network connected in sequence, wherein the first recurrent network includes a first input layer, at least one first recurrent layer and a first output layer connected in sequence, the nodes in the first input layer correspond to each element in the input vector respectively, each node of the first input layer is respectively connected to each node in the adjacent first recurrent layer, the first input layer is used to input an input vector respectively in each cycle, the number of cycles of the first recurrent layer is equal to the number of input vectors in the input vector sequence, each node of the first output layer is respectively connected to each node in the adjacent first recurrent layer, the first output layer is used to output a context vector after the last cycle, and the number of output nodes of the first output layer is equal to the number of elements in the context vector The same; the second recurrent network includes a second input layer, at least one second recurrent layer and a second output layer connected in sequence, the nodes in the second input layer correspond to each element in the context vector respectively, each node of the second input layer is connected to each node in the adjacent second recurrent layer respectively, the second input layer is used to input the context vector only in the first loop, the second recurrent layer is used to output in each loop according to the input of the second input layer and the state of the second recurrent layer in the previous loop, the number of nodes in the second output layer is one, the nodes of the second output layer are connected to each node in the adjacent second recurrent layer, the second output layer is used to output an access frequency prediction value in each loop, and multiple access frequency prediction values ​​constitute the output result of the preset recurrent neural network model output; Among them, according to the output results, the predicted access frequency is obtained, including: Selecting a maximum value from multiple access frequency prediction values ​​as the predicted access frequency; Among them, according to the predicted access frequency, preset priority and data volume, the recovery priority of each target data is obtained, including: The recovery priority of the target data is calculated by the following formula: ； in, Indicates the recovery priority of the target data. Indicates the preset priority of the target data. represents the predicted access frequency of the target data, is a natural constant, Indicates the amount of target data, is the preset unit adjustment factor, , and They have different preset weights.

Images