Storage space management method and device, electronic equipment and computer program product

Abstract

The invention provides a storage space management method and device, electronic equipment and a computer program product, and relates to the field of space management. The storage space management method comprises the following steps: inputting use information of a storage space into a dynamic prediction model to obtain a predicted use condition of the storage space; wherein the usage information comprises a historical usage amount sequence and service features; inputting the current use condition and the predicted use condition into an execution decision model, and determining a target operation; and managing the storage space according to the target operation. According to the storage space management method provided by the embodiment of the invention, the service scene can be sensed, the disk space can be accurately predicted, and the disk space can be dynamically managed according to the prediction result; and an execution result based on disk space management exists, optimization of the prediction model is realized, and the automation level and reliability of storage resource management are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of space management, in particular, to a storage space management party, device, electronic equipment and computer program product. BACKGROUND

[0002] The management of storage space is the key to guarantee system stability and business continuity. With the growth of data volume, insufficient storage space can cause service interruption, data loss or performance degradation, etc., which can affect user experience and increase enterprise operation and maintenance costs.

[0003] Currently, there are two main technologies for storage space management: one is based on static threshold alarm, and the other is simple trend prediction. Both of these two storage space management methods are one-way management logic, which cannot make specific analysis on the reasons for the growth of storage space, resulting in low overall efficiency and poor management effect of storage space management. SUMMARY

[0004] The purpose of the embodiments of the present application is to provide a storage space management party, device, electronic equipment and computer program product, which predicts the usage of storage space by combining business scenarios and user conditions, and performs corresponding target operations according to the prediction results through a special execution decision model after prediction, thereby realizing accurate prediction of storage space and effectively improving the automation level, reliability and accuracy of storage space management.

[0005] In a first aspect, the embodiments of the present application provide a storage space management method, which comprises: inputting the usage information of the storage space into a dynamic prediction model to obtain the predicted usage of the storage space; wherein the usage information comprises a historical usage sequence and business characteristics; inputting the current usage and the predicted usage into an execution decision model to determine a target operation; and managing the storage space according to the target operation.

[0006] In the above implementation process, the storage space management method provided by the embodiments of the present application, when predicting the usage of the storage space, considers business characteristics on the basis of time series prediction, making the prediction of the storage space more accurate. In addition, the storage space management method provided by the embodiments of the present application can dynamically manage the storage space based on the prediction result of the storage space; efficient storage management not only avoids resource waste caused by emergency expansion, but also avoids risks in advance through intelligent prediction, ensuring that critical businesses (such as databases and log systems) always have sufficient resource redundancy.

[0007] Optionally, in the embodiment of the present application, the dynamic prediction model comprises a usage prediction module and an attribution analysis module, the predicted usage comprises a predicted usage influencing factor and a predicted usage; obtaining the predicted usage of the storage space comprises: determining the predicted usage influencing factor based on the usage information of the storage space by the attribution analysis module; obtaining the predicted usage based on the usage information of the storage space and the predicted usage influencing factor by the usage prediction module.

[0008] In the above implementation process, the dynamic prediction model of the storage space management method provided by the embodiment of the present application comprises a usage prediction module and an attribution analysis module. The attribution analysis module quantifies the contribution value of business features (such as log generation rate and service request volume) to the growth of the storage space through the SHAP algorithm, accurately distinguishes natural growth from abnormal fluctuations; the usage prediction module realizes the consideration of business factors on the basis of baseline prediction based on the predicted usage influencing factor obtained by the attribution analysis, so as to accurately predict the predicted usage of the storage space and can clearly predict the driving factors of the prediction result; both solve the passivity of static threshold alarm and overcome the defect that simple trend prediction is not sensitive to business scenarios.

[0009] Optionally, in the embodiment of the present application, the dynamic prediction model comprises: = LSTM +α•MLP ; wherein, is a historical usage sequence before t time, is a predicted usage sequence of k time after t time, is a service request volume, is a log generation rate, and α is a dynamic weight.

[0010] Optionally, in the embodiment of the present application, the usage prediction module comprises a historical usage analysis network and a business feature analysis network; the predicted usage is obtained based on the usage information of the storage space and the predicted usage influencing factor, comprising: inputting the historical usage sequence into the historical usage analysis network to obtain a time sequence feature vector and determine a first usage; inputting the business features into the business feature analysis network to obtain a business feature vector; using a dynamic weight to weight the business feature vector to determine a second usage; wherein the dynamic weight is related to the predicted usage influencing factor; the predicted usage determined by the sum of the first usage and the second usage is obtained.

[0011] In the implementation process, the usage prediction module of the storage space management method provided in the embodiment of the application includes a historical usage analysis network and a service feature analysis network. The historical usage analysis network (LSTM) extracts periodicity, trend and other time sequence features by analyzing time sequence data, and outputs a first usage reflecting the natural growth law. The service feature analysis network (MLP) analyzes real-time service indicators (such as log generation rate and request volume) to generate a service feature vector, and outputs a second usage after weighting by a dynamic weight. The two are fused by weighted summation, so that the prediction result not only retains the stability of long-term trends, but also dynamically responds to business mutations.

[0012] Optionally, in the embodiment of the application, the determination manner of the dynamic weight comprises the following steps: fusing the time sequence feature vector and the service feature vector to obtain a fused feature vector; obtaining an initial weight through an attention mechanism for the fused feature vector; and modulating the initial weight according to a prediction usage influencing factor to obtain the dynamic weight.

[0013] In the implementation process, the dynamic weight mechanism of the storage space management method provided in the embodiment of the application realizes comprehensive improvement of the performance of the storage space prediction model through multi-stage collaborative optimization. At the feature processing level, the organic fusion of time sequence features and service features not only retains historical laws but also incorporates real-time business influences; the weight generation link realizes sensitive response to business abnormalities through an attention mechanism, so that the model can dynamically adjust the prediction focus. Therefore, the prediction accuracy is significantly improved, especially the judgment ability in the business fluctuation scenario, and the influence of business factors on the prediction result is clearly presented.

[0014] Optionally, in the embodiment of the application, the current usage includes a current usage amount and a current file distribution situation; and the determining the target operation comprises: inputting the current usage amount, the current file distribution situation and the predicted usage situation into a preset operation value function, and obtaining an operation value; and determining the target operation according to the operation value.

[0015] In the implementation process, the storage space management method provided in the embodiment of the application can intelligently balance the space release demand and the business risk through the Bellman equation of fusing the immediate return and the long-term value, wherein the risk cost calculation comprehensively considers the file access heat and the business criticality, effectively avoids the misdeletion of critical data, and significantly improves the accuracy and efficiency of the storage space automatic management under the premise of controllable risk.

[0016] Optionally, in the embodiment of the application, before the storage space is managed according to the target operation, the method further comprises: performing risk assessment on the target operation according to the access heat of the current storage space and the current business nature.

[0017] Optionally, in the embodiment of the application, the risk assessment is realized according to the following formula: ; wherein, is the maximum value of access hotness of all files in the current storage space, represents the recent access hotness of the target file, a is the target operation, represents the importance of the target file to the system stability and core function.

[0018] In the above implementation process, in the storage space management method provided by the embodiments of the present application, there is also risk assessment of the target operation, accurate identification of high-value files (such as high-frequency access core logs), and avoidance of accidental deletion of key data. While ensuring the efficiency of storage management, the business continuity risk is reduced to the minimum level.

[0019] Optionally, in the embodiments of the present application, after managing the storage space according to the target operation, the method further includes: monitoring the actual usage corresponding to the predicted usage, and evaluating the operation benefit of the target operation; and optimizing the dynamic prediction model and the execution decision model according to the actual usage and the operation benefit.

[0020] In the above implementation process, the storage space management method provided by the embodiments of the present application sets a closed-loop optimization mechanism, which realizes continuous optimization of the prediction model and the decision strategy through continuous feedback of the actual execution effect. The dynamic prediction model automatically adjusts the feature weight and the time sequence parameter based on the monitoring data, which significantly improves the prediction accuracy in abnormal scenarios; and the execution decision model makes the risk assessment more comprehensive and the operation selection more in line with the actual business needs by incorporating multi-dimensional business influence factors.

[0021] In a second aspect, the embodiments of the present application provide a storage space management device, which includes a dynamic prediction module, an execution decision module and a space management module; the dynamic prediction module is used to input the usage information of the storage space into a dynamic prediction model to obtain the predicted usage of the storage space; the execution decision module is used to input the current usage and the predicted usage into an execution decision model to determine a target operation; and the space management module is used to manage the storage space according to the target operation.

[0022] In a third aspect, the embodiments of the present application provide an electronic device, which includes a memory and a processor, and the memory stores program instructions; when the processor reads and runs the program instructions, the steps in any implementation manner of the first aspect are executed.

[0023] In a fourth aspect, the embodiments of the present application further provide a computer readable storage medium, which stores computer program instructions; when the computer program instructions are read and run by a processor, the steps in any implementation manner of the first aspect are executed.

[0024] In a fifth aspect, the embodiments of the present application further provide a computer program product, characterized by comprising computer programs / instructions, which are executed by a processor to perform the steps in any implementation manner of the first aspect. BRIEF DESCRIPTION OF DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments of the present application. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be regarded as a limitation to the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0026] Figure 1 The flow chart of the storage space management method provided by the embodiments of the present application; Figure 2 The first flow chart of the storage space prediction provided by the embodiments of the present application; Figure 3 The second flow chart of the storage space prediction provided by the embodiments of the present application; Figure 4 The flow chart of the dynamic weight generation provided by the embodiments of the present application; Figure 5 The flow chart of the target operation determination provided by the embodiments of the present application; Figure 6 The module schematic diagram of the storage space management device provided by the embodiments of the present application; Figure 7 The structural schematic diagram of the electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application. For example, the flowcharts and block diagrams in the drawings show the possible implementation architecture, functions and operations of the systems, methods and computer program products according to the embodiments of the present application. In this regard, each block in the flowchart or block diagram can represent a module, a program segment or a part of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in different order from that noted in the drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and they can also be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions. In addition, the functional modules in the various embodiments of the present application can be integrated together to form a separate part, or each module can exist independently, or two or more modules can be integrated to form a separate part.

[0028] Efficient management of storage space is the core foundation to ensure system stable operation and business continuity. In the digital era of explosive data growth, insufficient storage resources will directly trigger multiple risks; for example, service interruption causes key business downtime, data loss causes compliance and security crisis, and system performance drops dramatically, causing user experience collapse. These problems not only significantly increase enterprise operation and maintenance costs (such as emergency expansion expenses and fault recovery investment), but also can damage brand reputation and customer trust. Therefore, the management of storage space is very important to enterprises.

[0029] Currently, there are mainly two kinds of technologies for storage space management, one is based on static threshold alarm, and the other is simple trend prediction. Among them, the static threshold alarm triggers an alarm by presetting a fixed threshold (for example, triggering an alarm when the disk usage is greater than 90%), which completely relies on manual intervention and lacks forward-looking judgment of space usage trends; simple trend prediction is based on historical data (for example, storage space usage in the past 30 days) to predict future space occupancy using linear regression or moving average algorithms, but it does not analyze the growth causes combined with business characteristics, resulting in a prediction result that is out of touch with the real scene. Both of these two solutions use one-way management logic, which cannot distinguish between natural growth and abnormal fluctuations, and also does not form a linkage with automatic cleaning strategies, resulting in low overall management efficiency and poor results.

[0030] Based on this, the application provides a storage space management method, device, electronic equipment and computer program product. The storage space management method combines business scenarios and user conditions to predict the usage of the storage space. After prediction, a special execution decision model performs corresponding target operations, such as deleting part of the files, compressing part of the files, and the like, according to the prediction results, thereby realizing the management of the storage space. The storage space management method provided by the application can realize accurate prediction of the storage space, and effectively improve the automation level, reliability and accuracy of the management of the storage space.

[0031] Please refer to Figure 1 , Figure 1 The flowchart of the storage space management method provided by the embodiment of the application; the application provides a storage space management method, which can be executed by the electronic equipment. Figure 7 The storage space management method provided by the embodiment of the application includes the following steps: Step S100: inputting the usage information of the storage space into a dynamic prediction model to obtain the predicted usage of the storage space.

[0032] In the above step S100, the usage information of the storage space (for example, the historical usage sequence and the business characteristics) is input into the dynamic prediction model to obtain the predicted usage of the storage space.

[0033] The historical usage sequence refers to the disk space occupation record arranged in time sequence, such as The business characteristics refer to the business indicators that affect the storage space in real time, for example, the service request volume , the log generation rate .

[0034] In the embodiment of the application, the dynamic prediction model refers to a model that can dynamically predict the future usage or usage trend of the storage space based on the usage information of the storage space. Exemplarily, part of the dynamic prediction model can predict the growth area, and part of the dynamic prediction model can quantify the business impact. Therefore, the predicted usage output by the dynamic prediction model provided by the embodiment of the application is the predicted quantity considering both the user usage and the business scenario.

[0035] Step S200: inputting the current usage and the predicted usage into an execution decision model to determine the target operation.

[0036] In the above step S200, the predicted usage of the storage space predicted by the dynamic prediction model and the current usage of the storage space (for example, the current usage rate of the storage space and / or the file distribution situation of the storage space) are input into the execution decision model, and the target operation is output.

[0037] It should be noted that the decision model is executed in the embodiment of the application to obtain the optimal operation for the storage space, that is, to give the optimal operation for the current storage space according to the use condition of the current storage space and the predicted use condition of the storage space, such as deleting part of the file, compressing part of the file, retaining part of the file, and the like.

[0038] Step S300: managing the storage space according to the target operation.

[0039] In the above step S300, the target operation is executed to perform space release and the like on the storage space, thereby realizing the management of the storage space.

[0040] It should be noted that the storage space in the embodiment of the application can be a logical or physical storage area for saving files in various data storage media or systems, for example, a server local disk, a distributed storage cluster, a cloud storage service, a database table space, a log storage volume, a virtual machine or a container mounted virtual disk, and the like. It should be understood that as long as it involves space capacity management and file life cycle control, the technical solution of the application can be applied to block storage, file storage or object storage.

[0041] Through Figure 1 It can be known that the storage space management method provided in the embodiment of the application is more accurate in predicting the storage space on the basis of the time sequence prediction considering the business characteristics. In addition, the storage space management method provided in the embodiment of the application can dynamically manage the storage space based on the prediction result of the storage space; efficient storage management can not only avoid resource waste caused by emergency expansion, but also avoid risks in advance through intelligent prediction to ensure that the key business (such as a database and a log system) always has sufficient resource redundancy.

[0042] Please refer to Figure 2 , Figure 2 the first flowchart of the storage space prediction provided in the embodiment of the application; in the optional implementation manner of the embodiment of the application, the dynamic prediction model includes a usage prediction module and an attribution analysis module. The attribution analysis module is used to quantify the contribution value of each feature to the prediction result and to determine the core driving factor of the storage space growth; the usage prediction module is used to combine the result of the attribution analysis module to perform time sequence prediction on the storage space usage. The obtaining of the predicted use condition of the storage space in the above step S100 can be realized by the following steps: Step S110: determining the prediction usage influencing factor based on the use information of the storage space through the attribution analysis module.

[0043] In the above step S11, the attribution analysis module determines the prediction usage influencing factor based on the use information of the storage space, such as the historical disk usage and business characteristics, such as service request volume. Log generation rate Cache size, etc.

[0044] SHAP value for each feature Calculations are performed to reflect its marginal contribution to the predicted value. Specifically, Where F is the set of all feature techniques, such as, S is a feature subset, such as, |f(·)| is the model prediction function.

[0045] Factors influencing predicted usage are generally categorized as follows: natural growth factors, which are dominated by time-series characteristics and reflect baseline growth; and abnormal fluctuation factors, which are sudden increases in business characteristics and reflect potential log surges. In other words, it not only enables accurate prediction of usage but also identifies the driving factors behind the prediction results; for example, 95% of the usage in the next 7 days may be primarily due to a log surge.

[0046] Step S120: Using the usage prediction module, obtain the predicted usage based on storage space usage information and factors influencing predicted usage.

[0047] In step S120 above, the usage prediction module determines the predicted usage based on storage space usage information and factors influencing predicted usage. For example, it can combine attribution analysis results with dynamically weighted business characteristics to output a predicted value for future storage usage.

[0048] pass Figure 2 As can be seen, the dynamic prediction model of the storage space management method provided in this application includes a usage prediction module and an attribution analysis module. The attribution analysis module quantifies the contribution of business characteristics (such as log generation rate and service request volume) to storage space growth through the SHAP algorithm, accurately distinguishing between natural growth and abnormal fluctuations; the usage prediction module, based on the predicted usage influencing factors obtained from the attribution analysis, realizes the consideration of business factors on the basis of baseline prediction, enabling accurate prediction of storage space usage and clarifying the driving factors of the prediction results; it not only solves the passivity of static threshold alarms, but also overcomes the defect of simple trend prediction being insensitive to business scenarios.

[0049] Please refer to the following: Figure 3 , Figure 3 A second flowchart for storage space prediction is provided for embodiments of this application; in an optional embodiment, the usage prediction module includes a historical usage analysis network and a service characteristic analysis network. Step S120 above, based on storage space usage information and factors influencing predicted usage, obtains the predicted usage, which can be achieved through the following steps: Step S121: input the historical usage sequence into the historical usage analysis network to obtain a time sequence feature vector and determine a first usage amount.

[0050] Step S122: input the service feature into the service feature analysis network to obtain a service feature vector.

[0051] Step S123: use a dynamic weight to weight the service feature vector to determine a second usage amount.

[0052] Step S124: obtain a predicted usage amount determined by a sum of the first usage amount and the second usage amount.

[0053] The dynamic weight is related to a predicted usage amount influencing factor.

[0054] In the above steps S121 to S124, the service feature is input into the service feature analysis network to obtain the service feature vector, i.e., the first usage amount. The dynamic weight related to the service feature is determined in combination with the attribution analysis result to obtain the second usage amount. The final predicted usage amount is obtained after the first usage amount is adjusted by the second usage amount.

[0055] Exemplarily, in the embodiment of the present application, the usage prediction module described above is an LSTM-MLP hybrid model, where MLP is a multi-layer perceptron (Multi-Layer Perceptron) and LSTM is a long short-term memory network (Long Short-Term Memory). That is, first, time sequence modeling is performed to obtain the first usage amount LSTM to predict the natural growth trend; on the other hand, the service correction is performed using MLP, and MLP is used to correct the first usage amount based on the service factor. The second usage amount is obtained by weighting; and finally, the predicted amount composed of the first usage amount and the second usage amount is determined, where the second usage amount is used to adjust the first usage amount based on the service factor. The overall constructed prediction model is = LSTM + α • MLP ; where α is a service feature weight, different α is set according to different service conditions (considering the predicted usage amount influencing factor calculated above) to realize intelligent prediction of the storage space. Wherein, is the storage space usage amount before t time, is the usage prediction of k time after t time; for example, the prediction of the storage space growth amount in the future 7 days.

[0056] by Figure 3It can be known that the usage prediction module of the storage space management method provided in the embodiment of the application comprises a historical usage analysis network and a service feature analysis network. The historical usage analysis network (LSTM) extracts periodicity, trend and other time sequence features by analyzing time sequence data, and outputs a first usage reflecting a natural growth law. The service feature analysis network (MLP) analyzes real-time service indexes (such as log generation rate and request volume) to generate a service feature vector, and outputs a second usage after weighting by a dynamic weight. The two are fused by weighted summation, so that the prediction result not only retains the stability of long-term trends, but also dynamically responds to service mutations.

[0057] Please refer to Figure 4 , Figure 4 The dynamic weight generation flowchart is provided for the embodiment of the application. In the optional implementation of the embodiment of the application, the determination manner of the dynamic weight comprises the following steps: Step S1: fuse the time sequence feature vector and the service feature vector to obtain a fused feature vector.

[0058] In the above step S1, the time sequence feature vector from the LSTM and the service feature vector from the MLP are spliced and fused to obtain the fused feature vector which retains both the time sequence feature and the service feature.

[0059] Step S2: obtain an initial weight by the attention mechanism for the fused feature vector.

[0060] In the above step S2, the fused feature is processed by the attention mechanism. First, the fused vector is mapped to a scalar space by a learnable parameter matrix W and a bias b: Then, the Sigmoid function is used for normalization to compress the weight to [0, 1] to obtain the initial weight . Thus, in the case of service feature anomaly, the MLP vector increases significantly, the z value increases, closer to 1; in the case of service stability, the LSTM vector dominates, closer to 0.

[0061] Step S3: modulate the initial weight according to the prediction usage influencing factor to obtain a dynamic weight.

[0062] In the above step S3, the initial weight is modulated by the prediction usage influencing factor to obtain the dynamic weight α. Exemplarily, wherein β is a usage influencing factor modulation coefficient. Of course, the dynamic weight α is a number between [0, 1].

[0063] By Figure 4It can be known that the dynamic weight mechanism of the storage space management method provided in the embodiment of the application realizes comprehensive improvement of the performance of the storage space prediction model through multi-stage collaborative optimization. In the feature processing layer, the organic integration of the time sequence features and the business features not only retains the historical law but also incorporates the real-time business influence; the weight generation link realizes sensitive response to business abnormalities through the attention mechanism, so that the model can dynamically adjust the prediction focus. Therefore, the prediction accuracy is significantly improved, especially the judgment ability in the business fluctuation scenario, and the influence of business factors on the prediction result is clearly presented.

[0064] Please refer to Figure 5 , Figure 5 The target operation determination flowchart is provided for the embodiment of the application; in the optional implementation manner of the embodiment of the application, the current usage includes the current usage amount and the current file distribution situation; the determination of the target operation in the above step S200 includes the following steps. Step S210: inputting the current usage amount, the current file distribution situation and the predicted usage situation into a preset operation value function, and obtaining an operation value.

[0065] In the above step S210, according to the current usage amount, the current file distribution situation and the predicted usage situation, the operation value is calculated by using the preset operation value function.

[0066] Exemplarily, the operation value function is wherein, represents the immediate income, represents the future income. s represents the system state, such as the current usage amount and the current file distribution situation; A represents a candidate operation, and a represents an operation, for example, deletion, compression or retention; is a discount factor, generally a number between (0, 1), tends to 1, indicating that long-term income is valued, tends to 0, indicating that short-term immediate income is valued.

[0067] Step S220: determining the target operation according to the operation value.

[0068] In the above step S220, after the operation value of all candidate operations is calculated, the operation with the highest operation value is selected as the target operation. For example, in a relatively urgent scenario, such as when the space usage rate is more than 90%, non-core logs are deleted; for example, in a relatively stable scenario, such as when the space usage rate is less than 60%, a compression operation or a retention operation is performed.

[0069] Through Figure 5As can be seen, the storage space management method provided in this application embodiment can intelligently balance the demand for space release and business risks by integrating the Bellman equation, which combines immediate benefits and long-term value. The risk cost calculation comprehensively considers the popularity of file access and the criticality of business, effectively avoiding the accidental deletion of critical data, and significantly improving the accuracy and efficiency of automated storage space management under the premise of controllable risks.

[0070] In an optional embodiment, before managing the storage space according to the target operation, the method further includes: performing a risk assessment on the target operation based on the current access frequency of the storage space and the current business nature.

[0071] In the above implementation process, the access frequency of the current storage space, such as the number of accesses in the past 30 days, can be obtained through the file system audit log.

[0072] In this embodiment of the application, in addition to calculating the operational value of each operation, a risk assessment is also required before the operation is executed in order to quantify the potential business impact of each operation.

[0073] For example, ;in, =log(number of accesses in the last 7 days + 1), logarithmic compression avoids extreme values ​​and reflects the recent usage frequency of the file (e.g., 100 accesses correspond to a popularity of 2). Service Level Agreement (SLA) × File Relevance: This quantifies file importance using a service dependency graph (e.g., database logs: SLA=1.0, Relevance=0.9, Corresponding Criticality=0.9). For example, use the risk assessment method described above to conduct a risk assessment of the target operation. For example, the target operation is "delete a 10MB log file".

[0074] File A: 50 visits in 7 days (popularity = 1.7), SLA = 0.8, relevance = 0.6, criticality = 0.48. File B: 10 visits in 7 days (popularity = 1.0), SLA = 1.0, relevance = 0.3, criticality = 0.3. Maximum popularity = 2.3.

[0075] Then Risk=( ×0.48)+( (×0.3)≈0.35+0.13=0.48. Assuming the system sets the risk threshold to 0.5, the target operation of "deleting a 10MB log file" does not pose a security risk and can be executed.

[0076] Therefore, in the storage space management method provided by the embodiment of the application, risk assessment of the target operation is further performed, high-value files (such as high-frequency access core logs) are accurately identified, and misdeletion of key data is avoided, so that the business continuity risk is reduced to the minimum level while the storage management efficiency is ensured.

[0077] In an optional embodiment, after the storage space is managed according to the target operation, the method further comprises: monitoring actual usage corresponding to the predicted usage, and evaluating operation benefits of the target operation; and optimizing the dynamic prediction model and the execution decision model according to the actual usage and the operation benefits.

[0078] After the target operation is executed, the actual usage of the storage space is continuously monitored and compared with the predicted value. After the prediction deviation and the operation benefits are monitored, the dynamic prediction model is incrementally trained, the actual usage data is added to the training set to adjust the time sequence weight of the LSTM, and the business feature influence (such as the initial weight α of the log generation rate) is calibrated through SHAP analysis. At the same time, the risk assessment mechanism of the execution decision model is optimized, the query delay factor is added, and the space benefit coefficient is dynamically adjusted, so that the strategy is more suitable for the actual business tolerance.

[0079] Therefore, the storage space management method provided by the embodiment of the application sets a closed-loop optimization mechanism, continuously optimizes the prediction model and the decision strategy through continuous feedback of the actual execution effect, automatically adjusts the feature weight and the time sequence parameter of the dynamic prediction model based on the monitoring data, and significantly improves the prediction accuracy in abnormal scenarios; and the execution decision model is more comprehensive in risk assessment and more suitable for actual business needs in operation selection by incorporating multi-dimensional business influence factors.

[0080] Please refer to Figure 6 , Figure 6 a module schematic diagram of the storage space management apparatus provided by the embodiment of the application; the embodiment of the application provides a storage space management apparatus, which comprises a dynamic prediction module 110, an execution decision module 120, and a space management module 130.

[0081] The dynamic prediction module 110 is configured to input the usage information of the storage space into a dynamic prediction model to obtain predicted usage of the storage space.

[0082] The execution decision module 120 is configured to input the current usage and the predicted usage into an execution decision model to determine a target operation.

[0083] The space management module 130 is configured to manage the storage space according to the target operation.

[0084] In an optional embodiment, the dynamic prediction model comprises a usage prediction module and an attribution analysis module, and the predicted usage includes predicted usage factors and predicted usage; in the process of obtaining the predicted usage of the storage space, the dynamic prediction module 110 is specifically configured to: determine the predicted usage factors based on the usage information of the storage space through the attribution analysis module; and obtain the predicted usage based on the usage information of the storage space and the predicted usage factors through the usage prediction module.

[0085] In an optional embodiment, the usage prediction module comprises a historical usage analysis network and a service feature analysis network; in the process of obtaining the predicted usage based on the usage information of the storage space and the predicted usage factors, the dynamic prediction module 110 is specifically configured to: input the historical usage sequence into the historical usage analysis network to obtain a time sequence feature vector and determine a first usage; input the service feature into the service feature analysis network to obtain a service feature vector; and determine a second usage by weighting the service feature vector using a dynamic weight; wherein the dynamic weight is related to the predicted usage factors; and obtain the predicted usage determined by the sum of the first usage and the second usage.

[0086] In an optional embodiment, the determination manner of the dynamic weight comprises the following steps: fusing the time sequence feature vector and the service feature vector to obtain a fused feature vector; obtaining an initial weight by passing the fused feature vector through an attention mechanism; and obtaining the dynamic weight by modulating the initial weight according to the predicted usage factors.

[0087] In an optional embodiment, the dynamic prediction model comprises: = LSTM +α•MLP ; wherein, is a historical usage sequence before t time, is a predicted usage sequence of k time after t time, is a service request amount, is a log generation rate, and α is a dynamic weight.

[0088] In an optional embodiment, the current usage includes current usage and current file distribution; in the process of determining the target operation, the execution decision module 120 is specifically configured to: input the current usage, the current file distribution and the predicted usage into a preset operation value function, and obtain an operation value; and determine the target operation according to the operation value.

[0089] In an optional embodiment, the storage space management apparatus is further configured to perform risk assessment on the target operation according to the access heat of the current storage space and the current business nature before managing the storage space according to the target operation.

[0090] In an optional embodiment, the risk assessment is implemented according to the following formula: ; wherein, is the maximum value of the access hotness of all files in the current storage space, represents the recent access hotness of the target file, and a is the target operation, represents the importance of the target file to the system stability and core functions.

[0091] In an optional embodiment, the storage space management apparatus is further configured to monitor actual usage corresponding to the predicted usage and evaluate operation benefits of the target operation after managing the storage space according to the target operation; and optimize the dynamic prediction model and the execution decision model according to the actual usage and the operation benefits.

[0092] Please refer to Figure 7 , Figure 7 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. An electronic device 200 provided by an embodiment of the present application includes a processor 201 and a memory 202, the memory 202 stores machine readable instructions executable by the processor 201, and the machine readable instructions are executed by the processor 201 to perform the method described above.

[0093] Based on the same inventive concept, an embodiment of the present application further provides a computer program product, which includes computer programs / instructions, and the computer programs / instructions are executed by a processor to implement the steps of the storage space management method of the first aspect described above.

[0094] Based on the same inventive concept, an embodiment of the present application further provides a computer readable storage medium, which stores computer program instructions, and the computer program instructions are read and run by a processor to perform the steps in any implementation manner of the first aspect described above.

[0095] The computer readable storage medium can be a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), and various media that can store program codes.

[0096] In this document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0097] The embodiments of the present application described herein merely serve as examples and are not intended to limit the scope of the present application. The present application can be variously changed and modified without departing from the spirit and scope of the present application. Accordingly, any and all modifications, variations or equivalent arrangements that do not depart from the spirit and scope of the present application should be considered to be within the scope of the present application.

Claims

1. A storage space management method, characterized in that, The method includes: The storage space usage information is input into a dynamic prediction model to obtain the predicted usage of the storage space; wherein, the usage information includes historical usage sequences and business characteristics; Input the current usage and the predicted usage into the execution decision model to determine the target operation; Manage the storage space according to the stated target operation.

2. The method according to claim 1, characterized in that, in, The dynamic prediction model includes a usage prediction module and an attribution analysis module. The predicted usage includes factors influencing the predicted usage and the predicted usage. Obtaining the predicted usage of the storage space includes: Based on the storage space usage information, the attribution analysis module determines the factors influencing the predicted usage. The usage prediction module obtains the predicted usage based on the storage space usage information and the factors influencing the predicted usage.

3. The method according to claim 2, characterized in that, The usage prediction module includes a historical usage analysis network and a service characteristic analysis network; obtaining the predicted usage based on the storage space usage information and factors influencing predicted usage includes: The historical usage sequence is input into the historical usage analysis network to obtain a time-series feature vector and determine the first usage. The business features are input into the business feature analysis network to obtain the business feature vector; The business feature vector is weighted using dynamic weights to determine the second usage amount; wherein the dynamic weights are related to the factors influencing the predicted usage amount. The predicted usage amount is obtained by the sum of the first usage amount and the second usage amount.

4. The method according to claim 3, characterized in that, in, The determination of the dynamic weights includes the following steps: The time-series feature vector and the business feature vector are fused to obtain a fused feature vector; The fused feature vector is then processed through an attention mechanism to obtain initial weights; The initial weights are adjusted based on the factors influencing the predicted usage to obtain the dynamic weights.

5. The method according to claim 1, characterized in that, in, The current usage status includes current usage volume and current file distribution; the target determination operation includes: Input the current usage, current file distribution, and predicted usage into a preset operation value function to obtain the operation value; Based on the operational value, the target operation is determined.

6. The method according to claim 1, characterized in that, Prior to managing the storage space according to the target operation, the method further includes: A risk assessment is conducted on the target operation based on the current access frequency of the storage space and the nature of the current business.

7. The method according to claim 1, characterized in that, After managing the storage space according to the target operation, the method further includes: Monitor the actual usage corresponding to the predicted usage, and evaluate the operational benefits of the target operation; Based on the actual usage and operational benefits, optimize the dynamic prediction model and the execution decision model.

8. A storage space management device, characterized in that, The storage space management device includes: a dynamic prediction module, an execution decision module, and a space management module; The dynamic prediction module is used to input storage space usage information into the dynamic prediction model to obtain the predicted usage of the storage space. The execution decision module is used to input the current usage and the predicted usage into the execution decision model to determine the target operation; The space management module is used to manage the storage space according to the target operation.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores program instructions, and when the processor executes the program instructions, it performs the steps of the method according to any one of claims 1-7.

10. A computer program product, characterized in that, The computer program product includes a computer program / instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1-7.

Images