Method and system for automatically generating data backup strategy

Abstract

The invention relates to the technical field of data backup, in particular to a data backup strategy automatic generation method and system. The method comprises the following steps: acquiring and quantifying strategy generation input information; evaluating strategy-related attributes of the data assets, and estimating resource input costs meeting different preset protection levels according to strategy generation input information; based on the strategy generation input information, the strategy related attributes and the resource input cost, constructing and updating a dynamic model reflecting the strategy tradeoff relation in real time; performing strategy decision based on the dynamic model, wherein the strategy decision selects a backup strategy based on compliance laws and regulations in combination with real-time resource conditions; and deploying a backup strategy, continuously monitoring an execution effect of the backup strategy, and taking actual operation data of the execution effect as feedback for adjusting parameters of the dynamic model. The limitation of an existing data backup strategy automatic generation method in complex conditions is solved, and the ability of actual recovery result feedback optimization strategies is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data backup, in particular to a data backup strategy automatic generation method and system. BACKGROUND

[0002] In modern enterprise operations, data has become a core asset, and its security and recoverability are directly related to the continuity of business and the survival of the enterprise. In order to effectively protect the security of these massive data, the formulation and implementation of data backup strategy is particularly important. Traditional backup strategies often rely on manual experience, which not only consumes time and effort, but also easily leads to improper strategy configuration due to negligence, thereby affecting the efficiency and success rate of data recovery. In order to overcome these challenges, enterprises generally introduce data backup strategy automatic generation methods, aiming to automatically recommend or generate an optimized backup strategy by intelligently analyzing data characteristics and business needs, in order to improve management efficiency and the reliability of data protection. However, in actual application, with the rapid development of business and the continuous evolution of IT environment, these automated methods also face a series of new and complex challenges.

[0003] Specifically, the existing data backup strategy automatic generation method has obvious limitations in the face of complex situations such as real-time changes in data characteristics, intensified competition for backup resources, conflicts between compliance and cost targets, discrepancies between actual recovery performance and expectations, and sudden security incidents. For example, when the data characteristics of an enterprise change in real time due to sudden business growth, the original backup strategy generated based on static data analysis may no longer be applicable, resulting in the inability to meet new recovery point target requirements. At the same time, when multiple business systems simultaneously perform backup operations, shared resources within the data center, such as network data transmission capacity and storage device input / output performance, begin to show bottlenecks, affecting backup efficiency and even production system performance. In addition, the compliance requirements and cost control targets faced by enterprises are also constantly changing, and existing strategies may not fully meet new compliance requirements or generate unnecessary overhead. The deeper problem is that existing methods often only focus on the execution status of backup tasks, and lack the ability to iterate and optimize strategies based on actual recovery results, resulting in discrepancies between actual recovery performance and expected targets.

[0004] In view of the above problems, the existing technology needs to be improved. SUMMARY

[0005] The present application discloses a data backup strategy automatic generation method, which solves the limitations of existing data backup strategy automatic generation methods in complex situations and improves the ability to optimize strategies based on actual recovery results.

[0006] The technical solution of the present application is as follows: In a first aspect, the present application discloses a data backup strategy automatic generation method, comprising: acquire and quantify policy generation input information, the policy generation input information including compliance regulations of external data and cost control targets of enterprises internally; evaluate policy-related attributes of data assets, the policy-related attributes including data sensitivity, importance, current backup policy state, potential compliance risk, and recovery performance requirement, and estimate resource input costs for meeting different preset protection levels according to the policy generation input information; construct and update a dynamic model reflecting policy trade-off relationships in real time based on the policy generation input information, the policy-related attributes, and the resource input costs; make a policy decision based on the dynamic model, the policy decision selecting a backup policy based on compliance regulations in combination with real-time resource conditions, the real-time resource conditions including available capacity of storage devices, network data transmission capability, and input / output performance of storage devices; deploy the backup policy and continuously monitor execution effects of the backup policy, actual running data of the execution effects being used as feedback for adjusting parameters of the dynamic model.

[0007] Further, in the process of making a policy decision based on the dynamic model and deploying the backup policy, it further includes: collect and store micro-pattern characteristics of key network transmission paths, the micro-pattern characteristics including packet delay, delay jitter, packet interval time distribution, transmission control protocol window size change pattern, and throughput distribution, forming a normal operation baseline; when a large-scale data migration or backup task is started, real-time sampling and analysis of micro-pattern characteristics of task data streams are performed to obtain a task flow characteristic sequence; compare the task flow characteristic sequence with the normal operation baseline to calculate a deviation degree index; when the deviation degree index continuously exceeds a preset deviation threshold within a preset time window, it is determined that there is an implicit anomaly, an abnormality warning is triggered, and an abnormal event is provided as input to the policy decision for policy adjustment; prioritize preset emergency shunting plans according to recovery point target satisfaction, cost increment, and resource availability, the emergency shunting plans including alternative transmission paths and backup scheduling schemes; based on the abnormal event, recommend an alternative scheme with the highest priority from the emergency shunting plans, execute the alternative scheme, and adjust parameters of the dynamic model based on execution results of the alternative scheme as feedback.

[0008] Further, in the process of making a policy decision based on the dynamic model, it further includes: establishing a performance isolation threshold library of shared infrastructure components, the performance isolation threshold library recording maximum concurrent data flow quantity and / or input-output performance of each shared infrastructure component under different service loads and interference characteristics of specific types of data flow causing performance interference to other services; simulating resource occupation patterns of the candidate scheme on the shared infrastructure components according to task scale, data flow type and scheduling time of the candidate large-scale data migration or backup task; matching the resource occupation patterns with the interference characteristics in the performance isolation threshold library, marking the candidate scheme as a high-risk scheme when the matching result predicts that the candidate scheme will cause resource occupation of the shared infrastructure components to exceed the performance isolation threshold and / or overlap with the interference characteristics, and calculating an interference penalty cost; counting the interference penalty cost into total cost input of the candidate scheme in the dynamic model, and selecting a backup strategy from the candidate schemes that meet compliance regulations based on adjusted total cost input and in combination with real-time resource status.

[0009] Further, in the process of deploying the backup strategy and continuously monitoring the execution effect of the backup strategy, it also includes: establishing a business scenario recovery path map, the business scenario recovery path map being used to map internal structure, mutual dependency relationship of data assets and application environment supported thereby; when performing a business scenario recovery verification process, tracing and recording a complete path and corresponding time from data recovery to end-to-end function recovery of a business system based on the business scenario recovery path map; analyzing the complete path and corresponding time, identifying key dependent links and bottlenecks in the actual recovery path, and quantifying the influence of each key dependent link and bottleneck on overall business recovery time; comparing the overall business recovery time with a recovery time target and / or a recovery point target preset by a business department, and calculating a deviation degree; when the deviation degree exceeds a preset time deviation threshold, identifying actual deviation between actual recovery performance and business expectation caused by changes in internal structure, mutual dependency relationship or application environment; converting the actual deviation into a business scenario recovery benefit correction factor, which is used to adjust benefit function parameter in the dynamic model.

[0010] Further, in the process of evaluating the strategy-related attributes of the data assets, it also includes: continuously monitoring and analyzing business process iteration information, the business process iteration information including business system function update, data processing process adjustment and user access behavior change; identifying affected data assets and associated business scenarios according to the business process iteration information; For the affected data assets, analyze the call frequency, data update frequency and data access path of the data assets in the business system, determine the current business value and access mode in combination with the business system log and user behavior data, and re-evaluate the compliance classification based on the influence of the data processing method on the business process iteration information to obtain real-time detection results; Compare the real-time detection results with the policy-related attributes stored in the data asset metadata, and calculate the attribute deviation degree; When the attribute deviation degree exceeds the preset attribute threshold, automatically update the policy-related attributes of the data assets; And according to the updated policy-related attributes and policy generation input information, re-estimate the resource input cost required to meet different preset protection levels, and / or trigger real-time update of the dynamic model.

[0011] Further, continuously monitor and analyze business process iteration information, which also includes: Integrate enterprise service bus and / or message queue to monitor service update messages published by the business system; According to the service update message, configure the log parsing adapter to convert heterogeneous update log information into a standardized data structure; According to the standardized data structure, build an iteration information dependency graph to identify update events with conflicts and / or sequential dependencies; Use natural language processing technology to analyze unstructured business documents and change requests to extract keywords and semantic associations of business system function updates, data processing flow adjustments and user access behavior changes to obtain semantic association results; Integrate the standardized data structure, iteration information dependency graph and semantic association results to generate a business process overall change report for use when evaluating the policy-related attributes of the data assets.

[0012] Further, in the process of real-time detection of the affected data assets, it also includes: Deploy performance monitoring probes at the infrastructure level to collect read / write delays, IOPS, throughput of storage media, and bandwidth utilization, packet loss rate and link jitter of network equipment in real time; Deploy audit probes at the security policy execution level to record data desensitization operation execution logs and / or access control rule trigger events in real time; Correlate and analyze the infrastructure indicators collected by the performance monitoring probes with the access mode of the data assets to identify changes in data asset access performance caused by infrastructure changes; Correlate and analyze the security policy execution logs recorded by the audit probes with the compliance classification of the data assets to identify changes in data asset compliance risk caused by security policy changes; According to the result of the correlation analysis, the real-time detection result of the business value, compliance classification and / or access mode of the data asset is corrected; Based on the corrected real-time detection result, the calling frequency, data update frequency and data access path of the data asset in the business system are re-analyzed by combining the business system log and user behavior data, the current business value and access mode of the data asset are determined, and the compliance classification is re-evaluated.

[0013] Further, when the infrastructure indicators collected by the performance monitoring probe are correlated with the access mode of the data asset, the following steps are further included: The infrastructure indicator stream and the data asset access mode stream are transmitted in a streaming manner, and identification information is attached, the identification information including at least data asset identification and collection source identification; A timestamp is attached to each infrastructure indicator sample and each data asset access event during collection, and the infrastructure indicator stream and the data asset access mode stream are aligned based on the timestamps; The aligned infrastructure indicator stream and the data asset access mode stream are correlated using a sliding time window; After the infrastructure indicators show fluctuations exceeding a preset fluctuation threshold, the change trend of the data asset access performance in a subsequent preset time window is tracked, and when there is a consistent change trend in a continuous number of not less than a preset window count threshold, it is determined that there is a causal relationship; When it is determined that there is a causal relationship, the influence degree of the infrastructure change on the data asset access performance is quantified, and the quantification result is used to correct the parameters of the dynamic model and / or used for policy decision based on the dynamic model.

[0014] Further, when quantifying the influence degree of the infrastructure change on the data asset access performance, the following steps are further included: For a data asset, identify whether its access mode is a bursty access mode and / or a periodic access mode; When it is identified as a bursty access mode, after the infrastructure change occurs, the access performance indicators of the data asset are collected at a preset high-frequency sampling interval until the fluctuation amplitude of the access performance indicators in a continuous number of not less than a preset window count threshold in a preset time window does not exceed a preset stability threshold; the access performance indicators in the initial observation window after the change are compared with the access performance indicators in the equal-length reference window before the change to obtain an instantaneous difference, and the instantaneous difference is taken as the instantaneous influence degree; When it is identified as a periodic access mode, after the infrastructure change occurs, the access performance indicators of the data asset for at least one complete period are continuously collected, and the access performance indicators in the complete period after the change are compared with the access performance indicators in the corresponding period before the change to calculate a periodic average difference, and the periodic average difference is taken as the continuous influence degree; In combination with the transient influence degree and the persistent influence degree, a quantification result of the influence degree of the infrastructure change on the data asset access performance is obtained, and the quantification result is used to correct parameters of the dynamic model and / or used for a policy decision based on the dynamic model.

[0015] In a second aspect, the present application further discloses a data backup policy automatic generation system, comprising: An information acquisition module is configured to acquire and quantify policy generation input information, the policy generation input information comprising compliance regulations of external data and cost control targets of an enterprise internally; An asset evaluation module is configured to evaluate policy-related attributes of data assets, the policy-related attributes comprising data sensitivity, importance, current backup policy state, potential compliance risk, and recovery performance requirement, and estimate resource input costs for meeting different preset protection levels according to the policy generation input information; A model construction module is configured to construct and update a dynamic model reflecting a policy trade-off relationship in real time based on the policy generation input information, the policy-related attributes, and the resource input costs; A policy decision module is configured to make a policy decision based on the dynamic model, the policy decision being based on the compliance regulations in combination with real-time resource conditions to select a backup policy, the real-time resource conditions comprising available capacity of a storage device, network data transmission capability, and input / output performance of the storage device; A policy deployment and monitoring module is configured to deploy the backup policy and continuously monitor an execution effect of the backup policy, and use actual running data of the execution effect as feedback to adjust parameters of the dynamic model. Advantageous effects

[0016] The present application discloses a data backup policy automatic generation method, which acquires and quantifies policy generation input information, evaluates policy-related attributes of data assets, estimates resource input costs for meeting different preset protection levels, and constructs and updates a dynamic model reflecting a policy trade-off relationship in real time. On this basis, a policy decision is made based on the dynamic model, a backup policy is selected in combination with compliance regulations and real-time resource conditions, and the backup policy is deployed, the execution effect thereof is continuously monitored, and actual running data is used as feedback to adjust parameters of the dynamic model. The problems of relying on manual experience, low efficiency, and errors in the prior art are effectively solved, the automation and intelligent level of the data backup policy are significantly improved, and the reliability of data protection and the business continuity of an enterprise are enhanced. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 A flowchart of a data backup policy automatic generation method provided by the present application is shown.

[0018] Figure 2A module structure schematic diagram of a data backup strategy automatic generation system provided in the present application is shown.

[0019] In the figure: 1, information acquisition module; 2, asset evaluation module; 3, model construction module; 4, strategy decision module; 5, strategy deployment and monitoring module. DETAILED DESCRIPTION

[0020] The technical solutions in the present application will be described clearly and completely below in combination with the drawings in the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. The components of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0021] REFERENCE Figure 1 The present application proposes a data backup strategy automatic generation method, comprising: S1000: acquiring and quantifying strategy generation input information, the strategy generation input information including compliance regulations of external data and cost control targets of enterprise internal; S2000: evaluating strategy-related attributes of data assets, the strategy-related attributes including data sensitivity, importance, current backup strategy state, potential compliance risk and recovery performance demand, and estimating resource input costs for meeting different preset protection levels according to the strategy generation input information; S3000: constructing and real-time updating a dynamic model reflecting strategy trade-off relationship based on the strategy generation input information, the strategy-related attributes and the resource input costs; S4000: making a strategy decision based on the dynamic model, the strategy decision selecting a backup strategy based on compliance regulations combined with real-time resource conditions, the real-time resource conditions including available capacity of storage devices, network data transmission capability and input / output performance of storage devices; S5000: deploying the backup strategy, and continuously monitoring execution effect of the backup strategy, actual running data of the execution effect being used as feedback for adjusting parameters of the dynamic model.

[0022] The present application provides a data backup strategy automatic generation method, aiming to realize automatic generation, optimization and adjustment of data backup strategies through intelligent means. The core of the method is to construct a dynamic decision system capable of responding to external environmental changes and internal resource conditions in real time.

[0023] Among them, the policy generation input information is the basis for driving the entire policy generation process, which includes the compliance regulations of external data and the cost control targets of the enterprise. Compliance regulations refer to legal requirements for data storage, processing and transmission such as GDPR, HIPAA, etc., which directly affect the protection level and method of data backup. The cost control target represents the budget limit and efficiency pursuit of the enterprise in data protection, for example, the enterprise may set an annual total backup cost upper limit, or require that the backup storage cost not exceed a certain percentage.

[0024] The policy-related attributes of data assets are the key to assessing data value and risk, including data sensitivity, importance, current backup policy status, potential compliance risk, and recovery performance requirements. Data sensitivity may refer to personal identity information, financial data, etc., which usually require higher protection levels. Importance may refer to core business system data, which has more stringent recovery time objective (RTO) and recovery point objective (RPO) requirements. The current backup policy status reflects the protection level and backup period of the data at present. The potential compliance risk refers to the risk of non-compliance that the data may face under the existing policy. The recovery performance requirement refers to how long the data needs to be recovered to an available state when a failure occurs, and how much data loss can be tolerated.

[0025] The dynamic model is a mathematical or logical model that can be updated in real time to reflect the trade-off between cost, risk and benefit under different backup policy choices. The model can be a multi-objective optimization model, for example, balancing compliance, cost and recovery performance through linear programming or nonlinear programming.

[0026] Real-time resource status refers to the current available physical and logical resource status of the system when making policy decisions, including storage device available capacity, network data transmission capability and storage device input / output performance. For example, the storage device available capacity refers to the unoccupied storage space in the current storage array or cloud storage service. The network data transmission capability refers to the real-time bandwidth and delay of the internal or external network link of the data center. The storage device input / output performance refers to the number of read / write operations per second (IOPS) and data throughput that the storage system can handle.

[0027] First, in terms of acquiring and quantifying policy generation input information, the policy generation input information includes compliance regulations of external data and cost control targets of internal enterprises. For example, the latest compliance regulations such as data retention period, encryption requirements, etc. can be manually input or obtained from a compliance management system through an API interface. These regulations can be quantified as a series of constraints, for example, requiring that a certain type of data must be retained for at least five years, or AES-256 encryption must be used. The cost control targets of internal enterprises can be set through a financial system or management, for example, setting the monthly backup storage cost not to exceed 100,000 yuan, or the CPU and memory occupancy rate of backup operation not to exceed 5% of the production system. These targets can be quantified as cost items in the optimization function.

[0028] Second, in terms of evaluating the policy-related attributes of data assets, policy-related attributes include data sensitivity, importance, current backup policy state, potential compliance risk, and recovery performance requirements. For example, data sensitivity can be automatically identified and labeled by a data classification tool, such as labeling data containing an identity card number as "highly sensitive". Importance can be evaluated by business department priority setting or data access frequency, modification frequency, etc. indicators, such as core database data being labeled as "high importance". The current backup policy state can be obtained by querying the backup management system, for example, showing that a certain database currently uses a daily full backup plus hourly incremental backup strategy. Potential compliance risk can be evaluated by a compliance scanning tool, for example, finding that data in a certain storage bucket is not encrypted as required. Recovery performance requirements can be set by business department RTO and RPO requirements, for example, requiring that a certain key business system has an RTO of less than 4 hours and an RPO of less than 15 minutes. On this basis, the cost of resources required to meet different preset protection levels is estimated according to the policy generation input information. For example, for a highly sensitive data with an RPO of 1 hour and an RTO of 4 hours, the system can estimate the storage space, network bandwidth and computing resource cost required for a geo-redundancy plus snapshot backup scheme.

[0029] Third, in terms of constructing and updating a dynamic model reflecting the trade-off relationship of the policy based on policy generation input information, policy-related attributes and resource input cost. For example, a multi-objective optimization model can be constructed, in which the objective function includes minimizing total cost, minimizing compliance risk, maximizing recovery performance, etc. The constraints include compliance regulation requirements, resource availability restrictions, etc. The model can be solved using linear programming, integer programming or heuristic algorithms. When the policy generation input information or policy-related attributes change, for example, new compliance regulations are released or the sensitivity of a certain data asset is improved, the dynamic model will update its parameters and constraints in real time to reflect the latest trade-off relationship.

[0030] Then, in terms of policy decision based on the dynamic model, the policy decision selects a backup strategy based on compliance regulations combined with real-time resource conditions. Real-time resource conditions include storage device available capacity, network data transmission capability, and storage device input-output performance. For example, when making decisions, the dynamic model queries the available capacity of the current storage device. If the capacity of a certain storage pool is insufficient, it will preferentially select other storage pools or cloud storage. At the same time, the model will consider the current network data transmission capability to avoid large-scale data transmission during network peak periods to prevent affecting the production system. Storage device input-output performance is also taken into account to ensure that the backup task does not cause excessive load on the storage system. By comprehensively considering these real-time resource conditions, the dynamic model can select a backup strategy that not only meets compliance requirements but also efficiently executes under current resource conditions.

[0031] Finally, in terms of deploying the backup strategy and continuously monitoring the execution effect of the backup strategy, the actual running data of the execution effect is used as feedback to adjust the parameters of the dynamic model. For example, once the policy decision module selects a backup strategy, the strategy will be deployed to the backup management system. The system will continuously monitor the execution of the backup task, including backup success rate, backup time, data transmission volume, resource occupation, etc. These actual running data will be collected and fed back to the dynamic model. For example, if the actual execution time of a certain backup strategy always exceeds expectations or resource occupation is too high, the dynamic model will adjust its internal parameters based on this feedback data, such as adjusting the weight of a certain resource cost, to avoid similar problems in future decision-making, thereby achieving continuous optimization and adaptive adjustment of the strategy.

[0032] In another embodiment of the present application, it is further proposed that the process of policy decision based on the dynamic model and deployment of the backup strategy further includes: S6000: Collect and store the micro-pattern characteristics of the key network transmission path, including packet delay, delay jitter, packet interval time distribution, transmission control protocol window size variation pattern, and throughput distribution, to form a normal operation baseline; S7000: When a large-scale data migration or backup task is started, real-time sampling and analysis of the micro-pattern characteristics of the task data stream are performed to obtain a task flow characteristic sequence; S8000: Compare the task flow characteristic sequence with the normal operation baseline to calculate a deviation degree index; S9000: When the deviation degree index continuously exceeds a preset deviation threshold within a preset time window, it is determined that there is a non-explicit anomaly, an abnormality warning is triggered, and the abnormal event is provided as input to the policy decision for policy adjustment; S10000: Prioritize the preset emergency shunt plan according to the recovery point target satisfaction, cost increment, and resource availability, the emergency shunt plan including an alternative transmission path and a backup scheduling scheme; S11000: Based on the abnormal event, recommend the highest-priority alternative scheme from the emergency shunt plan, execute the alternative scheme, and use the execution result of the alternative scheme as feedback to adjust the parameters of the dynamic model.

[0033] Specifically, the micro mode characteristics of the key network transmission path refer to the fine-grained behavior patterns exhibited by the network layer and the transmission layer during data transmission. Packet delay refers to the time required for a data packet to be sent from the source end to the destination end, delay jitter refers to the variation amplitude of the packet delay, packet interval time distribution reflects the regularity of data packet arrival, transmission control protocol window size variation pattern reveals the behavior of transmission control protocol flow control, and throughput distribution represents the amount of successfully transmitted data per unit time and its fluctuation. The collection of these characteristics can be achieved by deploying probes at key network nodes or using the monitoring functions of existing network devices (such as routers and switches). Through long-term and continuous collection and statistical analysis of these micro mode characteristics, a baseline of the network in a normal operating state can be established, which represents the typical performance of the network under normal conditions.

[0034] When a large-scale data migration or backup task is initiated, the system will sample and analyze the micro mode characteristics of the task data stream in real time, generating a task flow feature sequence. This task flow feature sequence is then compared with the pre-established normal operation baseline to calculate the deviation degree index. The deviation degree index can be calculated using various statistical methods, such as Euclidean distance, Mahalanobis distance, or machine learning-based anomaly score, to quantify the difference between the current task flow characteristics and the normal baseline.

[0035] Among them, the non-explicit abnormality refers to the network performance degradation or behavior anomaly that does not immediately lead to service interruption or obvious failure, but can be identified through continuous deviation of micro mode characteristics. When the deviation degree index continuously exceeds the preset deviation threshold within a preset time window, such as consecutive minutes or hours, the system will determine that there is such a non-explicit abnormality and immediately trigger an abnormal warning. This abnormal event is then provided as a key input to the strategy decision module, so that the system can timely adjust the current backup strategy.

[0036] To cope with possible non-explicit anomalies, the system pre-sets multiple emergency shunting plans. These plans include but are not limited to switching to a backup network transmission path, adjusting data transmission priority, suspending part of non-critical tasks, or adopting a different backup scheduling scheme (e.g., switching from full backup to incremental backup). When an anomaly occurs, the system prioritizes these plans based on multiple dimensions such as recovery point objective satisfaction (RPO satisfaction based on a pre-set evaluation function), cost increment, and resource availability. The recovery point objective satisfaction evaluates the control ability of different plans over data loss risk, the cost increment considers the additional overhead that may be brought by executing the plan, and the resource availability ensures that the selected plan is feasible under the current resource conditions. Based on the analysis of the abnormal event and the prioritization of the plans, the system will recommend and execute the highest priority alternative. The execution results of the alternative, including its impact on network performance, backup progress, and resource consumption, will be used as feedback data to adjust the parameters of the dynamic model in real time, so that the model can better adapt to changes in the actual operating environment, and improve the accuracy and robustness of future strategy decisions.

[0037] In another embodiment of the present application, it is further proposed that the process of making strategy decisions based on the above dynamic model includes the following steps: S12000: Establish a performance isolation threshold library of shared infrastructure components, which records the maximum number of concurrent data streams and / or input / output performance of each shared infrastructure component under different business loads, and the interference characteristics of specific types of data streams on other services; S13000: Simulate the resource occupation mode of the candidate scheme on the shared infrastructure components according to the task size, data stream type, and scheduling time of the candidate large-scale data migration or backup task; S14000: Match the resource occupation mode with the interference characteristics in the performance isolation threshold library. When the matching result predicts that the candidate scheme will cause the resource occupation of the shared infrastructure components to exceed the performance isolation threshold and / or overlap with the interference characteristics, mark the candidate scheme as a high-risk scheme, and calculate the interference penalty cost; S15000: Include the interference penalty cost in the total cost input of the candidate scheme in the dynamic model, and select the backup strategy from the candidate schemes that meet the compliance regulations based on the adjusted total cost input and the real-time resource conditions.

[0038] In particular, a shared infrastructure component can refer to a storage array, a network switch, a virtualization platform, or a computing cluster that is shared within an enterprise. The performance isolation threshold library is designed to predefine the capacity boundary of these shared resources under different business pressure, and the performance impact pattern that a specific type of data flow (e.g., high IOPS database backup, large file transfer, etc.) can cause to other businesses (e.g., online transaction system, real-time analytics platform, etc.). For example, the threshold library can record the maximum IOPS or throughput of a certain storage array that can be used for backup tasks when it is carrying OLTP business, and how the characteristics of the backup task such as packet size distribution, burstiness, etc. affect the latency of the OLTP business.

[0039] Before making a policy decision, the system simulates the resource occupation pattern that a candidate large-scale data migration or backup task can cause on the shared infrastructure component, according to its expected task size (e.g., data volume), data flow type (e.g., sequential read / write, random read / write), and scheduled time. This simulation process can be done by using historical data, prediction algorithms, or simulation techniques.

[0040] Subsequently, the resource occupation pattern is matched with the pre-defined interference characteristics in the performance isolation threshold library. If the matching result predicts that the candidate scheme will cause the resource occupation of the shared infrastructure component to exceed its performance isolation threshold, e.g., exceed the bandwidth or IOPS upper limit reserved for other critical businesses, or its data flow characteristics overlap with known interference characteristics, e.g., the bursty read / write pattern of the backup task conflicts with the sensitive read / write pattern of the online business, the candidate scheme will be marked as a high-risk scheme. At this time, the system will calculate an interference penalty cost (e.g., interference penalty cost = Σ (affected system importance weight x performance decline ratio x influence duration x unit business loss)), which quantifies the potential loss caused by the possible performance decline, business interruption, or additional resource investment of the scheme.

[0041] Finally, the interference penalty cost will be included in the total cost input of the candidate scheme in the dynamic model. A scheme that can cause a conflict of shared resources, even if its direct resource consumption is low, will become "more expensive" because of the potential interference penalty cost. Based on the adjusted total cost input, combined with the real-time resource status such as the available capacity of storage devices, network data transmission capability, and storage device input / output performance, the system will select the backup strategy with the lowest comprehensive cost (including direct cost and potential interference cost) and the lowest risk from all candidate schemes that meet the compliance regulations.

[0042] As a specific implementation: Suppose a company has a shared storage area network (SAN) and a core database server that hosts high-concurrency online transaction services. The SAN is also used to store various backup data. When automatically generating a data backup strategy, the system needs to decide whether to perform a TB-level large-scale database backup task during the working hours.

[0043] First, the system queries a pre-established performance isolation threshold library. The threshold library records the minimum IOPS and bandwidth reserved for the database server by the SAN during the peak period of online transaction services, as well as the delay jitter interference characteristics that a large-scale backup task (e.g., sequential read-write mode) may cause to the database IO of random read-write mode.

[0044] Next, the system simulates the resource occupation mode of the TB-level backup task on the SAN during the scheduled time (e.g., working hours), including its expected IOPS, throughput, and packet characteristics.

[0045] Then, the system matches the simulated resource occupation mode with the performance isolation threshold and interference characteristics in the threshold library. If the matching result predicts that the execution of the backup task during the working hours will cause the IOPS or bandwidth occupation of the SAN to exceed the threshold reserved for online transaction services, and there is a serious overlap between the sequential read-write mode of the backup task and the random read-write mode of the database, which may cause a significant increase in the database response time, the backup scheme will be marked as a high-risk scheme. At this time, the system will calculate an interference penalty cost, for example, based on the quantitative value of the potential loss or user experience decline of the database business per minute.

[0046] Finally, the interference penalty cost will be included in the total cost of the backup scheme. For example, if the direct backup cost is 1000 yuan, but the interference penalty cost is 5000 yuan, the total cost becomes 6000 yuan. In contrast, a backup scheme executed during non-working hours may have a direct cost of 1200 yuan, but the interference penalty cost is 0 yuan, and the total cost is 1200 yuan. Based on the adjusted total cost, the system will preferentially select the backup scheme executed during non-working hours, thereby avoiding potential interference to the core business and ensuring the continuity and stability of the business.

[0047] In another embodiment of the present application, it is further proposed that the process of deploying a backup strategy and continuously monitoring the execution effect of the backup strategy further includes: S16000: Establish a business scenario recovery path diagram, which is used to map the internal structure, mutual dependency relationship of data assets, and the application environment supported thereby; S17000: When the business scenario recovery verification process is performed, the complete path and corresponding time from data recovery to end-to-end function recovery of the business system are tracked and recorded based on the business scenario recovery path graph; S18000: Analyze the complete path and corresponding time, identify key dependent links and bottlenecks in the actual recovery path, and quantify the impact of each key dependent link and bottleneck on the overall business recovery time; S19000: Compare the overall business recovery time with the recovery time target and / or recovery point target preset by the business department, and calculate the deviation; S20000: When the deviation exceeds the preset time deviation threshold, identify the actual deviation between the actual recovery performance and the business expectation caused by internal structure, mutual dependence or application environment change; S21000: Convert the actual deviation into a business scenario recovery benefit correction factor, which is used to adjust the benefit function parameter in the dynamic model.

[0048] Specifically, the establishment of the business scenario recovery path graph aims to provide a comprehensive view for clearly depicting all data assets on which a specific business function or service depends, the logical and physical connections between these data assets, and the software and hardware environments in which they run. The path graph can be represented in a graphical interface or a structured data model, and its purpose is to reveal data flow, service call chain, and potential single point of failure or recovery bottleneck. For example, for an online transaction system, its recovery path graph may include databases, application servers, message queues, cache services, and external interfaces, and details of their dependency order and data synchronization mechanism. When the business scenario recovery verification process is performed, it means simulating or actually performing a recovery operation of the business system to verify the effectiveness of the backup strategy. During this process, the system will track and record every step, every link, and the time spent by each link based on the pre-established business scenario recovery path graph, starting from the initial data recovery operation (e.g., recovering the database from the backup storage) to the complete recovery of the entire business system to its end-to-end function (e.g., users can log in and complete transactions normally). This includes data loading time, application startup time, service registration time, dependent service readiness time, etc.

[0049] By analyzing the complete path and time from data recovery to business function recovery, the longest time-consuming, error-prone and critical links are identified, and the influence of each dependency and bottleneck on the overall recovery time is quantified by sensitivity analysis or causal analysis. The actual total recovery time is compared with the recovery time target and recovery point target set in the business continuity planning or disaster recovery planning to obtain the deviation degree; if the deviation exceeds the threshold, the cause is located (such as database schema change, service dependency addition, operating system or hardware change). Accordingly, the deviation is converted into a business scenario recovery benefit correction factor to adjust the dynamic model benefit function: a large deviation increases the weight of rapid recovery, and a small deviation reduces it to explore more cost-effective strategies.

[0050] Further proposed in another embodiment of the present application is that the process of evaluating the policy-related attributes of the data assets further includes: S2100: Continuously monitor and analyze business process iteration information, including business system function updates, data processing process adjustments, and user access behavior changes; S2200: According to the business process iteration information, identify the affected data assets and their associated business scenarios; S2300: For the affected data assets, analyze the calling frequency, data update frequency and data access path of the data assets in the business system, determine their current business value and access mode in combination with the business system log and user behavior data, and reevaluate their compliance classification based on the influence of the business process iteration information on the data processing method to obtain real-time detection results; S2400: Compare the real-time detection results with the policy-related attributes stored in the data asset metadata to calculate the attribute deviation degree; S2500: When the attribute deviation degree exceeds the preset attribute threshold, automatically update the policy-related attributes of the data assets; S2600: And according to the updated policy-related attributes and policy generation input information, re-estimate the resource input cost required to meet different preset protection levels, and / or trigger real-time update of the dynamic model.

[0051] Specifically, continuously monitoring and analyzing business process iteration information refers to the system continuously collecting and analyzing the changes of internal business processes. These changes may include business system function updates, such as adding new modules, optimizing functions, or abandoning old functions; data processing flow adjustments, such as modifying data cleaning rules, changing data flow paths, or changing data storage formats; and user access behavior changes, such as significant increases or decreases in access frequency to specific data assets, or changes in access patterns from batch processing to real-time querying. These information can be obtained through integrating enterprise internal message queues, service buses, version control systems, or by parsing business documents, change requests, etc. Data processing methods include but are not limited to data collection, storage, replication, transmission, encryption, desensitization, compression, archiving, destruction, and other processing activities that affect the data lifecycle.

[0052] According to the business process iteration information, identifying the affected data assets and their associated business scenarios refers to the system automatically identifying which specific data assets (such as database tables, files, API interface data, etc.) may be affected based on the parsed business process changes, and the changes that may occur in the business scenarios (such as order processing, customer management, financial statements, etc.) supported by these data assets. For example, if a new function involves processing of customer sensitive information, the related customer data assets will be identified as affected.

[0053] For the affected data assets, analyze the calling frequency, data update frequency, and data access path of the data assets in the business system, and determine their current business value and access mode based on business system logs and user behavior data, and re-evaluate their compliance classification based on the impact of business process iteration information on data processing methods, to obtain real-time detection results. This step aims to in-depth analyze the actual use of affected data assets. Calling frequency and data update frequency reflect the activity and importance of data assets; data access path reveals data flow and dependency. Combined with business system logs (such as operation logs, error logs) and user behavior data (such as access records, operation statistics), the current business value (for example, high-frequency access and support for core business data value is higher) and access mode (for example, read more, write less, burst access, etc.) of data assets can be more accurately determined. At the same time, based on the impact of business process iteration information on data processing methods, such as adjustment of data desensitization rules or change of data storage location, re-evaluate its compliance classification (for example, from general data to sensitive data, or from being subject to specific regulations to being subject to more stringent regulations), so as to obtain a real-time detection result reflecting the latest state of data assets.

[0054] The real-time detection results are compared with the policy-related attributes stored in the data asset metadata, and the attribute deviation is calculated, which refers to comparing the latest detected data asset attributes (such as real-time business value, access mode, compliance classification) with the policy-related attributes (such as the initial set data sensitivity, importance, potential compliance risk, etc.) of the data asset pre-stored in the metadata management system. The attribute deviation quantifies the difference between the actual situation and the initial assumption. For example, if the compliance classification of a data asset changes from "general" to "sensitive", the deviation is higher.

[0055] When the attribute deviation exceeds the preset attribute threshold, the policy-related attributes of the data asset are automatically updated. The preset attribute threshold is a configurable parameter that defines the degree of deviation that needs to trigger attribute update. Once the deviation exceeds this threshold, the system will automatically correct the policy-related attributes of the data asset stored in the metadata, ensuring that they always conform to the latest business status.

[0056] According to the updated policy-related attributes and policy generation input information, the resource investment cost required to meet different preset protection levels is re-estimated, and / or real-time updating of the dynamic model is triggered. After the attribute update, the system will use these latest policy-related attributes, combined with the original policy generation input information (such as compliance regulations, cost control targets), to re-calculate the resource investment cost required to meet different protection levels (for example, backup solutions with stricter RPO / RTO requirements). In addition, these updates may directly trigger real-time updating of the dynamic model to ensure that the model always makes policy trade-offs and decisions based on the latest and most accurate data asset status.

[0057] In some embodiments of the present application described above, during the evaluation of the policy-related attributes of the data asset, it is necessary to continuously monitor and analyze business process iteration information. Specifically, continuously monitoring and analyzing business process iteration information can include: S2110: Integrate enterprise service bus and / or message queue to monitor service update messages published by the business system; S2120: Configure log parsing adapters according to service update messages to convert heterogeneous update log information into standardized data structures; S2130: Build an iteration information dependency graph based on the standardized data structure to identify update events with conflicts and / or sequential dependencies; S2140: Use natural language processing techniques to analyze unstructured business documents and change requests to extract keywords and semantic associations of business system function updates, data processing process adjustments, and user access behavior changes, and obtain semantic association results; S2150: Fuse the standardized data structure, the iteration information dependency graph, and the semantic association result to generate a business process overall change report for use in evaluating the policy-related attributes of the data assets.

[0058] Among them, the integrated enterprise service bus and / or message queue aims to establish a unified communication backbone for capturing service update messages from different business systems. The enterprise service bus can be used as middleware to realize message routing, conversion and coordination between different systems; the message queue provides an asynchronous communication mechanism to ensure reliable delivery and decoupling of messages. By listening to these service update messages, real-time access to business system function updates, data processing process adjustments, and user access behavior changes can be obtained. According to the service update message, configure the log parsing adapter, the purpose of which is to uniformly convert the update log information from different sources and formats into a standardized data structure. For example, the adapter can parse logs in XML, JSON, CSV and other formats, and map them to a predefined, structured data format for subsequent automated processing and analysis.

[0059] According to the standardized data structure, an iteration information dependency graph can be constructed. The graph is used to visualize and identify conflicts and / or sequential dependencies of update events in the business process iteration information. For example, if the adjustment of a data processing process depends on the update of a business system function, the graph can explicitly represent this dependency, which helps to consider the update order and impact range in subsequent evaluation. Natural language processing techniques are used to parse unstructured business documents and change requests to extract valuable business process iteration information from text descriptions. For example, natural language processing models can identify keywords and semantic associations such as "add user registration function", "modify order processing logic", "optimize data query performance", etc., to supplement details that may be missed by structured messages, and obtain semantic association results. Finally, by fusing the standardized data structure, the iteration information dependency graph, and the semantic association result, a business process overall change report can be generated. The report comprehensively summarizes the iteration information of the business process, providing comprehensive and accurate input for subsequent evaluation of the policy-related attributes of the data assets.

[0060] In another embodiment of the present application, further proposed is that in the process of real-time detection of the affected data assets, further comprising: S2310: Deploy performance monitoring probes at the infrastructure level to collect read / write latency, IOPS, throughput of storage media, and bandwidth utilization, packet loss rate and link jitter of network devices in real time; S2320: Deploy audit probes at the security policy enforcement level to record execution logs of data desensitization operations and / or trigger events of access control rules in real time; S2330: Correlate and analyze the infrastructure indicators collected by the performance monitoring probe with the access patterns of the data assets, to identify the changes in data asset access performance caused by infrastructure changes; S2340: Correlate and analyze the security policy enforcement logs recorded by the audit probe with the compliance classification of the data assets, to identify the changes in data asset compliance risk caused by security policy changes; S2350: Based on the results of the correlation analysis, revise the real-time detection results of the business value, compliance classification, and / or access patterns of the data assets; S2360: Based on the revised real-time detection results, re-analyze the call frequency, data update frequency, and data access path of the data assets in the business system, determine their current business value and access patterns, and re-evaluate the compliance classification, based on the business system logs and user behavior data.

[0061] Specifically, the performance monitoring probe at the infrastructure level can be understood as a software agent or hardware module deployed on key infrastructure components such as storage devices and network devices, which aims to continuously and finely collect the running status and performance indicators of these components. For example, for storage media, indicators such as read-write delay, input-output operations per second (IOPS), and data throughput can be collected; for network devices, indicators such as bandwidth utilization, packet loss rate, and link jitter can be collected. These indicators directly reflect the performance of the physical or virtual resources on which the data assets depend. The audit probe at the security policy enforcement level refers to a monitoring tool deployed on security components such as data de-sensitization systems and access control systems, which aims to record events related to the execution of data security policies in real time. For example, when data is de-sensitized, the probe records the execution log of the de-sensitization operation, including operation time, operation object, operation result, etc.; when an access control rule is triggered, the probe records the triggering event, including the access, the accessed resource, the rule type, the decision result, etc. These logs and events directly reflect the actual operation of the compliance guarantee measures of the data assets. In practical applications, correlating and analyzing the infrastructure indicators collected by the performance monitoring probe with the access patterns of the data assets means that through data analysis techniques, the performance data of the underlying infrastructure is compared and fused with the access behavior data of the upper layer data assets to identify the potential causal relationship between the two. For example, when the read-write delay of the storage medium suddenly increases, it can be analyzed whether it is associated with the change in access pattern of a certain specific data asset, so as to judge whether the infrastructure change has affected the access performance of the data asset.

[0062] Correlation analysis between the security policy execution log recorded by the audit probe and the compliance classification of the data asset refers to evaluating the impact of the security policy execution on the compliance risk of the data asset by analyzing the execution of the security policy. For example, if the desensitization operation execution log of a certain sensitive data asset shows that the desensitization failure rate is rising, or the access control rules are frequently bypassed or triggered abnormally, it may indicate that the compliance risk of the data asset is increasing.

[0063] According to the results of the above correlation analysis, the real-time detection results of the business value, compliance classification and / or access pattern of the data asset can be corrected. For example, if the infrastructure performance decreases, causing the access performance of a certain data asset to deteriorate significantly, the business value of the data asset may be re-evaluated as reduced; if there is a problem with the execution of the security policy, causing the compliance risk to increase, the compliance classification of the data asset may be adjusted to a stricter level. The corrected real-time detection results will be combined with the business system logs and user behavior data for more comprehensive re-analysis to finally determine the current business value, access pattern of the data asset, and re-evaluate its compliance classification, so as to ensure that the evaluation of the data asset attributes is more accurate and comprehensive.

[0064] In another embodiment of the present application, it is further proposed that when the infrastructure indicators collected by the performance monitoring probe are correlated with the access pattern of the data asset, the following steps are further included: S2331: transmitting the infrastructure indicator stream and the data asset access pattern stream in a streaming manner, and attaching identification information, the identification information including at least the data asset identification and the collection source identification; S2332: respectively attaching a timestamp to each infrastructure indicator sample and each data asset access event at the time of collection, and aligning the infrastructure indicator stream and the data asset access pattern stream based on the timestamps; S2333: performing correlation analysis on the aligned infrastructure indicator stream and the data asset access pattern stream using a sliding time window; S2334: after a fluctuation of the infrastructure indicator exceeding a preset fluctuation threshold, tracking the change trend of the data asset access performance in a subsequent preset time window, and when there is a continuous and consistent change trend in a continuous number of not less than a preset window count threshold, it is determined that there is a causal relationship; S2335: when it is determined that there is a causal relationship, quantifying the influence degree of the infrastructure change on the data asset access performance, and using the quantification result to correct the parameters of the dynamic model, and / or for the policy decision based on the dynamic model.

[0065] Specifically, transmitting infrastructure metrics stream and data asset access pattern stream in streaming manner refers to transmitting the index data collected by the performance monitoring probe at the infrastructure layer (such as the read-write delay of the storage medium, IOPS, throughput, bandwidth utilization of the network device, packet loss rate and link jitter, etc.) and the access pattern data of the data asset (such as the call frequency, data update frequency, data access path, etc.) in real time and continuously as data stream. The identification information, such as the data asset identification, can uniquely identify a specific data asset, and the collection source identification is used to indicate that the data comes from which monitoring probe or system, ensuring the traceability of the data. The attachment of these identification information and time stamp aims to provide accurate context and alignment benchmark for subsequent correlation analysis.

[0066] The time stamp is attached to each infrastructure index sample and each data asset access event during collection, and the infrastructure index stream and the data asset access pattern stream are aligned based on the time stamp, which aims to ensure that different types of data from different sources can be accurately matched in the time dimension. The accuracy of the time stamp is crucial for identifying the sequence and synchronization of events, and through time stamp alignment, the data misplacement problem caused by data transmission delay or system clock difference can be eliminated, laying a foundation for subsequent causal relationship determination.

[0067] Correlation analysis of the aligned infrastructure index stream and the data asset access pattern stream by using a sliding time window refers to defining a fixed-length time window, and letting the window continuously slide on the time axis to perform real-time or near-real-time correlation calculation on the data in the window. This way can dynamically capture the short-term and long-term correlation between data streams, adapt to the changing system behavior, and avoid the lag or inaccuracy that may be caused by static analysis.

[0068] After the infrastructure index appears a fluctuation exceeding a preset fluctuation threshold, the change trend of the data asset access performance in a subsequent preset time window is tracked, and when there is a continuous and consistent change trend in the preset time window for not less than a preset window count threshold, it is determined that there is a causal relationship. The preset fluctuation threshold is used to define the significant change of the infrastructure index, for example, when the storage IOPS decreases by more than a preset percentage, it can be considered as a fluctuation. The preset time window defines the time range for observing the change of the data asset access performance, for example, within a certain number of minutes after the infrastructure fluctuation occurs. The preset window count threshold requires that the consistent change trend appears in multiple consecutive observation windows, for example, the access delay of the data asset continuously increases in multiple preset time windows, which is considered as strong evidence of causal relationship. This multiple condition judgment mechanism aims to improve the accuracy of causal relationship determination and avoid false positives.

[0069] When a causal relationship is determined, the degree of impact of the infrastructure change on the data asset access performance is quantified, and the quantified result is used to correct the parameters of the dynamic model, and / or used for policy decisions made based on the dynamic model. The degree of impact can be quantified by calculating the average amount of change, the maximum amount of change, or the rate of change of performance indicators such as latency, throughput, etc. These quantified results directly reflect the specific impact of infrastructure changes on data asset performance, providing accurate inputs for the above dynamic model, enabling it to more accurately assess risks, costs, and benefits, and thus optimize the generation and adjustment of backup strategies.

[0070] In some preferred embodiments, the following is illustrated by a specific example: Suppose a company's data center is running a critical online transaction system, and its core data assets are stored on a high-performance storage array. The IOPS (Input / Output Operations Per Second) of the storage array is one of the key infrastructure indicators.

[0071] At a certain point in time, the performance monitoring probe detects that the IOPS indicator of the storage array suddenly decreases and continues to exceed the preset fluctuation threshold (for example, IOPS decreases from 100,000 to 80,000, exceeding the 15% fluctuation threshold). At this time, the system will immediately start tracking the access performance of the data assets (e.g., transaction database) associated with the storage array.

[0072] Specifically, the system will transmit the IOPS indicator stream of the storage array and the access latency stream of the transaction database in a streaming manner. During data collection, each IOPS sample and each database access event is attached with an accurate timestamp, and the timestamps are aligned based on these timestamps. Subsequently, a 5-minute sliding time window is used to perform correlation analysis on the aligned data streams.

[0073] When the fluctuation of IOPS decrease occurs, the system starts tracking the change trend of the access latency of the transaction database in the subsequent preset time window (e.g., every 5-minute window). If the system observes that the average access latency of the transaction database continues to increase (e.g., from 10 milliseconds to 20 milliseconds, and then to 30 milliseconds) in consecutive not less than a preset window count threshold (e.g., 3 consecutive 5-minute windows), and this increasing trend is consistent with the direction of IOPS decrease, then the system will determine that there is a causal relationship between the decrease of IOPS of the storage array and the increase of access latency of the transaction database.

[0074] Once the causality is determined, the system quantifies the degree of this impact. For example, it is calculated that for every 10% drop in IOPS, the access latency to the transactional database increases by an average of 10 milliseconds. This quantified result is then used to correct the parameters of the dynamic model described above. For example, the parameter in the dynamic model about the relationship between storage performance and data recovery time is adjusted to more accurately reflect the recovery performance under the current infrastructure conditions. Based on this adjusted dynamic model, the policy decision module may re-evaluate the current backup strategy, for example, consider migrating part of the high-sensitivity data to another storage area with more stable performance, or adjust the backup window to avoid peak hours, so that the compliance and cost control targets of data backup and recovery can still be met when the infrastructure performance fluctuates.

[0075] In another embodiment of the present application, it is further proposed that when quantifying the degree of impact of infrastructure changes on data asset access performance, it further includes: 23351: For the above data asset, identify the type of its access pattern as bursty access pattern and / or periodic access pattern; 23352: When identified as a bursty access pattern, after the infrastructure change occurs, collect the access performance indicators of the above data asset at a preset high-frequency sampling interval until the fluctuation amplitude of the above access performance indicators in a continuous not less than a preset window count threshold for a preset time window does not exceed a preset stability threshold; compare the access performance indicators in the initial observation window after the change with the access performance indicators in the equal-length reference window before the change to obtain the instantaneous difference, and the above instantaneous difference is taken as the instantaneous impact degree; 23353: When identified as a periodic access pattern, after the infrastructure change occurs, continuously collect the access performance indicators of the above data asset for at least one complete cycle, and compare the access performance indicators in the changed one complete cycle with the access performance indicators in the corresponding cycle before the change to calculate the periodic average difference, and the above periodic average difference is taken as the continuous impact degree; 23354: Combine the above instantaneous impact degree and the above continuous impact degree to obtain the quantification result of the impact of the infrastructure change on the access performance of the above data asset, and the above quantification result is used to correct the parameters of the above dynamic model, and / or used for policy decisions based on the above dynamic model.

[0076] Specifically, identifying the access pattern type of a data asset can be based on historical access logs, business system configuration information, or by applying machine learning algorithms to analyze real-time access data. The bursty access pattern is usually characterized by a high concurrency of read-write requests in a short period of time, such as report generation at the end of the month, data processing during a bursty marketing campaign, etc. The periodic access pattern is characterized by repeated access behavior at fixed time intervals, such as daily data synchronization, weekly data backup, monthly bill generation, etc.

[0077] The preset high-frequency sampling interval refers to continuously collecting performance indicators at a short time interval (e.g., seconds or milliseconds) after the infrastructure change occurs, in order to capture the rapid changes of the bursty access pattern. The preset window count threshold and the preset stability threshold are used to determine whether the performance has stabilized, in order to avoid ending the sampling too early due to temporary fluctuations. The calculation of the instantaneous difference aims to quickly assess the immediate impact of the infrastructure change on the bursty access. For the periodic access pattern, performance indicators are continuously collected for at least one complete cycle, in order to ensure that the inherent access regularity can be fully covered, so as to calculate a more representative periodic average difference, reflecting the continuous impact of the change on long-term and regular access. Finally, by combining the instantaneous impact degree and the continuous impact degree, a more comprehensive and accurate quantitative result can be obtained, which can more finely guide the parameter correction and strategy decision of the dynamic model.

[0078] The specific embodiments of the present application also disclose a data backup strategy automatic generation system, comprising: An information acquisition module 1 is configured to acquire and quantify strategy generation input information, and the strategy generation input information comprises compliance regulations of external data and cost control targets of an enterprise.

[0079] An asset evaluation module 2 is configured to evaluate strategy-related attributes of a data asset, and the strategy-related attributes comprise data sensitivity, importance, current backup strategy state, potential compliance risk and recovery performance requirement, and to estimate resource input costs for meeting different preset protection levels according to the strategy generation input information.

[0080] A model construction module 3 is configured to construct and update a dynamic model reflecting a strategy trade-off relationship in real time based on the strategy generation input information, the strategy-related attributes and the resource input costs. The module can be implemented as a model management service, which internally contains various optimization algorithms (e.g., multi-objective optimization algorithm, machine learning model) for constructing and adjusting the dynamic model according to the input data.

[0081] A policy decision module 4 is configured to make a policy decision based on the dynamic model, the policy decision being based on a compliance regulation combined with real-time resource conditions, the real-time resource conditions including storage device available capacity, network data transfer capability and storage device input / output performance.

[0082] A policy deployment and monitoring module 5 is configured to deploy the backup policy and continuously monitor the execution effect of the backup policy, and use actual running data of the execution effect as feedback to adjust parameters of the dynamic model.

[0083] The above merely provides an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for automatically generating a data backup policy, characterized by, The method comprises the following steps: acquiring and quantifying policy generation input information, which comprises compliance regulations of external data and cost control targets of an enterprise; evaluating policy-related attributes of data assets, which comprise data sensitivity, importance, current backup policy status, potential compliance risk, and recovery performance requirements, and estimating resource input costs for meeting different preset protection levels according to the policy generation input information; constructing and updating a dynamic model reflecting policy trade-off relationships in real time based on the policy generation input information, the policy-related attributes, and the resource input costs; making a policy decision based on the dynamic model, which selects a backup policy based on the compliance regulations in combination with real-time resource conditions, the real-time resource conditions comprising available capacity of a storage device, network data transmission capability, and input-output performance of the storage device; deploying the backup policy and continuously monitoring the execution effect of the backup policy, and using actual operation data of the execution effect as feedback to adjust parameters of the dynamic model.

2. The method of claim 1, wherein, In the process of making a policy decision based on the dynamic model and deploying the backup policy, the method further comprises the following steps: collecting and storing micro-pattern characteristics of key network transmission paths, which comprise packet delay, delay jitter, packet interval time distribution, transmission control protocol window size variation pattern, and throughput distribution, to form a normal operation baseline; when a large-scale data migration or backup task is started, real-time sampling and analyzing the micro-pattern characteristics of task data streams to obtain a task flow characteristic sequence; comparing the task flow characteristic sequence with the normal operation baseline to calculate a deviation degree index; when the deviation degree index continuously exceeds a preset deviation threshold within a preset time window, determining that there is an implicit anomaly, triggering an anomaly warning, and providing an anomaly event as input to the policy decision for policy adjustment; prioritizing preset emergency shunting plans according to recovery point target satisfaction, cost increment, and resource availability, the emergency shunting plans comprising alternative transmission paths and backup scheduling schemes; based on the anomaly event, recommending an alternative scheme with the highest priority from the emergency shunting plans, executing the alternative scheme, and adjusting parameters of the dynamic model based on the execution result of the alternative scheme as feedback.

3. The method of claim 1, wherein, In the process of making a policy decision based on the dynamic model, the method further comprises the following steps: establishing a performance isolation threshold library of shared infrastructure components, which records maximum concurrent data stream quantity and / or input-output performance of each shared infrastructure component under different business loads and interference characteristics of specific types of data streams on performance of other businesses; simulating resource occupation patterns of candidate schemes on the shared infrastructure components according to task size, data stream type, and scheduling time of candidate large-scale data migration or backup tasks; matching the resource occupation mode with interference features in the performance isolation threshold library, marking the candidate scheme as a high-risk scheme when a matching result predicts that the candidate scheme will cause resource occupation of a shared infrastructure component to exceed a performance isolation threshold and / or to overlap with the interference features, and calculating an interference penalty cost; counting the interference penalty cost into total cost input of the candidate scheme in the dynamic model, and selecting a backup strategy from candidate schemes that meet the compliance regulations based on adjusted total cost input and in combination with the real-time resource status.

4. The method of claim 1, wherein, In the process of deploying the backup strategy and continuously monitoring the execution effect of the backup strategy, further comprising: establishing a business scenario recovery path diagram, the business scenario recovery path diagram being used to map the internal structure, mutual dependency relationship of data assets and the application environment supported thereby; when a business scenario recovery verification process is performed, tracing and recording a complete path and corresponding time from data recovery to end-to-end function recovery of a business system based on the business scenario recovery path diagram; analyzing the complete path and corresponding time, identifying key dependent links and bottlenecks in the actual recovery path, and quantifying the influence of each key dependent link and bottleneck on the overall business recovery time; comparing the overall business recovery time with a recovery time target and / or a recovery point target preset by a business department, and calculating a deviation degree; when the deviation degree exceeds a preset time deviation threshold, identifying actual deviation between actual recovery performance and business expectation caused by changes in internal structure, mutual dependency relationship or application environment; converting the actual deviation into a business scenario recovery benefit correction factor, which is used to adjust a benefit function parameter in the dynamic model.

5. The method of claim 1, wherein, In the process of evaluating the policy-related attributes of data assets, further comprising: continuously monitoring and analyzing business process iteration information, the business process iteration information including business system function updates, data processing process adjustments and user access behavior changes; identifying affected data assets and their associated business scenarios according to the business process iteration information; for the affected data assets, analyzing the calling frequency, data update frequency and data access path of the data assets in the business system, determining their current business value and access mode in combination with business system logs and user behavior data, and reevaluating their compliance classification based on the influence of the business process iteration information on data processing methods to obtain real-time detection results; comparing the real-time detection results with policy-related attributes stored in data asset metadata, and calculating an attribute deviation degree; when the attribute deviation degree exceeds a preset attribute threshold, automatically updating the policy-related attributes of the data assets; and reestimating resource input costs required to meet different preset protection levels and / or triggering real-time updating of the dynamic model according to the updated policy-related attributes and the policy generation input information.

6. The method of claim 5, wherein, The continuous monitoring and analysis of business process iteration information further comprises: integrating enterprise service buses and / or message queues to monitor service update messages published by business systems; According to the service update message, a log parsing adapter is configured to convert heterogeneous update log information into a standardized data structure; According to the standardized data structure, an iterative information dependency graph is constructed to identify update events that exist conflicts and / or sequential dependencies; Using natural language processing techniques, unstructured business documents and change requests are parsed to extract keywords and semantic associations of business system function updates, data processing flow adjustments, and user access behavior changes, resulting in semantic association results; Fusion of the standardized data structure, the iterative information dependency graph and the semantic association results, generate business process overall change report, for use in evaluating the policy-related attributes of data assets.

7. The method of claim 5, wherein, In the process of real-time detection of the affected data assets, further comprising: Deploying infrastructure-level performance monitoring probes to collect real-time storage media read / write latency, IOPS, throughput, and network device bandwidth utilization, packet loss rate and link jitter; Deploying security policy enforcement layer audit probes to record real-time data de-identification operation execution logs and / or access control rule trigger events; Correlate and analyze the infrastructure indicators collected by the performance monitoring probes with the access patterns of the data assets to identify changes in data asset access performance caused by infrastructure changes; Correlate and analyze the security policy enforcement logs recorded by the audit probes with the compliance classification of the data assets to identify changes in data asset compliance risk caused by security policy changes; According to the results of the correlation analysis, correct the real-time detection results of the business value, compliance classification and / or access pattern of the data assets; And based on the corrected real-time detection results, combined with business system logs and user behavior data, re-analyze the call frequency, data update frequency and data access path of the data assets in the business system to determine their current business value and access pattern, and re-evaluate the compliance classification.

8. The method of claim 7, wherein, In the correlation analysis of the infrastructure indicators collected by the performance monitoring probes and the access patterns of the data assets, further comprising: Streaming infrastructure indicator streams and data asset access pattern streams with identification information, including at least data asset identification and collection source identification; When collecting, respectively, each infrastructure indicator sample and each data asset access event, add a timestamp, and align the infrastructure indicator stream and the data asset access pattern stream based on the timestamp; Correlate and analyze the aligned infrastructure indicator stream and data asset access pattern stream using a sliding time window; After the infrastructure indicators fluctuate beyond a preset fluctuation threshold, track the change trend of data asset access performance in the subsequent preset time window, and when there is a consistent change trend in consecutive not less than a preset window count threshold, determine that there is a causal relationship; When the causal relationship is determined, quantify the impact of infrastructure changes on the data asset access performance, and use the quantification results to correct the parameters of the dynamic model, and / or for policy decisions based on the dynamic model.

9. The method of claim 8, wherein, In quantifying the degree of influence of the infrastructure change on the data asset access performance, further comprising: For the data asset, identifying the type of its access pattern as a bursty access pattern and / or a periodic access pattern; When identified as a bursty access pattern, after the infrastructure change occurs, collecting the access performance indicators of the data asset at a preset high-frequency sampling interval until the fluctuation range of the access performance indicators within a continuous preset time window of no less than a preset window count threshold does not exceed a preset stability threshold; comparing the access performance indicators in the initial observation window after the change with the access performance indicators in an equal-length reference window before the change to obtain an instantaneous difference, which is taken as an instantaneous degree of influence; When identified as a periodic access pattern, after the infrastructure change occurs, continuously collecting the access performance indicators of the data asset for at least one complete period, and comparing the access performance indicators in the complete period after the change with the access performance indicators in the corresponding period before the change to calculate a periodic average difference, which is taken as a continuous degree of influence; Combining the instantaneous degree of influence and the continuous degree of influence, obtaining a quantification result of the degree of influence of the infrastructure change on the data asset access performance, which is used to correct the parameters of the dynamic model and / or used for policy decision based on the dynamic model.

10. A data backup policy automatic generation system characterized by comprising: Comprise: An information acquisition module for acquiring and quantifying policy generation input information, the policy generation input information including compliance regulations of external data and cost control targets within an enterprise; An asset evaluation module for evaluating policy-related attributes of data assets, the policy-related attributes including data sensitivity, importance, current backup policy state, potential compliance risk, and recovery performance requirement, and estimating resource input costs for meeting different preset protection levels according to the policy generation input information; A model construction module for constructing and real-time updating a dynamic model reflecting policy trade-off relationships based on the policy generation input information, the policy-related attributes, and the resource input costs; A policy decision module for making policy decisions based on the dynamic model, the policy decisions selecting a backup policy based on the compliance regulations combined with real-time resource conditions, the real-time resource conditions including storage device available capacity, network data transmission capability, and storage device input / output performance; A policy deployment and monitoring module for deploying the backup policy and continuously monitoring the execution effect of the backup policy, taking actual running data of the execution effect as feedback for adjusting parameters of the dynamic model.

Images