A data configuration method and device, a storage medium and an electronic device

Abstract

The embodiment of the specification discloses a data configuration method and device, a storage medium and an electronic device, wherein the method comprises: obtaining a service data load size to be transferred in a target period, an internal demand prediction deviation factor and an external environment disturbance parameter of data transfer, and determining a lower limit of a basic buffer data quota and an upper limit of a data service target buffer in combination with a preset safety threshold; determining an internal demand buffer data amount and an external environment buffer adjustment amount based on the internal demand prediction deviation factor, the external environment disturbance parameter and the lower limit to obtain a buffer reserve data amount; constructing a theoretical data operation water level based on the lower limit and the buffer reserve data amount, and performing data water level limiting and bottom configuration processing using the upper limit and the lower limit in sequence to obtain a dynamic resource configuration target value, and configuring data redundancy for a data resource pool according to the dynamic resource configuration target value. The embodiment of the specification realizes high-reliability and dynamic precise configuration of data transfer resources.

Description

Technical Field

[0001] This specification relates to the field of computer technology, and in particular to a data configuration method, apparatus, storage medium, and electronic device. Background Technology

[0002] In modern resource data scheduling and management scenarios, resource data can be any schedulable, quantifiable, cost-effective, and quota-constrained flow of data. Whether it's processing financial resource data streams, allocating power resource loads, or managing carbon resource emission quotas, resource data processing faces a common data transfer task (i.e., transferring data from one node to another). A common scenario in data transfer tasks is that service platforms that provide data services (i.e., service platforms associated with "platform data accounts") need to continuously acquire and schedule resource data such as computing power data, power data, carbon data, and financial data from multiple data source nodes (i.e., data source nodes associated with "institutional data accounts") to meet the service platform's dynamically changing data compliance goals (such as data buffer pool levels, grid security margins, or carbon compliance goals), thereby better providing data services to users.

[0003] In real-world scenarios, service platforms face increasingly demanding precise and dynamic resource allocation when dealing with large-scale data transfer tasks. Traditional resource allocation methods typically rely on historical static experience values ​​or fixed redundancy reserves. However, during actual business execution, data transfer tasks often encounter multi-dimensional uncertainties: on the one hand, internal demand forecasts may be inaccurate due to limitations in algorithmic logic or data sample bias; on the other hand, random fluctuations in the external communication environment and hardware load can significantly impact data transmission efficiency. Summary of the Invention

[0004] This specification provides a data configuration method, apparatus, storage medium, and electronic device, which achieves highly reliable, dynamic, and precise configuration of data transfer resources. The technical solution is as follows: Firstly, embodiments of this specification provide a data configuration method applied to a service platform, the method comprising: The system obtains the business data load scale to be transferred within the target time period, the internal demand prediction deviation factor for data transfer, and the external environmental disturbance parameters for data transfer. Based on the preset security threshold and the business data load scale, it determines the lower limit of the basic buffer data quota and the upper limit of the target buffer for data transfer service operation. The amount of internal demand buffer data is determined based on the data transfer internal demand prediction deviation factor and the lower limit of the basic buffer data quota. The amount of external environment buffer data adjustment is determined based on the external environment disturbance parameters and the lower limit of the basic buffer data quota. The amount of internal demand buffer data and the amount of external environment buffer data adjustment are combined to obtain the amount of buffer reserve data. Based on the lower limit of the basic buffer data quota and the amount of buffer backup data, a theoretical data operation level for data transfer service is constructed. The upper limit of the data service target buffer is used to perform data level limiting processing on the theoretical data operation level to obtain a reference data reservation result. The lower limit of the basic buffer data quota is used to perform a bottom-line configuration processing on the reference data reservation result to obtain a dynamic resource configuration target value. Data redundancy configuration is performed on the data resource pool for data transfer services based on the dynamic resource configuration target value.

[0005] Secondly, embodiments of this specification provide a data configuration device, the device comprising: The data processing module is used to obtain the business data load scale to be transferred within the target time period, the internal demand prediction deviation factor of data transfer, and the external environmental disturbance parameters of data transfer. Based on the preset security threshold and the business data load scale, it determines the lower limit of the basic buffer data quota and the upper limit of the data business target buffer for the operation of the data transfer service. The inference module is configured to determine the amount of internal demand buffer data based on the data transfer internal demand prediction deviation factor and the lower limit of the basic buffer data limit, determine the amount of external environment buffer data adjustment based on the external environment disturbance parameters and the lower limit of the basic buffer data limit, and comprehensively process the amount of internal demand buffer data and the amount of external environment buffer data adjustment to obtain the amount of buffer reserve data. The configuration inference module is used to construct the theoretical data operation level of the data transfer service based on the lower limit of the basic buffer data quota and the amount of buffer backup data, perform data level limiting processing on the theoretical data operation level using the upper limit of the data service target buffer to obtain the reference data reservation result, and use the lower limit of the basic buffer data quota to perform bottom-line configuration processing on the reference data reservation result to obtain the dynamic resource configuration target value. The reserved configuration module is used to configure data redundancy for data transfer services of the data resource pool according to the dynamic resource configuration target value.

[0006] Thirdly, embodiments of this specification provide a computer storage medium storing a plurality of instructions adapted for loading by a processor and executing the above-described method steps.

[0007] Fourthly, this specification provides a computer program product storing at least one instruction adapted to be loaded by a processor and to execute the method steps of one or more embodiments of this specification.

[0008] Fifthly, this specification provides a computer program product storing at least one instruction adapted to be loaded by a processor and to execute the method steps of one or more embodiments of this specification.

[0009] Fifthly, embodiments of this specification provide an electronic device that may include: a processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to execute the above-described method steps.

[0010] The beneficial effects of the technical solutions provided in some embodiments of this specification include at least the following: In one or more embodiments of this specification, by comprehensively considering the scale of business data load, internal demand prediction deviation factors, and external environmental disturbance parameters, the accurate quantification and decoupling of intrinsic algorithm errors and extrinsic physical environment noise during the data transfer process are achieved, effectively overcoming the technical defects of traditional single static configuration methods that are difficult to adapt to dynamic and complex environments. On this basis, a theoretical data operation level is constructed, and a two-way constraint mechanism is introduced, which combines level limiting processing based on the upper limit of the business target buffer with bottom-line configuration processing based on the lower limit of the basic buffer quota. This mechanism not only uses the upper limit constraint to strongly prevent excessive resource reservation and waste caused by extreme disturbance calculations, but also uses the lower limit to ensure the minimum survivability and basic availability of the data transfer service under harsh conditions, thereby determining a dynamic balance point between resource constraints and the high stability requirements of the business. Finally, based on this dynamic resource configuration target value, the data resource pool is accurately redundantly configured, improving the system robustness and resource utilization efficiency of the service platform when facing high concurrency load impacts and complex physical environment fluctuations, ensuring the continuity, security, and economy of large-scale data transfer services. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart illustrating a data configuration method provided in an embodiment of this specification; Figure 2This is a flowchart illustrating the process of determining the amount of internal demand buffer data provided in the embodiments of this specification; Figure 3 This is a flowchart illustrating the determination of external environment buffer data adjustment amount provided in the embodiments of this specification; Figure 4 This is a schematic diagram of a process based on configuration parameter adjustment provided in the embodiments of this specification; Figure 5 This is a schematic diagram of the structure of a data configuration device provided in the embodiments of this specification; Figure 6 This is a schematic diagram of the structure of an electronic device provided in the embodiments of this specification. Detailed Implementation

[0013] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this specification.

[0014] In the description of this specification, it should be understood that the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. In the description of this specification, it should be noted that, unless otherwise expressly specified and limited, "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices. Those skilled in the art can understand the specific meaning of the above terms in this specification based on the specific circumstances. Furthermore, in the description of this specification, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship.

[0015] In related technologies, data allocation mechanisms often lack comprehensive quantitative analysis methods for the aforementioned internal and external disturbances, making it difficult to balance stability and economy in resource reservation schemes. Excessive redundancy leads to low utilization of the data resource pool, resulting in significant waste of computing power and bandwidth; insufficient redundancy, on the other hand, easily causes task backlogs, response timeouts, or even system service interruptions when facing sudden environmental disturbances or demand fluctuations. Therefore, how to achieve dynamic allocation of data transfer resources with both high robustness and high utilization in a multi-dimensional disturbance environment has become a key technical challenge that current service platforms urgently need to solve.

[0016] The present specification will now be described in detail with reference to specific embodiments.

[0017] In one embodiment, such as Figure 1 As shown, a data configuration method is proposed, which can be implemented using a computer program and run on a data configuration device based on the von Neumann architecture. This computer program can be integrated into applications or run as a standalone utility application. The data configuration device can be a service platform.

[0018] The data transfer business processing scenarios described in one or more embodiments of this specification are primarily aimed at service platforms (such as cloud storage migration gateways, distributed data resource computing scheduling centers, and financial data clearing systems) that need to cope with high concurrency, fluctuating loads, and complex physical environment interference requiring continuous data resource scheduling (such as data streams, power loads, carbon quotas, etc.). In this scenario, the core task of the service platform is to accurately configure the underlying "data resource pool" (including, but not limited to, storage IOPS, network bandwidth slicing, or memory buffers in some resource data scheduling scenarios) for the "business data load" to be processed within a target time period (such as peak business windows or specific batch processing cycles). The core purpose of this configuration is to ensure that system resources are always maintained at a dynamic balance point that can both absorb the current load and withstand potential risks. This balance point is the "dynamic resource configuration target value" (S106) in this specification, designed to prevent data backlog and packet loss due to insufficient resource reservation, or resource idleness and waste due to excessive reservation.

[0019] However, in this specification, the data configuration process is influenced by a series of complex and dynamically changing constraints, specifically including: This specification introduces a load baseline constraint: the starting point for configuration decisions depends on the "business data load scale" (S102). However, relying solely on static load cannot cope with fluctuations. Based on this, a "basic buffer data quota lower limit" is set, which is the rigid resource bottom line required to maintain the minimum availability of services.

[0020] This specification introduces an intrinsic uncertainty constraint: business demands are often not entirely predictable. Due to the lag in data transfer volume forecasting or sample bias, the platform faces the challenge of the "internal demand forecasting deviation factor for data transfer" (S102 / S104). That is, the actual business throughput may deviate from the forecast trajectory at any time. Based on this, this specification introduces the concept of reserving "internal demand buffer data volume" to accommodate the uncertainty of data resource usage.

[0021] This specification introduces external environmental constraints: data transfer depends on physical links, and the physical environment has inherent noise. Based on this, this specification introduces "external environmental disturbance parameters for data transfer" (S102 / S104) to characterize uncontrollable environmental factors such as network jitter, hardware aging, or third-party interface latency. External disturbances reduce the actual output efficiency per unit of resource, forcing the platform to calculate "external environment buffer data adjustment amount" for compensation.

[0022] This specification introduces a two-way water level boundary constraint: resource allocation cannot be infinitely expanded or arbitrarily compressed. On the one hand, limited by physical capacity or cost budget, the "data service target buffer upper limit" (S102) dictates that if any theoretically calculated "theoretical data operation water level" (S106) exceeds this value, "data water level limiting processing" (S106) must be performed to avoid system crashes. On the other hand, to prevent excessive negative adjustment from causing service paralysis, a basic lower limit must be used for "safety net configuration processing" (S106).

[0023] Therefore, the solution in this specification is to intelligently construct a "theoretical data operating level" and derive the optimal "dynamic resource allocation target value" in a complex environment with multi-source disturbances (internal deviations, external environment) and two-way boundary constraints (lower limit support, upper limit amplitude), thereby achieving high robustness and high utilization of the data resource pool.

[0024] Specifically, the data configuration method includes: S102: Obtain the business data load scale to be transferred within the target time period, the internal demand prediction deviation factor for data transfer, and the external environmental disturbance parameters for data transfer. Based on the preset security threshold and the business data load scale, determine the lower limit of the basic buffer data quota and the upper limit of the target buffer for data transfer service operation, respectively. The target time period refers to the specific time window in which the service platform plans to execute the data transfer task. This time period can be a preset fixed period (such as the business off-peak period from 02:00 to 04:00 every day), or a dynamic time slice based on real-time event triggers (such as the next 30 minutes after a storage water level alarm event is detected).

[0025] Business data load scale refers to the amount of business resource data (such as data streams, power load, carbon quotas, etc.) that is expected to require continuous data resource scheduling within the target period.

[0026] The data transfer internal demand forecasting deviation factor is used to quantify the internal uncertainty of the data configuration process. Due to the randomness of business traffic, there is an error between the predicted data configuration value and the actual data configuration value. This data transfer internal demand forecasting deviation factor characterizes the endogenous error rate within the platform.

[0027] The external environmental disturbance parameter for data transfer is used to quantify the degree to which the physical environment impairs data transmission efficiency. This parameter comprehensively reflects the degree of external interference such as network congestion and service delays in data transfer.

[0028] The minimum basic buffer data limit refers to the minimum data resource reserve level of the platform to ensure that data transfer services are not interrupted or severely backlogged.

[0029] The upper limit of the data service target buffer refers to the maximum level of reserved data resources that the data transfer task is allowed to occupy, which is limited by objective factors such as the total capacity of physical resources, cost budget, and global resource quota.

[0030] In one feasible implementation, a static determination method based on preset rules can be adopted. The service platform can establish a connection with the historical business database and determine the business data load scale to be transferred within the target period by querying the average business traffic data within the same time window for a specific period in the past (e.g., the past 30 days). At the same time, the system's preset parameter configuration table is called to match the corresponding empirical values ​​according to the business type of the data to be transferred (e.g., log data, transaction data, or backup data). Based on the empirical values, the internal demand prediction deviation factor for data transfer and the external environmental disturbance parameters for data transfer are obtained respectively. Based on this, according to the pre-set security strategy, a basic security coefficient greater than the reference value (e.g., 1.2) is selected as the preset security threshold. The business data load scale is multiplied by this basic security coefficient, that is: basic buffer data limit lower limit = business data load scale * preset security threshold, thereby determining the basic buffer data limit lower limit to ensure the minimum availability of the service. At the same time, in order to prevent a single task from excessively consuming resources, the service platform selects a preset resource circuit breaker coefficient (e.g., 2.0), and multiplies the business data load scale by this resource circuit breaker coefficient, that is: data service target buffer upper limit = business data load scale * resource circuit breaker coefficient, or reads the maximum quota value currently allocated to this business segment by the current resource pool to determine the data service target buffer upper limit.

[0031] In one feasible implementation, a dynamic determination method based on real-time monitoring can be adopted. The service platform accesses the metadata stream of a real-time message queue, calculates the backlog of tasks at the current moment based on the metadata stream, and combines this with time-series prediction models (such as Large Language Modeling, LLM) to extrapolate the business data load scale within the future target time period. Simultaneously, the real-time computing platform calculates the variance between the predicted and actual values ​​at the previous time step to generate an internal demand prediction deviation factor. Furthermore, network probes measure environmental characteristics such as round-trip latency and packet loss rate in real time to generate external environmental disturbance parameters. When determining boundaries, instead of relying on fixed coefficients, the minimum resource amount required to maintain the service level can be calculated backward from the maximum allowable latency time specified in the service level agreement, and this is set as the lower limit of the basic buffer data quota. Further, the remaining idle capacity of the current global resource pool is scanned, and combined with the concurrency preemption probability of other high-priority business types, a maximum allowable resource occupancy threshold at the current moment is dynamically calculated and set as the upper limit of the data service target buffer. This processing method allows the determined upper and lower limit boundaries to adaptively float with the overall system load status, thereby adapting to highly volatile business scenarios.

[0032] S104: Determine the amount of internal demand buffer data based on the data transfer internal demand prediction deviation factor and the lower limit of the basic buffer data quota; determine the amount of external environment buffer data adjustment based on the external environment disturbance parameters and the lower limit of the basic buffer data quota; and comprehensively process the amount of internal demand buffer data and the amount of external environment buffer data adjustment to obtain the amount of buffer reserve data. The internal demand buffer data volume refers to the amount of additional data resources that need to be reserved to cope with the deviation between the platform's forecast results and the actual business, based on the data transfer internal demand forecast deviation factor.

[0033] External environment buffer data adjustment amount refers to the amount of data resources determined to compensate for changes in the physical environment (such as network speed reduction, IOPS decrease) based on external environmental disturbance parameters for data transfer.

[0034] The buffer backup data volume refers to the total amount of redundant resources determined after integrating the aforementioned internal requirements with the impact of the external environment. It is a safety margin prepared by the system to ensure service continuity even in the worst-case scenario (i.e., when both the forecast is low and the environment is harsh).

[0035] In one feasible implementation, the lower limit of the basic buffer data quota is used as a benchmark value, and linear multiplication is performed with both the internal demand forecasting deviation factor and the external environmental disturbance parameter for data transfer. Specifically, the lower limit of the basic buffer data quota is multiplied by the internal demand forecasting deviation factor to obtain the internal demand buffer data amount; simultaneously, the lower limit of the basic buffer data quota is multiplied by the external environmental disturbance parameter to obtain the external environmental buffer data adjustment amount. After obtaining these two independent buffer components, a linear superposition method is used for comprehensive processing, that is, the internal demand buffer data amount and the external environmental buffer data adjustment amount are directly added together to obtain the final buffer reserve data amount.

[0036] S106: Based on the lower limit of the basic buffer data quota and the amount of buffer spare data, construct the theoretical data operation level of the data transfer service, use the upper limit of the data service target buffer to perform data level limiting processing on the theoretical data operation level to obtain the reference data reservation result, and use the lower limit of the basic buffer data quota to perform bottom-line configuration processing on the reference data reservation result to obtain the dynamic resource configuration target value. The theoretical data operation level refers to the ideal total amount of reserved resources calculated based on basic security needs and internal and external disturbances, without considering the upper limit of data service target buffer.

[0037] Data level limiting refers to a resource protection mechanism that forcibly cuts off the demand to the maximum allowable value when the theoretical demand exceeds the maximum available quota.

[0038] The fallback configuration handling refers to a safety net mechanism that ensures the final allocated resource amount will not fall below the minimum standard required to maintain basic service operation. Even under extreme limiting or negative adjustment logic, this fallback configuration handling mechanism can guarantee that the business has a certain amount of resources.

[0039] The target value for dynamic resource configuration refers to the final data resource redundancy configuration value for the data resource pool after a series of calculations, verifications, and corrections.

[0040] In one feasible implementation, the theoretical data operation level for constructing the data transfer service is obtained by superimposing the lower limit of the basic buffer data quota and the amount of buffer backup data. The upper limit of the data service target buffer is introduced as a comparison benchmark. Data level limiting processing is performed on the theoretical data operation level. Specifically, it is determined whether the theoretical level is greater than the target buffer upper limit. If so, the target buffer upper limit is taken as the reference data reservation result; otherwise, the theoretical level is retained as the reference data reservation result. Furthermore, in order to prevent insufficient resource supply due to abnormal parameters (such as the target buffer upper limit being set incorrectly and thus lower than the basic lower limit), the reference data reservation result is configured with a safety net using the lower limit of the basic buffer data quota. That is, the reference data reservation result is compared with the lower limit of the basic buffer data quota, and the larger of the two is taken as the final dynamic resource configuration target value.

[0041] Specifically, the process of using the upper limit of the data service target buffer to limit the theoretical data operating level to obtain a reference data reservation result, and then using the lower limit of the basic buffer data quota to perform a safety net configuration on the reference data reservation result to obtain a dynamic resource configuration target value, can be performed as follows: Step S2: Compare the theoretical data operating level with the upper limit of the data service target buffer. If the theoretical data operating level is greater than the upper limit of the data service target buffer, the reference data reservation result is determined as the upper limit of the data service target buffer. If the theoretical data operating level is less than or equal to the upper limit of the data service target buffer, the reference data reservation result is determined as the theoretical data operating level. Step S4: Compare the reference data reservation result with the lower limit of the basic buffer data quota, and select the maximum value between the reference data reservation result and the lower limit of the basic buffer data quota as the target value for dynamic resource allocation.

[0042] S108: Configure data redundancy for data transfer service in the data resource pool according to the dynamic resource configuration target value.

[0043] A data resource pool refers to the collection of underlying physical or virtual resources that support the operation of data transfer services.

[0044] In this embodiment, data redundancy configuration refers to reserving or adding corresponding data resource reserves to the resource pool based on the calculated dynamic resource configuration target value. This configuration ensures that even under load fluctuations or environmental interference, the service platform still has sufficient redundancy to handle data transfer services, preventing service overload.

[0045] For example, taking a data flow scheduling scenario, the "business data load" is represented as a binary bit stream, and the data resource pool is represented as a collection of computing, storage, and network resources. The specific configuration process is as follows: After the service platform obtains the dynamic resource configuration target value (e.g., a transmission bandwidth of 1500MB / s and a write capability of 5000IOPS) after water level limiting and bottom-line correction, it performs the following redundant configuration on the underlying physical resources: Network resource dimension: The interface of the software-defined network controller is invoked to modify the traffic shaping strategy for this data transfer task. Specifically, the promised burst rate parameter for this task is adjusted to match the target value of the dynamic resource configuration, ensuring that the network channel has sufficient bandwidth margin to accommodate instantaneous data surges and prevent packet loss. Storage resource dimension: Instructions are sent to the target storage cluster to perform pre-allocation operations at the file system level. That is, before the data actually arrives, contiguous storage blocks of the same size as the target value are pre-defined or zeroed on the disk, and write permissions for this area are locked to prevent write delays caused by disk fragmentation or insufficient space. Computing resource dimension: The elastic scaling group threshold of the data decompression / verification microservice is adjusted, and the CPU load threshold triggering expansion is lowered according to the target value, thereby preheating the computing nodes and achieving redundant readiness of computing resources.

[0046] In the embodiments described in this specification, by comprehensively considering the scale of business data load, internal demand prediction deviation factors, and external environmental disturbance parameters, the accurate quantification and decoupling of intrinsic algorithm errors and extrinsic physical environment noise during the data transfer process are achieved. This effectively overcomes the technical shortcomings of traditional single static configuration methods, which are difficult to adapt to dynamic and complex environments. On this basis, a theoretical data operation level is constructed, and a two-way constraint mechanism is introduced, which combines level limiting processing based on the upper limit of the business target buffer with bottom-line configuration processing based on the lower limit of the basic buffer quota. This mechanism not only uses the upper limit constraint to strongly prevent excessive resource reservation and waste caused by extreme disturbance calculations, but also uses the lower limit to ensure the minimum survivability and basic availability of the data transfer service under harsh conditions, thereby determining a dynamic balance point between resource constraints and the high stability requirements of the business. Finally, based on this dynamic resource configuration target value, the data resource pool is accurately configured with redundancy, which improves the system robustness and resource utilization efficiency of the service platform when facing high concurrency load impacts and complex physical environment fluctuations, ensuring the continuity, security, and economy of large-scale data transfer services.

[0047] Please see Figure 2 , Figure 2 This is a flowchart illustrating a method for determining the amount of internal demand buffer data as described in this specification. Specifically, the method for determining the amount of internal demand buffer data based on the data transfer internal demand forecasting deviation factor and the lower limit of the basic buffer data amount can be referenced as follows: S202: In the preset internal factor mapping relationship corresponding to the data transfer internal demand prediction deviation factor, the target time domain compensation coefficient is matched according to the first internal mapping time label corresponding to the target time period, and the target time domain compensation coefficient is determined as the current internal factor parameter value of the data transfer internal demand prediction deviation factor; Pre-defined internal factor mapping relationships refer to a type of relational data structure that is pre-stored in the service platform's database or memory. It describes the correspondence between temporal characteristics and the degree of prediction bias.

[0048] The first internal mapping time tag refers to the index identifier used for retrieval in the above mapping relationship. It is a time classification tag after business logic processing, such as "09:00-11:00 (peak data transfer window)" and "02:00-04:00 (cold backup idle window)".

[0049] Target time-domain compensation coefficient: refers to the prediction bias correction value determined at a specific time label. The larger the value, the higher the risk of prediction inaccuracy within that period, and the larger the buffer that needs to be reserved.

[0050] In one feasible implementation, the target time period for data transfer is identified (e.g., 10:00:00 on October 1, 2023). A time parser is invoked to map this physical time to a first internally mapped time label in the business logic. For example, this time point is determined to belong to the weekday-morning peak label. A preset internal factor mapping relationship (i.e., a pre-defined strategy configuration table) is accessed, which stores the corresponding entries of different time labels and compensation coefficients. Based on the first internally mapped time label, a precise match is performed in the table, and the corresponding value is extracted as the target time-domain compensation coefficient. This coefficient is then determined as the current internal factor parameter value of the data transfer internal demand prediction deviation factor.

[0051] S204: Extract the backup data resource fluctuation index corresponding to the target time period, and select the maximum value of the backup data resource fluctuation among the backup data resource fluctuation index; Backup data resources: refers to the portion of the data resource pool that is currently not occupied and can theoretically be used to undertake new tasks.

[0052] Backup data resource fluctuation index: A quantitative parameter used to characterize the stability of idle resources within a corresponding indicator observation window (such as the current quarter window, the same period last year window, and a future time period window). It reflects the random fluctuation amplitude of available resources over a time observation window in the absence of explicit sudden business events. This random fluctuation amplitude is the fluctuation index, and different indicator observation windows correspond to different backup data resource fluctuation indices.

[0053] The maximum fluctuation value of backup data resources refers to the highest peak value of the aforementioned fluctuation amplitude within the statistical sample interval composed of all backup data resource fluctuation indicators. Selecting the maximum value reflects the principle of maximizing risk coverage in the security design of this scheme.

[0054] In one feasible implementation, a target time period (e.g., 10:00-11:00 daily) is determined, and using this time period as an index, the system operation logs of several past indicator observation windows (e.g., the past 7 days, 30 days) within the same time period are retrieved back. The theoretical idle resource value and the actual available resource value for each sampling point are extracted from the logs. The difference between the two is defined as the standby data resource fluctuation index, or the ratio of the difference to the theoretical idle resource value is defined as the standby data resource fluctuation index. The set of these differences is defined as the standby data resource fluctuation index. Subsequently, this index set is sorted or scanned traversally, and the largest difference is selected as the maximum standby data resource fluctuation value.

[0055] In one feasible implementation, specifically performing the extraction of the backup data resource fluctuation index corresponding to the target time period and selecting the maximum value of the backup data resource fluctuation among the backup data resource fluctuation index can be done in the following way: Step A2: Extract the volatility of the first historical backup data resource within the preset historical quarterly cycle of the target time period; The preset historical quarterly period refers to the current quarter or the most recent statistical quarter (e.g., Q1, Q2, etc.) in which the target period falls. This period reflects the overall operating status and environmental characteristics of the system in the recent period (e.g., whether system upgrades or expansions have been carried out recently). The first historical backup data resource volatility refers to the ratio of the average or maximum fluctuation of the system resource level to the benchmark level within the aforementioned quarterly period, which represents the recent trend.

[0056] Indicatively, identify the calendar quarter to which the target period belongs (e.g., the fourth quarter, Q4), query all historical operation logs from the start of the quarter to the current day, calculate the dispersion (e.g., coefficient of variation) of the data resource buffer level for each day within the period based on the historical operation logs, and determine the mean or maximum value as the first historical backup data resource volatility.

[0057] Step A4: Extract the volatility of the second historical backup data resource corresponding to the same period of the previous year for the target period; The volatility of the second historical backup data resource refers to the ratio of fluctuations in the data resource buffer level within the same time window of the previous year, which represents the periodic / seasonal pattern. For example, going back to the previous year, we identify the historical date range that is the same as the target period (e.g., the same month and day of last year), retrieve the archived data for that historical date range, query the resource level fluctuation ratio at that time based on the archived data, and determine it as the second historical backup data resource volatility rate.

[0058] Step A6: Determine the cumulative value of the predicted load fluctuation for the target time period within the current cycle for the business data load scale, and calculate the future time period volatility based on the cumulative value of the predicted load fluctuation for the business data load scale. The cumulative value of predicted load fluctuations refers to the total amount obtained by integrating or accumulating the changes in data resource buffers at each time step within a future period using a time series prediction algorithm.

[0059] Future volatility refers to the ratio of potential future resource instability calculated based on forecast data.

[0060] In a schematic manner, historical data is used to perform mathematical modeling to generate a data resource buffer curve for the target time period, and the cumulative predicted load fluctuation value of this curve relative to the preset data resource buffer mean (i.e., the sum of the absolute values ​​of all predicted points deviating from the benchmark value) is calculated. The business data load scale within the target time period is weighted and averaged (e.g., the load closer to the current time has a higher weight) to obtain the weighted average load value. The cumulative predicted load fluctuation value is then divided by the weighted average load value to normalize and obtain the volatility for the future time period.

[0061] Step A8: Select the maximum value among the first historical backup data resource volatility, the second historical backup data resource volatility, and the future time period volatility as the maximum value of backup data resource volatility.

[0062] S206: The internal demand buffer data volume is obtained by comprehensively processing the current internal factor parameter value, the lower limit of the basic buffer data limit, and the maximum value of the fluctuation of the backup data resources.

[0063] Comprehensive processing includes, but is not limited to, the process of transforming parameters of different dimensions and units into numerical values ​​with unified business meaning through specific mathematical models (such as linear weighting, vector summation, or nonlinear amplification).

[0064] For example, the current internal factor parameter value, the lower limit of the basic buffer data limit, and the maximum fluctuation value of the backup data resources can be multiplied to obtain the internal demand buffer data amount; In the embodiments of this specification, by using a pre-set internal factor mapping relationship and time label matching mechanism, the general prediction deviation factor is accurately anchored to the specific target time domain, realizing dynamic calibration of the differentiated performance of the prediction model under different time windows (such as business peak and off-peak periods). By extracting the fluctuation index of backup data resources and locking its maximum value, the inherent jitter noise of the underlying physical resources is included in the buffer calculation scope, establishing the boundary for dealing with extreme hardware fluctuations. The algorithm deviation component after time domain correction and the maximum value of physical fluctuations are comprehensively processed to ensure that under complex and ever-changing endogenous interference, the buffer resources reserved by the platform can accurately cover the prediction blind zone and environmental physical data oscillations, thereby improving the operational stability and risk resistance of the data transfer service.

[0065] Optional, please see Figure 3 , Figure 3 This is a flowchart illustrating a method for determining the adjustment amount of external environment buffer data as described in this specification. Specifically, the method for determining the adjustment amount of external environment buffer data based on the external environment disturbance parameters and the lower limit of the basic buffer data limit can be referenced as follows: S302: In the mapping relationship between business environment time stamp and popularity index, match the current data transfer environment popularity index according to the business environment time stamp corresponding to the target time period; Business environment time tags refer to the classification identifiers obtained after abstracting the target time period into business semantics. Unlike the time tags of internal business logic, business environment time tags focus on the time characteristics at the social or public network level, such as public holidays, weekday evening peak hours, and the end of a quarter or month.

[0066] The mapping relationship between business environment time tags and popularity index refers to the pre-built mapping relationship used to describe the relationship between different business environment time tags and popularity.

[0067] The current data transfer environment heat index is a scalar value used to quantify the current level of congestion in the external environment. If set to a normalized range (e.g., 0-100), a higher value indicates a more congested external environment.

[0068] As an illustration, the service platform calls the built-in calendar service component to perform business attribute parsing on the target time period, identifying whether it belongs to a statutory holiday, whether it is a working day, and the specific time period type (such as morning peak, evening peak, or early morning off-peak). The service platform obtains the business attribute parsing results, converts the physical timestamp into a business environment time tag with semantic information based on the business attribute parsing results, loads the business environment time tag and popularity index mapping relationship table pre-stored in the database. This mapping table is a set of key-value pairs built based on historical long-term network statistical data. Using the generated business environment time tag as the index key, the service platform performs an exact matching query in the mapping relationship table, extracts the preset value corresponding to the tag, and determines the preset value as the current data transfer environment popularity index.

[0069] S304: Determine the degree of deviation of the current data transfer environment heat index from the preset neutral balance threshold, and determine the data transfer environment heat factor based on the degree of deviation; The preset neutral balance threshold refers to the baseline score defined when the external environment is in a standard or healthy operating state. For example, in a scoring system of 0-100, 50 is set as the threshold. When the index is 50, it means that the network smoothness is as expected and no additional intervention is required; below 50 means the environment is idle, and above 50 means the environment is congested.

[0070] The degree of shift refers to the algebraic difference between the currently measured heat index and the aforementioned neutral threshold. This difference reflects the severity of environmental deterioration (positive shift) or improvement (negative shift).

[0071] The data transfer environment heat factor refers to the final adjustment coefficient calculated based on the degree of offset. It is a dimensionless value (e.g., 0.15, 0.2).

[0072] To illustrate, the system reads a preset neutral balance threshold (e.g., 50) defined in the system configuration file and subtracts the current data transfer environment heat index from this threshold, calculating the algebraic difference as the original offset degree. This original offset degree undergoes positive / negative verification and noise reduction. If the calculated original offset degree is less than or equal to zero, it means the current environment heat is below or at the baseline level, and the system determines that no additional environmental compensation is needed, directly setting the data transfer environment heat factor to zero. If the calculated offset degree is greater than zero, a preset sensitivity conversion coefficient (e.g., 0.01) is introduced, and the offset degree is multiplied by this conversion coefficient, thereby converting the difference in score form into a proportional data transfer environment heat factor. This ensures that resource reservations are increased only when the environment becomes unfavorable (i.e., overheated), and the basic quota is not actively reduced when the environment is favorable.

[0073] In one feasible implementation, specifically determining the degree of deviation of the current data transfer environment heat index from a preset neutral balance threshold, and determining the data transfer environment heat factor based on the degree of deviation, can be done in the following manner: Step B2: Calculate the difference between the current data transfer environment heat index and the preset neutral balance threshold; Step B4: If the difference is greater than the reference value, the ratio of the difference to the preset neutral balance threshold is determined as the data transfer environment heat factor; if the difference is less than or equal to the reference value, the reference value is determined as the data transfer environment heat factor.

[0074] S306: The external environment buffer data adjustment amount is obtained by comprehensively processing the data transfer environment heat factor, the external environment disturbance parameter and the basic buffer data limit.

[0075] Comprehensive processing includes, but is not limited to, the process of transforming parameters of different dimensions and units into numerical values ​​with unified business meaning through specific mathematical models (such as linear weighting, vector summation, or nonlinear amplification).

[0076] For example, the data transfer environment heat factor, the external environment disturbance parameter, and the lower limit of the basic buffer data quota can be multiplied to obtain the external environment buffer data adjustment amount.

[0077] This specification introduces a mapping logic between business environment time stamps and heat indexes, overcoming the limitations of relying solely on real-time physical link detection (such as packet loss rate and latency). It transforms the macro-level congestion trend of the external environment at specific time points (such as peak holiday periods and active trading windows) into a quantifiable environmental heat index, thereby enabling the prediction of potential risks. By calculating the degree of deviation of this heat index from the neutral equilibrium threshold, the heat factor is determined, ensuring that the intensity of resource adjustment is positively correlated with the degree of environmental deterioration. In addition, the heat factor, which represents the macro-level situation, and the external environmental disturbance parameters, which represent the micro-level link quality, are comprehensively processed. This allows the service platform to lock in sufficient compensatory buffer resources in advance, even in high-risk scenarios where the physical link is temporarily normal but the macro-level environment is extremely congested. This improves the survivability and service continuity of data transfer services in complex public environments.

[0078] Optional, please refer to Figure 4 , Figure 4 This is a flowchart illustrating a process based on configuration parameter adjustments. Specifically, after executing the data redundancy configuration for the data resource pool based on the dynamic resource configuration target value, the following methods can also be referenced: S402: Monitor the actual resource consumption of the data resource pool and the environmental operation characteristics under the current data transfer scenario, and calculate the prediction error deviation between the actual resource consumption and the target value of the dynamic resource configuration; Actual resource consumption refers to the peak or cumulative value of resources actually occupied by the data resource pool during the execution of the data transfer task. Specific indicators depend on the business type, such as actual peak bandwidth in network transmission scenarios, actual write IOPS in storage scenarios, and actual load output in power scenarios. Environmental operating characteristics refer to a real-time snapshot of the external environment within the time window of task execution. For example, recording the network packet loss rate, link jitter, server CPU temperature, and power grid frequency deviation at that time.

[0079] As an illustration, when the data transfer task is started, the service platform activates a fine-grained monitoring probe deployed in the underlying infrastructure. This probe continuously collects the actual resource consumption of the data resource pool at the millisecond or second time granularity, such as real-time throughput or memory usage, and simultaneously records the environmental operating characteristics under the current data transfer scenario. It maintains a sliding time window in memory, compares the collected actual resource consumption with the dynamic resource configuration target value in real time, and calculates the difference between the two through subtraction to obtain the instantaneous prediction error deviation.

[0080] S404: Construct a data reservation configuration error profile based on the predicted error deviation and the environmental operating characteristics; Data reservation configuration error profile information refers to a high-dimensional feature set or structured descriptive text that characterizes the execution status of a single data transfer task at the resource configuration level. This data reservation configuration error profile information not only includes the magnitude and direction of the error (too large or too small), but also encapsulates the external environmental state at the time the error occurred (such as network congestion and hardware health). It can be understood as a snapshot of the configuration deviation event.

[0081] In one implementation of attribution analysis for Large Language Model (LLM), semantic description logic is employed. Pre-defined prompt word engineering templates are invoked to transform numerical prediction error deviations and environmental operational characteristics into descriptions in natural human language. Specifically, result descriptions are generated based on the magnitude of the deviation, and background descriptions are generated based on environmental characteristics, such as extreme congestion in the external network or stable hardware operation. These descriptive fragments are filled into the corresponding slots of the templates and assembled into a complete narrative text (e.g., under stable external conditions, the task exhibited moderate resource over-reservation, with actual consumption far below the target value). This text containing causal logic clues is then identified as the data reservation configuration error profile information.

[0082] S406: Based on the data reserved configuration error profile information, the configuration parameter adjustment large language model is used to perform error attribution judgment processing to generate parameter evolution instructions for the data transfer internal demand prediction deviation factor and / or the data transfer external environment disturbance parameters. The parameter evolution instructions are used to perform parameter fine-tuning processing on the data transfer internal demand prediction deviation factor and / or the data transfer external environment disturbance parameters.

[0083] The configuration parameter adjustment of large language models refers to a special AI model based on a general large language model base (such as the Transformer architecture), which utilizes massive historical data transfer task logs, resource scheduling expert knowledge bases, and "error-cause" to perform supervised fine-tuning or reinforcement learning optimization on samples.

[0084] Error attribution determination refers to the logical reasoning process of model analysis of error sources. That is, it identifies whether the current configuration deviation is mainly caused by the subjective inaccuracy of the internal demand forecasting model (endogenous factors), or by the objective changes in the external physical environment (exogenous factors), or by the combined effect of both.

[0085] Parameter evolution instructions refer to the control code used to guide the updating of the system's core parameter library. They are usually expressed as structured JSON instructions, which include the object to be adjusted (internal factor / external parameter), the direction of adjustment (increase / decrease), and the adjustment step size.

[0086] In a schematic manner, a pre-trained configuration parameter adjustment language model is loaded. The constructed data-reserved configuration error profile information is input into the configuration parameter adjustment language model as part of the prompt words. The configuration parameter adjustment language model is controlled to perform error attribution judgment processing according to the preset thinking chain logic. First, the error direction in the profile is analyzed (such as resource idleness or insufficiency). Then, counterfactual reasoning is performed in combination with the environmental operation characteristics in the profile. For example, if the profile shows "a large amount of resources are idle" and "the environmental characteristics are extremely stable", the model infers that the external disturbance parameter has not been triggered and the root cause of the error is that the internal demand prediction deviation factor is set too high. If the profile shows "resources are in short supply" and "the network environment is suddenly congested", the model infers that the safety margin set by the external environmental disturbance parameter is insufficient to resist the actual risk. Based on the attribution conclusion of the above thinking chain logic, the language model generates a targeted parameter evolution instruction. The parameter evolution instruction clearly specifies the target parameter (internal factor or external parameter) to be adjusted and the specific fine-tuning value (e.g., "reduce the internal factor by 0.02"). The system then parses the instruction and updates the parameter database to complete the parameter fine-tuning processing.

[0087] The pre-defined thought chain logic requires the large language model to generate a series of intermediate reasoning steps before giving the final conclusion (i.e., parameter adjustment instructions). The system provides the following pre-defined logical reasoning templates through prompts: observation logic (e.g., parsing the input profile information to clarify the error direction (overflow / insufficiency) and environmental state (stable / congested)), counterfactual reasoning logic (e.g., assuming a certain parameter is normal, deducing the expected result, and comparing it with the actual result), attribution locking logic (e.g., eliminating interference items, locking the fundamental parameter (internal factor or external parameter) that causes the error to obtain the attribution result), and decision-making logic (e.g., determining the adjustment direction and step size based on the attribution result).

[0088] The configuration parameter adjustment of the large language model is based on the preset thought chain logic, and performs the following step-by-step reasoning: 1. Error nature definition and environmental background verification based on observation logic: The model first analyzes the prediction error deviation in the profile. If the deviation is negative, it is marked as a resource idle / wasteful state, meaning the configuration strategy is too conservative. If the deviation is positive, it is marked as a resource squeeze / insufficiency state, meaning the configuration strategy is too aggressive and risky. Simultaneously, the environmental operating characteristics in the reserved configuration error profile information are read to determine whether the external environment, as a "disturbing variable," participates in the formation of the error. 2. Based on counterfactual reasoning logic and attribution-locking logic, multidimensional attribution reasoning is performed to obtain attribution conclusions; 3. Based on the decision-making logic, attribution conclusions are used to generate instructions and evolve parameters, outputting structured parameter evolution instructions.

[0089] Optionally, the following illustrates a model training process for a large language model by adjusting configuration parameters: 1. Obtain a pre-trained base language model (LLM), create an initial configuration parameter adjustment plugin model for data transfer resource configuration scenarios, and combine the base LLM and the initial configuration parameter adjustment plugin model to form an initial configuration parameter adjustment LLM. The base LLM includes, but is not limited to, pre-trained models with general semantic and logical reasoning capabilities such as DeepSeek, GPT, and Llama series LLM. The initial configuration parameter adjustment plugin model can be created based on lightweight fine-tuning techniques (such as LoRA, Adapter) or machine learning models to carry professional knowledge in the resource scheduling domain.

[0090] 2. Obtain sample data for data transfer scenarios. This sample data includes "sample data reservation configuration error profile information" and corresponding "sample parameter evolution instructions" for data transfer scenarios. The sample data reservation configuration error profile information covers prediction error deviation, service load characteristics, and environmental operating characteristics such as network packet loss rate and latency in historical tasks. The sample parameter evolution instructions contain expert-level correction suggestions for internal demand prediction deviation factors and external environmental interference factors.

[0091] 3. Based on the attribution requirements in the data transfer scenario, the sample data is labeled. Specifically, each group of sample error profiles is labeled with the corresponding "dominant perturbation dimension" label (such as internal prediction dimension or environmental perturbation dimension) and "parameter correction gradient" label (including adjustment direction and step size value), forming the labeled samples required for supervised learning.

[0092] 4. Input the labeled sample data into the initial configuration parameters and adjust the large language model to perform at least one round of model training.

[0093] During the forward training of the model: The initial configuration parameters are adjusted. The large language model reserves configuration error profile information for the input sample data, performs multi-dimensional semantic feature decoupling, and performs logical alignment with the built-in domain knowledge to obtain the dominant perturbation dimension of the prediction and the parameter evolution instructions of the prediction.

[0094] During model back-training: Based on the predicted dominant perturbation dimension, the predicted parameter correction gradient, and the labeled parameters, a model loss function (such as cross-entropy loss function or mean squared error loss function) is used to calculate the model loss value (e.g., the model loss value consists of the dominant prediction loss and the instruction prediction loss; the dominant prediction loss is calculated using the model loss function based on the predicted dominant perturbation dimension and the dominant perturbation dimension label, while the instruction prediction loss is calculated using the model loss function based on the predicted parameter correction gradient and the parameter correction gradient label). Based on this model loss value, gradient backpropagation and parameter adjustment are performed on the initial configuration parameter-adjusted large language model (focusing on the plug-in model parameters) until the model training termination condition is met, resulting in a trained configuration parameter-adjusted large language model.

[0095] The training termination conditions for the model include, but are not limited to: the value of the loss function being less than or equal to a preset loss function threshold, the model's attribution accuracy on the validation set reaching a preset percentage, or the number of iterations reaching a preset threshold. Specific termination conditions can be determined based on the actual business requirements for resource reservation accuracy.

[0096] In this specification, S402 to S406 construct a resource allocation self-evolution closed-loop mechanism driven by a large language model. By monitoring the actual consumption of the resource pool and the environmental operation characteristics in real time, the abstract prediction error is transformed into error profile information containing causal clues. By utilizing the logical reasoning and feature decoupling capabilities of the large language model, intelligent attribution of error sources is achieved, thereby generating gradient correction instructions that can accurately drive parameter evolution. Through continuous parameter fine-tuning, resource idle waste is reduced and safety redundancy blind spots are eliminated.

[0097] In one feasible implementation, the specific execution of the error attribution determination process based on the data reservation configuration error profile information using a large language model with configuration parameter adjustment to generate parameter evolution instructions for the data transfer internal demand prediction deviation factor and / or the data transfer external environment disturbance parameters can refer to the following method: Step C2: Use the configuration parameters to adjust the large language model to perform multi-dimensional semantic feature decoupling on the data reserved configuration error profile information, so as to separate the internal prediction failure component and the environmental disturbance variation component; Multidimensional semantic feature decoupling refers to mapping a descriptive text or feature vector containing complex causal relationships to a high-dimensional semantic space, and then using a predefined semantic orthogonal basis to decompose the mixed features into independent feature vectors that are unrelated to each other. In this embodiment, it can be understood as the process of separating deviations at the business logic level from noise at the physical environment level.

[0098] The internal prediction failure component refers to the weight or value of the error source specifically attributed to the inaccuracy of the business load prediction model.

[0099] The environmental disturbance variation component refers to the weight or value of the error source specifically attributed to the deviation of the external physical environment from the expected state and the resulting resource loss.

[0100] In a schematic manner, the data reservation configuration error profile information is input into the configuration parameter adjustment language model. The configuration parameter adjustment language model internally activates a multi-head attention mechanism, assigning weights to keywords in the data reservation configuration error profile information. The first attention head focuses on the platform's business semantic features, extracting the causal correlation strength between these business semantic features and the total error, quantifying and normalizing it into an internal prediction failure component (e.g., a score of 0.7 indicates 70% responsibility lies with internal prediction). Simultaneously, the second attention head focuses on the semantic features related to the external environment, extracting and quantifying their correlation strength with the error into an environmental disturbance variation component (e.g., a score of 0.3).

[0101] Step C4: Adjust the large language model using the configuration parameters to perform logical alignment processing on the internal prediction failure component and the environmental disturbance variation component with the historical parameter configuration knowledge graph, so as to determine the dominant disturbance dimension that causes the prediction error deviation. A historical parameter configuration knowledge graph is a graph-based data structure that is maintained. Its nodes represent historical environmental states, internal load characteristics, error types, and parameter adjustment actions; edges represent causal probabilities or evolutionary paths between these elements. This graph stores all the platform's past trial-and-error experiences and expert knowledge. Association logic alignment refers to the process by which the configuration parameters of a large language model are adjusted to map the current real-time feature components to the knowledge graph, and to find the most similar subgraph or the optimal path. The dominant perturbation dimension refers to one or more dimension categories that have been verified by historical experience to play a decisive role in the current prediction error. In a schematic manner, the large language model for controlling configuration parameter adjustment transforms internal predicted failure components and environmental disturbance variation components into internal query vectors and environmental query vectors. Based on the internal query vectors and environmental query vectors, graph vector retrieval and similarity calculation are performed in the historical parameter configuration knowledge graph. Each historical node in the historical parameter configuration knowledge graph has been pre-converted into a fixed historical feature key value vector through the same encoder and is divided into internal cause clusters and external environment clusters according to its attributes.

[0102] During logical alignment, the following parallel operations are performed: 1) Internal alignment calculation: Using the internal query vector as a probe, an approximate nearest neighbor search is performed in the "internal cause cluster" of the graph to find the top k most matching historical nodes, and the cosine similarity between the query vector and these historical node vectors is calculated. The maximum similarity value is taken and denoted as the internal alignment confidence score_Int. 2) Context alignment calculation: Using the context query vector as a probe, a search is performed in the external context cluster of the graph. Similarly, the maximum cosine similarity value is calculated and extracted and denoted as the context alignment confidence score_Ext.

[0103] The control configuration parameters are adjusted to compare the two confidence levels mentioned above to determine the dominant perturbation dimension: if Score_Int is greater than Score_Ext and the difference exceeds a preset threshold (e.g., 0.1), this geometrically means that the current error feature vector is closer to the cluster center of the "internal algorithm failure" in history, and the angle is smaller. Then it is determined that the internal prediction failure component is highly aligned with the historical experience logic, and the dominant perturbation dimension is determined to be the internal prediction dimension.

[0104] If Score_Ext is greater than Score_Int and meets the threshold, it means that the current error characteristics are more in line with the historical pattern of "external environment deterioration". Therefore, the environmental disturbance variation component is determined to be the main contradiction, and the dominant disturbance dimension is determined to be the environmental disturbance dimension.

[0105] Step C6: If the dominant disturbance dimension is the internal prediction dimension, then adjust the internal parameter correction gradient value of the large language model based on the internal prediction failure component to solve the internal parameter correction gradient value for the data transfer internal demand prediction deviation factor through the configuration parameters, and generate parameter evolution instructions based on the internal parameter correction gradient value. The intrinsic parameter correction gradient value refers to the numerical increment vector used to update the intrinsic predictors. It includes: the sign attribute, which determines the direction of adjustment (positive increase to improve reservation, negative decrease to reduce waste); and the magnitude attribute, which determines the magnitude or step size of the adjustment.

[0106] Once the configuration parameter adjustment language model confirms that the dominant perturbation dimension is the internal prediction dimension, it infers the internal prediction failure component with the pre-trained domain knowledge base to estimate the partial derivative of the current data transfer internal demand prediction deviation factor with respect to the total resource allocation target. This involves analyzing "the change in the final buffer limit response when the deviation factor changes by one unit step." Then, a simulated "error back-mapping" operation is performed: to offset the deviation represented by the internal prediction failure component, the configuration parameter adjustment language model is used to deduce the compensatory change that the internal factor should undergo, defining it as the internal parameter correction gradient value. For example, if the actual load exceeds expectations, the model calculates a positive gradient value (e.g., +0.05); if resources are largely idle, a negative gradient value (e.g., -0.03) is calculated. Finally, this gradient value is encapsulated into a structured parameter evolution instruction and sent to the system's configuration center to achieve automated parameter fine-tuning of the internal demand prediction deviation factor.

[0107] Specifically, the configuration parameter adjustment language model associates the internal parameter correction gradient value with the globally unique identifier of the target parameter to clarify that the controlled object corresponding to the gradient value is the data transfer internal demand prediction deviation factor. According to the preset system security policy, the internal parameter correction gradient value is processed for compliance mapping, that is, the new parameter value is determined by querying the parameter configuration threshold table, and it is judged whether the new parameter value after adding the internal parameter correction gradient value is within the preset safe extreme value range (for example, the internal factor must not exceed 2.0 and must not be lower than 1.0). If the parameter evolution trend caused by the internal parameter correction gradient value touches the extreme value boundary, the system will automatically perform a truncation and smoothing operation to correct the gradient value to the critical increment that makes the parameter reach the boundary value, thereby ensuring the security of the instruction.

[0108] After completing the security verification, the semantic and instruction protocol conversion stage begins. The verified new parameter values, adjustment direction (the sign of the new parameter value determines the adjustment direction), and effective timestamp are encapsulated into a structured instruction message header. For example, the model generates a memory parameter hot update instruction containing operation type field, gradient value field, and execution strategy field by filling a preset instruction template.

[0109] Step C8: If the dominant perturbation dimension is the environmental perturbation dimension, then adjust the environmental compensation gradient value of the large language model based on the environmental perturbation variation component to solve for the external data interaction environmental interference factor through the configuration parameters, and generate parameter evolution instructions based on the environmental compensation gradient value.

[0110] Once the configuration parameters are adjusted and the large language model confirms that the dominant disturbance dimension is the environmental disturbance dimension, deep causal reasoning is performed based on the environmental disturbance variation components and the pre-trained domain knowledge base. This aims to simulate and calculate the transfer function between the external data interaction environment interference factor and the environmental dynamic loss, thereby estimating the partial derivative of the current interference factor with respect to the total resource configuration target. In other words, it quantitatively analyzes the change in the system's response redundancy to external environmental physical noise (such as network jitter, link packet loss, etc.) when the environmental interference factor changes by one unit step.

[0111] Subsequently, a simulated environmental noise reverse compensation operation is performed. Based on the principle of eliminating bias, the large language model, adjusted by configuration parameters, derives the compensatory shift that external interference factors should undergo to completely offset the resource gaps or surpluses caused by the environmental disturbance variation components, and defines it as the environmental compensation gradient value. For example, if a sharp increase in retransmission overhead is detected due to external network congestion, the model will calculate a positive environmental compensation gradient value (e.g., +0.08) to enhance defense; if the external environment is in an extremely stable low-noise state, resulting in excessive reserves, a negative gradient value (e.g., -0.04) will be calculated to release redundancy.

[0112] During the instruction generation phase, the model forcibly associates the environmental compensation gradient value with the globally unique identifier of the external environmental disturbance factor to ensure that the adjustment object is accurately locked as the external disturbance resistance parameter, and performs compliance mapping processing on the gradient value according to the preset environmental security strategy. At this time, the system determines the new value of the parameter to be adjusted by querying the environmental parameter threshold table, and closely monitors whether the new value is within the preset safe extreme value range to prevent the system from falling into resource deadlock due to over-compensation or defense collapse due to insufficient compensation; once the gradient value-driven evolution trend reaches the safe extreme value, the system will automatically trigger the truncation smoothing algorithm to forcibly correct the gradient value to the critical increment that returns the parameter to the boundary value.

[0113] After completing the aforementioned security checks, the semantic and instruction protocol conversion phase begins. The validated new values ​​of environmental parameters, the adjustment direction based on symbolic determination, and the timestamp of the current task are co-encapsulated into a structured instruction header. Finally, the large language model automatically generates a memory parameter hot-update instruction by filling in a dedicated instruction template optimized for the external environment. This instruction includes fields for the environment operation type, the environment compensation gradient value, and the execution strategy based on real-time requirements. This instruction is then immediately sent to the system's configuration center, thereby achieving adaptive closed-loop fine-tuning against external data interaction environment interference factors.

[0114] In this specification, the above steps are performed by decoupling the complex mixed error profile through multidimensional semantic features, separating the different contributions of internal prediction inaccuracies and external environmental disturbances to the error, and combining them with historical parameter configuration knowledge graphs for logical alignment. This ensures that parameter correction gradient values ​​with clear directions and step sizes can be accurately calculated for the dominant disturbance dimension, thereby achieving targeted fine-tuning and closed-loop evolution of internal demand prediction deviation factors and external environmental disturbance factors, and improving the response accuracy and robustness of the resource allocation system under complex operating conditions.

[0115] The following will combine Figure 5 This specification provides a detailed description of the data configuration device provided in the embodiments. It should be noted that... Figure 5 The data configuration device shown is used to execute this specification. Figures 1-4 The methods shown in the embodiments are illustrated for ease of explanation, showing only the parts related to the embodiments of this specification. For specific technical details not disclosed, please refer to this specification. Figures 1-4 The example shown.

[0116] Please see Figure 5 This diagram illustrates the structure of a data configuration device according to an embodiment of this specification. The data configuration device 1 can be implemented as all or part of a device through software, hardware, or a combination of both. According to some embodiments, the data configuration device 1 includes a data processing module 11, a configuration inference module 12, and a reserved configuration module 13, specifically used for: Data processing module 11 is used to obtain the business data load scale to be transferred within the target time period, the internal demand prediction deviation factor of data transfer, and the external environmental disturbance parameters of data transfer. Based on the preset security threshold and the business data load scale, it determines the lower limit of the basic buffer data quota and the upper limit of the data business target buffer for the operation of the data transfer service. The inference module 12 is configured to determine the amount of internal demand buffer data based on the data transfer internal demand prediction deviation factor and the lower limit of the basic buffer data limit, determine the amount of external environment buffer data adjustment based on the external environment disturbance parameters and the lower limit of the basic buffer data limit, and comprehensively process the amount of internal demand buffer data and the amount of external environment buffer data adjustment to obtain the amount of buffer reserve data. The configuration inference module 12 is used to construct the theoretical data operation level of the data transfer service based on the lower limit of the basic buffer data quota and the amount of buffer backup data, use the upper limit of the data service target buffer to perform data level limiting processing on the theoretical data operation level to obtain the reference data reservation result, and use the lower limit of the basic buffer data quota to perform bottom-line configuration processing on the reference data reservation result to obtain the dynamic resource configuration target value. The reserved configuration module 13 is used to configure data redundancy for data transfer services of the data resource pool according to the dynamic resource configuration target value.

[0117] Optionally, determining the amount of internal demand buffer data based on the data transfer internal demand forecasting deviation factor and the lower limit of the basic buffer data quota includes: In the preset internal factor mapping relationship corresponding to the data transfer internal demand prediction deviation factor, the target time domain compensation coefficient is matched according to the first internal mapping time label corresponding to the target time period, and the target time domain compensation coefficient is determined as the current internal factor parameter value of the data transfer internal demand prediction deviation factor. Extract the backup data resource fluctuation index corresponding to the target time period, and select the maximum value of the backup data resource fluctuation among the backup data resource fluctuation index; The internal demand buffer data volume is obtained by comprehensively processing the current internal factor parameter value, the lower limit of the basic buffer data limit, and the maximum fluctuation value of the backup data resources.

[0118] Optionally, the step of extracting the backup data resource fluctuation index corresponding to the target time period and selecting the maximum value of the backup data resource fluctuation among the backup data resource fluctuation index includes: Extract the volatility of the first historical backup data resource within the preset historical quarterly cycle of the target time period; Extract the volatility of the second historical backup data resource corresponding to the same period of the previous year for the target time period; Determine the cumulative value of the predicted load fluctuation for the target time period within the current cycle for the business data load scale, and obtain the volatility rate for the future time period by weighting the business data load scale based on the cumulative value of the predicted load fluctuation; The maximum value among the first historical backup data resource volatility, the second historical backup data resource volatility, and the future time period volatility is selected as the maximum value of backup data resource volatility.

[0119] Optionally, determining the external environment buffer data adjustment amount based on the external environment disturbance parameters and the lower limit of the basic buffer data quota includes: In the mapping relationship between business environment time tags and popularity index, the current data transfer environment popularity index is matched according to the business environment time tag corresponding to the target time period. Determine the degree of deviation of the current data transfer environment heat index from a preset neutral balance threshold, and determine the data transfer environment heat factor based on the degree of deviation; The external environment buffer data adjustment amount is obtained by comprehensively processing the data transfer environment heat factor, the external environment disturbance parameter, and the basic buffer data limit.

[0120] Optionally, determining the degree of deviation of the current data transfer environment heat index from a preset neutral balance threshold, and determining the data transfer environment heat factor based on the degree of deviation, includes: Calculate the difference between the current data transfer environment heat index and the preset neutral balance threshold; If the difference is greater than the reference value, the ratio of the difference to the preset neutral balance threshold is determined as the data transfer environment heat factor; if the difference is less than or equal to the reference value, the reference value is determined as the data transfer environment heat factor.

[0121] Optionally, the step of using the upper limit of the data service target buffer to perform data level limiting processing on the theoretical data operating level to obtain a reference data reservation result, and using the lower limit of the basic buffer data quota to perform a safety net configuration processing on the reference data reservation result to obtain a dynamic resource configuration target value, includes: The theoretical data operating level is compared with the upper limit of the data service target buffer. If the theoretical data operating level is greater than the upper limit of the data service target buffer, the reference data reservation result is determined as the upper limit of the data service target buffer. If the theoretical data operating level is less than or equal to the upper limit of the data service target buffer, the reference data reservation result is determined as the theoretical data operating level. The reference data reservation result is compared with the lower limit of the basic buffer data limit, and the maximum value between the reference data reservation result and the lower limit of the basic buffer data limit is selected as the target value for dynamic resource allocation.

[0122] Optionally, after configuring data redundancy for data transfer services on the data resource pool according to the dynamic resource configuration target value, the method further includes: Monitor the actual resource consumption of the data resource pool and the environmental operation characteristics under the current data transfer scenario, and calculate the prediction error deviation between the actual resource consumption and the target value of the dynamic resource configuration; Based on the predicted error deviation and the environmental operating characteristics, a data reservation configuration error profile is constructed. Based on the data reservation configuration error profile information, a configuration parameter adjustment language model is used to perform error attribution judgment processing to generate parameter evolution instructions for the data transfer internal demand prediction deviation factor and / or the data transfer external environment disturbance parameters. The parameter evolution instructions are used to perform parameter fine-tuning processing on the data transfer internal demand prediction deviation factor and / or the data transfer external environment disturbance parameters.

[0123] Optionally, the error attribution determination processing based on the data reservation configuration error profile information using a configuration parameter adjustment large language model is performed to generate parameter evolution instructions for the data transfer internal demand prediction deviation factor and / or the data transfer external environment disturbance parameters, including: The configuration parameters are used to adjust the large language model to perform multi-dimensional semantic feature decoupling on the data reserved configuration error profile information, so as to separate the internal prediction failure component and the environmental disturbance variation component. By adjusting the configuration parameters of the large language model, the internal prediction failure component and the environmental disturbance variation component are logically aligned with the historical parameter configuration knowledge graph to determine the dominant disturbance dimension that causes the prediction error deviation. If the dominant disturbance dimension is an internal prediction dimension, then the large language model is adjusted by the configuration parameters to calculate the internal parameter correction gradient value of the prediction deviation factor of the data transfer internal demand based on the internal prediction failure component, and a parameter evolution instruction is generated based on the internal parameter correction gradient value. If the dominant perturbation dimension is the environmental perturbation dimension, then the configuration parameters are used to adjust the environmental compensation gradient value of the large language model based on the environmental perturbation variation component for the external data interaction environmental interference factor, and the parameter evolution instruction is generated based on the environmental compensation gradient value.

[0124] It should be noted that the data configuration device provided in the above embodiments is only illustrated by the division of the above functional modules when executing the data configuration method. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the data configuration device and the data configuration method embodiments provided in the above embodiments belong to the same concept, and the implementation process can be found in the method embodiments, which will not be repeated here.

[0125] The example numbers in this specification are for descriptive purposes only and do not represent the superiority or inferiority of the examples.

[0126] This specification also provides a computer storage medium that can store multiple instructions adapted to be loaded and executed by a processor as described above. Figures 1-4 The data configuration method described in the illustrated embodiment can be found in the following documentation for a detailed execution process: Figures 1-4 The specific details of the illustrated embodiments will not be elaborated here.

[0127] This specification also provides a computer program product that stores at least one instruction, said at least one instruction being loaded and executed by the processor as described above. Figures 1-4 The data configuration method described in the illustrated embodiment can be found in the following documentation for a detailed execution process: Figures 1-4 The specific details of the illustrated embodiments will not be elaborated here.

[0128] Please refer to Figure 6 This is a structural block diagram of an electronic device provided in an embodiment of this specification. The electronic device in this specification may include one or more of the following components: a processor 1010, a memory 1020, an input device 1030, an output device 1040, and a bus 1050. The processor 1010, memory 1020, input device 1030, and output device 1040 may be connected to each other via the bus 1050.

[0129] Processor 1010 may include one or more processing cores. Processor 1010 connects to various parts of the electronic device using various interfaces and lines, and performs various functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 1020, and by calling data stored in memory 1020. Optionally, processor 1010 may be implemented using at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). Processor 1010 may integrate one or a combination of central processing unit (CPU), graphics processing unit (GPU), and modem. The CPU mainly handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the displayed content; and the modem is used for wireless communication. It is understood that the modem may also not be integrated into processor 1010 and may be implemented separately through a communication chip.

[0130] The memory 1020 may include random access memory (RAM) or read-only memory (ROM). Optionally, the memory 1020 may include non-transitory computer-readable storage medium. The memory 1020 may be used to store instructions, programs, code, code sets, or instruction sets.

[0131] The input device 1030 is used to receive input instructions or data, and includes, but is not limited to, a keyboard, mouse, camera, microphone, or touch device. The output device 1040 is used to output instructions or data, and includes, but is not limited to, a display device and a speaker. In this embodiment, the input device 1030 can be a temperature sensor for acquiring the operating temperature of the electronic device. The output device 1040 can be a speaker for outputting audio signals.

[0132] In addition, those skilled in the art will understand that the structure of the electronic device shown in the above figures does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown, or combine certain components, or have different component arrangements. For example, the electronic device may also include radio frequency circuits, input units, sensors, audio circuits, wireless fidelity (WIFI) modules, power supplies, Bluetooth modules, etc., which will not be described in detail here.

[0133] In the embodiments of this specification, the executing entity for each step can be the electronic device described above. Optionally, the executing entity for each step can be the operating system of the electronic device. The operating system can be Android, iOS, or other operating systems; this specification does not limit this.

[0134] exist Figure 6 In the electronic device, the processor 1010 can be used to call a program stored in the memory 1020 and execute it to implement the data configuration method as described in the various method embodiments of this specification.

[0135] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory, or random access memory, etc.

[0136] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.), and signals involved in the embodiments of this specification are all authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. For example, the business data involved in this specification were all obtained under full authorization.

[0137] The above-disclosed embodiments are merely preferred embodiments of this specification and should not be construed as limiting the scope of this specification. Therefore, any equivalent variations made in accordance with the claims of this specification shall still fall within the scope of this specification.

Claims

1. A data configuration method, characterized in that, Applied to a service platform, the method includes: The system obtains the business data load scale to be transferred within the target time period, the internal demand prediction deviation factor for data transfer, and the external environmental disturbance parameters for data transfer. Based on the preset security threshold and the business data load scale, it determines the lower limit of the basic buffer data quota and the upper limit of the target buffer for data transfer service operation. The amount of internal demand buffer data is determined based on the data transfer internal demand prediction deviation factor and the lower limit of the basic buffer data quota. The amount of external environment buffer data adjustment is determined based on the external environment disturbance parameters and the lower limit of the basic buffer data quota. The amount of internal demand buffer data and the amount of external environment buffer data adjustment are combined to obtain the amount of buffer reserve data. Based on the lower limit of the basic buffer data quota and the amount of buffer backup data, a theoretical data operation level for data transfer service is constructed. The upper limit of the data service target buffer is used to perform data level limiting processing on the theoretical data operation level to obtain a reference data reservation result. The lower limit of the basic buffer data quota is used to perform a bottom-line configuration processing on the reference data reservation result to obtain a dynamic resource configuration target value. Data redundancy configuration is performed on the data resource pool for data transfer services based on the dynamic resource configuration target value; The determination of the internal demand buffer data amount based on the data transfer internal demand forecasting deviation factor and the lower limit of the basic buffer data quota includes: In the preset internal factor mapping relationship corresponding to the data transfer internal demand prediction deviation factor, the target time domain compensation coefficient is matched according to the first internal mapping time label corresponding to the target time period, and the target time domain compensation coefficient is determined as the current internal factor parameter value of the data transfer internal demand prediction deviation factor; the spare data resource fluctuation index corresponding to the target time period is extracted, and the maximum value of spare data resource fluctuation in the spare data resource fluctuation index is selected; the current internal factor parameter value, the lower limit of the basic buffer data quota, and the maximum value of spare data resource fluctuation are comprehensively processed to obtain the internal demand buffer data volume; or, The amount of internal demand buffer data is obtained by multiplying the lower limit of the basic buffer data amount by the data transfer internal demand forecast deviation factor.

2. The method according to claim 1, characterized in that, The step of extracting the backup data resource fluctuation index corresponding to the target time period and selecting the maximum value of the backup data resource fluctuation among the backup data resource fluctuation index includes: Extract the volatility of the first historical backup data resource within the preset historical quarterly cycle of the target time period; Extract the volatility of the second historical backup data resource corresponding to the same period of the previous year for the target time period; Determine the cumulative value of the predicted load fluctuation for the target time period within the current cycle for the business data load scale, and obtain the volatility rate for the future time period by weighting the business data load scale based on the cumulative value of the predicted load fluctuation; The maximum value among the first historical backup data resource volatility, the second historical backup data resource volatility, and the future time period volatility is selected as the maximum value of backup data resource volatility.

3. The method according to claim 1, characterized in that, The step of determining the external environment buffer data adjustment amount based on the external environment disturbance parameters and the lower limit of the basic buffer data quota includes: In the mapping relationship between business environment time tags and popularity index, the current data transfer environment popularity index is matched according to the business environment time tag corresponding to the target time period. Determine the degree of deviation of the current data transfer environment heat index from a preset neutral balance threshold, and determine the data transfer environment heat factor based on the degree of deviation; The external environment buffer data adjustment amount is obtained by comprehensively processing the data transfer environment heat factor, the external environment disturbance parameter, and the basic buffer data limit.

4. The method according to claim 3, characterized in that, The step of determining the degree of deviation of the current data transfer environment heat index from a preset neutral balance threshold, and determining the data transfer environment heat factor based on the degree of deviation, includes: Calculate the difference between the current data transfer environment heat index and the preset neutral balance threshold; If the difference is greater than the reference value, the ratio of the difference to the preset neutral balance threshold is determined as the data transfer environment heat factor; if the difference is less than or equal to the reference value, the reference value is determined as the data transfer environment heat factor.

5. The method according to claim 1, characterized in that, The process of using the upper limit of the data service target buffer to perform data level limiting processing on the theoretical data operating level to obtain a reference data reservation result, and using the lower limit of the basic buffer data quota to perform a safety net configuration processing on the reference data reservation result to obtain a dynamic resource configuration target value, includes: The theoretical data operating level is compared with the upper limit of the data service target buffer. If the theoretical data operating level is greater than the upper limit of the data service target buffer, the reference data reservation result is determined as the upper limit of the data service target buffer. If the theoretical data operating level is less than or equal to the upper limit of the data service target buffer, the reference data reservation result is determined as the theoretical data operating level. The reference data reservation result is compared with the lower limit of the basic buffer data limit, and the maximum value between the reference data reservation result and the lower limit of the basic buffer data limit is selected as the target value for dynamic resource allocation.

6. The method according to claim 1, characterized in that, After configuring data redundancy for data transfer services in the data resource pool according to the dynamic resource configuration target value, the method further includes: Monitor the actual resource consumption of the data resource pool and the environmental operation characteristics under the current data transfer scenario, and calculate the prediction error deviation between the actual resource consumption and the target value of the dynamic resource configuration; Based on the predicted error deviation and the environmental operating characteristics, a data reservation configuration error profile is constructed. Based on the data reservation configuration error profile information, a configuration parameter adjustment language model is used to perform error attribution judgment processing to generate parameter evolution instructions for the data transfer internal demand prediction deviation factor and / or the data transfer external environment disturbance parameters. The parameter evolution instructions are used to perform parameter fine-tuning processing on the data transfer internal demand prediction deviation factor and / or the data transfer external environment disturbance parameters.

7. The method according to claim 6, characterized in that, The error attribution determination process based on the data reservation configuration error profile information uses a configuration parameter adjustment large language model to generate parameter evolution instructions for the data transfer internal demand prediction deviation factor and / or the data transfer external environment disturbance parameters, including: The configuration parameters are used to adjust the large language model to perform multi-dimensional semantic feature decoupling on the data reserved configuration error profile information, so as to separate the internal prediction failure component and the environmental disturbance variation component. By adjusting the configuration parameters of the large language model, the internal prediction failure component and the environmental disturbance variation component are logically aligned with the historical parameter configuration knowledge graph to determine the dominant disturbance dimension that causes the prediction error deviation. If the dominant disturbance dimension is an internal prediction dimension, then the large language model is adjusted by the configuration parameters to calculate the internal parameter correction gradient value of the prediction deviation factor of the data transfer internal demand based on the internal prediction failure component, and a parameter evolution instruction is generated based on the internal parameter correction gradient value. If the dominant perturbation dimension is the environmental perturbation dimension, then the configuration parameters are used to adjust the environmental compensation gradient value of the large language model based on the environmental perturbation variation component for the external data interaction environmental interference factor, and the parameter evolution instruction is generated based on the environmental compensation gradient value.

8. A data reservation configuration device, characterized in that, The device, applied to a service platform, includes: The data processing module is used to obtain the business data load scale to be transferred within the target time period, the internal demand prediction deviation factor of data transfer, and the external environmental disturbance parameters of data transfer. Based on the preset security threshold and the business data load scale, it determines the lower limit of the basic buffer data quota and the upper limit of the data business target buffer for the operation of the data transfer service. The inference module is configured to determine the amount of internal demand buffer data based on the data transfer internal demand prediction deviation factor and the lower limit of the basic buffer data limit, determine the amount of external environment buffer data adjustment based on the external environment disturbance parameters and the lower limit of the basic buffer data limit, and comprehensively process the amount of internal demand buffer data and the amount of external environment buffer data adjustment to obtain the amount of buffer reserve data. The configuration inference module is used to construct the theoretical data operation level of the data transfer service based on the lower limit of the basic buffer data quota and the amount of buffer backup data, perform data level limiting processing on the theoretical data operation level using the upper limit of the data service target buffer to obtain the reference data reservation result, and use the lower limit of the basic buffer data quota to perform bottom-line configuration processing on the reference data reservation result to obtain the dynamic resource configuration target value. The reserved configuration module is used to configure data redundancy for data transfer services of the data resource pool according to the dynamic resource configuration target value; The determination of the internal demand buffer data amount based on the data transfer internal demand forecasting deviation factor and the lower limit of the basic buffer data quota includes: In the preset internal factor mapping relationship corresponding to the data transfer internal demand prediction deviation factor, the target time domain compensation coefficient is matched according to the first internal mapping time label corresponding to the target time period, and the target time domain compensation coefficient is determined as the current internal factor parameter value of the data transfer internal demand prediction deviation factor; the spare data resource fluctuation index corresponding to the target time period is extracted, and the maximum value of spare data resource fluctuation in the spare data resource fluctuation index is selected; the current internal factor parameter value, the lower limit of the basic buffer data quota, and the maximum value of spare data resource fluctuation are comprehensively processed to obtain the internal demand buffer data volume; or, The amount of internal demand buffer data is obtained by multiplying the lower limit of the basic buffer data amount by the data transfer internal demand forecast deviation factor.

9. A computer storage medium, characterized in that, The computer storage medium stores a plurality of instructions adapted for loading by a processor and executing the steps of the method as described in any one of claims 1 to 7.

10. A computer program product, characterized in that, The computer program product stores at least one instruction, which is loaded by a processor and executed according to any one of claims 1 to 7.

11. An electronic device, characterized in that, include: A processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to execute the steps of the method as described in any one of claims 1 to 7.

Images