Enterprise server data optimization platform applied to multiple scenarios

Abstract

The application discloses an enterprise server data optimization platform applied to multiple scenes, comprising the following steps: monitoring bidirectional interaction data in a data interaction period, and completing data label setting through multi-dimensional classification; calculating single-time node server short-time pressure parameters, predicting long-time pressure trend based on an LSTM neural network, then fusing long-time and short-time pressure features, and realizing data priority grading through deviation degree calculation of a standard pressure value; finally, executing a differentiated storage strategy through a multi-level cache architecture in the next period, synchronizing high-priority data to the cloud, and only locally caching low-priority data. The application combines long-time and short-time two-dimensional pressure evaluation and a graded cache mechanism, effectively improves server data access efficiency, reduces bandwidth consumption, and adapts to the data storage optimization needs of multiple scenes such as e-commerce, finance and manufacturing.

Description

Technical Field

[0001] This invention relates to the field of server data optimization, and more specifically, to an enterprise server data optimization platform applicable to multiple scenarios. Background Technology

[0002] With the rapid development of the digital economy, the scale of data carried by enterprise-level servers is growing exponentially. Especially in the e-commerce industry, enterprise servers need to process multi-dimensional data such as core transaction orders, user behavior interactions, system operation and maintenance monitoring, and network security protection at the same time. The data interaction frequency is high, the data types are complex, and the peak access fluctuations are large, which puts extremely high demands on server performance and data storage efficiency.

[0003] In existing technologies, enterprise servers and cloud data storage mostly adopt a unified upload and download synchronization model. All interactive data is synchronized to the cloud and cached locally without differentiation, lacking a differentiated processing mechanism based on data characteristics. On the one hand, high-frequency access and high-pressure fluctuating data are directly uploaded to the cloud, resulting in high network bandwidth utilization and increased data access latency. In high-concurrency scenarios, this can easily lead to server response timeouts and order data processing bottlenecks. On the other hand, low-frequency and low-pressure data occupy local cache resources for extended periods, causing a decrease in cache hit rate, wasted storage space, and difficulty in optimizing storage resource allocation.

[0004] Therefore, there is an urgent need to develop an enterprise server data optimization platform that can combine long-term and short-term pressure assessment, realize intelligent data classification and multi-level caching collaboration, and adapt to multiple application scenarios, in order to solve the technical problems of low storage efficiency, insufficient pressure prediction and poor scenario adaptability in existing technologies. Summary of the Invention

[0005] This invention overcomes the shortcomings of existing technologies and proposes an enterprise server data optimization platform applicable to multiple scenarios.

[0006] The first aspect of this invention provides a method for optimizing enterprise server data in multiple scenarios, comprising: S1. During a data interaction cycle on the enterprise server, monitor the data interaction between the user data terminal and the cloud data terminal, and classify and label the real-time interaction data to obtain multiple sets of interaction data. S2. During the interactive monitoring process, analyze the network pressure monitoring pressure and data interaction pressure monitoring of the user data terminal for each set of interactive data in a single time node to obtain short-term pressure parameters. Based on LSTM prediction, serialize and predict long-term features of the short-term pressure parameters corresponding to multiple time nodes to obtain long-term pressure parameters. S3. Construct a real-time pressure feature vector by using short-term and long-term pressure parameters, introduce standard pressure feature values ​​and real-time pressure features for deviation comparison, evaluate the overall pressure of real-time interactive data on enterprise servers, and classify the real-time interactive data. S4. Based on the hierarchical classification, map the data to the data tags. In the next data interaction cycle, set up a multi-level caching mechanism to classify and cache the real-time interactive data, and upload the highest-level data to the cloud data terminal.

[0007] In this solution, the real-time interactive data in S1 includes core transaction and order full-link data, real-time user behavior and interaction data, server and application performance operation and maintenance data, and real-time network traffic and security data.

[0008] In this solution, S1 further includes: The dimensions for classifying real-time interactive data include user geographic location, product category preference, access time period, and data volume per access. During a data interaction cycle, multiple sets of interaction data are obtained by classifying real-time interaction data from different users. The same set of interaction data has similar access frequency, data volume, and data pressure characteristics.

[0009] In this solution, S2 specifically refers to: Short-term stress parameters include short-term network stress values ​​and short-term server stress values; The short-term network pressure value is obtained by weighting three indicators after normalization: the average response time of a set of interactive data at a single time node, the request success rate, and the data traffic consumption. The short-term server pressure value is obtained by weighted calculation of three indicators of enterprise server: CPU utilization, memory utilization, and disk I / O utilization, after normalization. Using a set of interactive data as the target data for pressure monitoring, we analyze the corresponding short-term pressure parameters at N time points. We then arrange the short-term pressure parameters of N consecutive time points in chronological order to construct a multi-dimensional time series as the input to the LSTM model.

[0010] In this solution, S2 further includes: Construct an LSTM prediction model, which includes an input layer, two LSTM hidden layers, a fully connected layer, and an output layer. The input layer receives time series with an N×2 dimension, corresponding to two indicators: network pressure and server pressure at N time points. LSTM hidden layers are used to extract long-short-term dependency features from time-series data; The output layer outputs the pressure prediction values ​​for the next M time points; the arithmetic mean of the M prediction values ​​is taken as the long-term pressure parameter of a certain set of interactive data, including the long-term network pressure value and the long-term server pressure value.

[0011] In this solution, S3 specifically refers to: The short-term and long-term pressure parameters are numerically vectorized to generate a real-time pressure feature vector. We introduce standard pressure eigenvalues ​​and real-time pressure eigenvectors to perform deviation analysis based on Manhattan distance, and characterize the overall pressure of real-time interactive data on enterprise servers based on the deviation. Based on the degree of deviation, the real-time interactive data is classified and categorized.

[0012] In this solution, S4 specifically refers to: Based on the hierarchical classification of multiple sets of interactive data, group labels are associated with priorities. In the next data interaction cycle, real-time interactive data from different groups are classified and cached based on a multi-level caching mechanism, and the highest-level data is uploaded to the cloud data terminal.

[0013] In this scheme, the multi-level caching mechanism in S4 is specifically as follows: The multi-level caching mechanism adopts a three-level caching architecture. The first-level cache is a memory-level cache, deployed in the server memory, used to store high-priority data. The second-level cache is a local disk cache, deployed on the server's local disk array, with access latency in the sub-second range; Tier 3 storage is cloud data endpoint storage, deployed in cloud-based object storage services, enabling multi-copy data backup and cross-node sharing.

[0014] A second aspect of the present invention also provides an enterprise server data optimization platform applicable to multiple scenarios. The platform includes: a memory, a processor, a user data terminal, and a cloud data terminal. The memory includes an enterprise server data optimization program applicable to multiple scenarios. When executed by the processor, the enterprise server data optimization program applicable to multiple scenarios performs the following steps: S1. During a data interaction cycle on the enterprise server, monitor the data interaction between the user data terminal and the cloud data terminal, and classify and label the real-time interaction data to obtain multiple sets of interaction data. S2. During the interactive monitoring process, analyze the network pressure monitoring pressure and data interaction pressure monitoring of the user data terminal for each set of interactive data in a single time node to obtain short-term pressure parameters. Based on LSTM prediction, serialize and predict long-term features of the short-term pressure parameters corresponding to multiple time nodes to obtain long-term pressure parameters. S3. Construct a real-time pressure feature vector by using short-term and long-term pressure parameters, introduce standard pressure feature values ​​and real-time pressure features for deviation comparison, evaluate the overall pressure of real-time interactive data on enterprise servers, and classify the real-time interactive data. S4. Based on the hierarchical classification, map the data to the data tags. In the next data interaction cycle, set up a multi-level caching mechanism to classify and cache the real-time interactive data, and upload the highest-level data to the cloud data terminal.

[0015] A third aspect of the present invention also provides a computer-readable storage medium comprising an enterprise server data optimization program applicable to multiple scenarios, wherein when the enterprise server data optimization program applicable to multiple scenarios is executed by a processor, the program implements the steps of the enterprise server data optimization method applicable to multiple scenarios as described in any of the preceding claims.

[0016] This invention discloses an enterprise server data optimization platform applicable to multiple scenarios, comprising: monitoring bidirectional interactive data during the data interaction cycle and setting data labels through multi-dimensional classification; calculating short-term server pressure parameters at a single time node, predicting long-term pressure trends based on an LSTM neural network, and then fusing short- and long-term pressure features to achieve data priority classification by calculating the deviation from standard pressure values; finally, in the next cycle, executing a differentiated storage strategy through a multi-level caching architecture, with high-priority data synchronized to the cloud and low-priority data cached locally only. This invention combines short- and long-term dual-dimensional pressure assessment with a graded caching mechanism to effectively improve server data access efficiency, reduce bandwidth consumption, and adapt to the data storage optimization needs of multiple scenarios such as e-commerce, finance, and manufacturing. Attached Figure Description

[0017] Figure 1 The flowchart of an enterprise server data optimization method applicable to multiple scenarios according to the present invention is shown; Figure 2 The flowchart of the short-time pressure time series prediction input construction process of the present invention is shown; Figure 3 The diagram shows a block diagram of an enterprise server data optimization platform applicable to multiple scenarios according to the present invention. Detailed Implementation

[0018] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0019] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0020] Figure 1 The flowchart illustrates a method for optimizing enterprise server data in multiple scenarios according to the present invention.

[0021] like Figure 1 As shown, the first aspect of the present invention provides a method for optimizing enterprise server data in multiple scenarios, including: S1. During a data interaction cycle on the enterprise server, monitor the data interaction between the user data terminal and the cloud data terminal, and classify and label the real-time interaction data to obtain multiple sets of interaction data. S2. During the interactive monitoring process, analyze the network pressure monitoring pressure and data interaction pressure monitoring of the user data terminal for each set of interactive data in a single time node to obtain short-term pressure parameters. Based on LSTM prediction, serialize and predict long-term features of the short-term pressure parameters corresponding to multiple time nodes to obtain long-term pressure parameters. S3. Construct a real-time pressure feature vector by using short-term and long-term pressure parameters, introduce standard pressure feature values ​​and real-time pressure features for deviation comparison, evaluate the overall pressure of real-time interactive data on enterprise servers, and classify the real-time interactive data. S4. Based on the hierarchical classification, map the data to the data tags. In the next data interaction cycle, set up a multi-level caching mechanism to classify and cache the real-time interactive data, and upload the highest-level data to the cloud data terminal.

[0022] According to an embodiment of the present invention, in step S1, the real-time interactive data includes core transaction and order full-link data, real-time user behavior and interaction data, server and application performance operation and maintenance data, and real-time network traffic and security data.

[0023] It should be noted that the core transaction and order end-to-end data includes structured data covering the entire process of order placement, payment, inventory deduction, and logistics synchronization. Real-time user behavior and interaction data includes logs of product browsing, searching, favorites, adding to cart, and rating activities. Server and application performance and maintenance data includes operational metrics such as CPU utilization, memory utilization, disk I / O utilization, and interface response time for each service node. Real-time network traffic and security data includes inbound traffic, outbound traffic, attack interception records, and abnormal request logs. Enterprise servers include both user data endpoints and cloud data endpoints; the user data endpoint includes local databases and user interaction endpoints.

[0024] It is worth mentioning that in traditional e-commerce enterprise data solutions, the storage solutions for data on the enterprise data terminal and in the cloud are inefficient. Data is generally uploaded and downloaded in a unified manner, lacking caching and matching mechanisms. This results in low storage access efficiency for some data that is not suitable for immediate upload to the cloud, and makes it difficult to guarantee data analysis and efficient decision support in the cloud. This invention can effectively solve the above problems, is suitable for the data storage optimization needs of e-commerce servers in multiple scenarios, and effectively improves data interaction efficiency.

[0025] According to an embodiment of the present invention, step S1 further includes: The dimensions for classifying real-time interactive data include user geographic location, product category preference, access time period, and data volume per access. During a data interaction cycle, multiple sets of interaction data are obtained by classifying real-time interaction data from different users. The same set of interaction data has similar access frequency, data volume, and data pressure characteristics.

[0026] It should be noted that one or more dimensions can be used for classification here, such as user geographic location, product category preference, access time period, and data volume per access. Generally speaking, one dimension can be selected for classification to construct user groups and user interaction data groups with the same characteristics.

[0027] Here, the data tags include four core fields: group ID, data type code, category dimension attribute, and data generation timestamp. The data tags are bound to the corresponding interactive data and stored throughout the entire lifecycle of data collection, stress assessment, classification, and caching, and are used for full-link data tracking and policy mapping.

[0028] In this invention, the enterprise server can generally be an e-commerce server. In e-commerce enterprise servers or other service platforms with cloud requirements, there is a large amount of interactive data between the server and the user end. Especially for large-scale users with real-time synchronous access cycles, core data often needs to be uploaded to the cloud in real time for decision analysis or real-time storage based on data consistency requirements. However, traditional data storage is often based on fixed periodic data storage in the cloud, lacking data optimization and priority analysis processes, resulting in low storage efficiency. It is only suitable for single application scenarios with low complexity of user interaction data and has a certain data storage lag.

[0029] Figure 2 The flowchart of the short-time pressure time series prediction input construction process of the present invention is shown; According to an embodiment of the present invention, step S2 specifically includes: Short-term stress parameters include short-term network stress values ​​and short-term server stress values; The short-term network pressure value is obtained by weighting three indicators after normalization: the average response time of a set of interactive data at a single time node, the request success rate, and the data traffic consumption. The short-term server pressure value is obtained by weighted calculation of three indicators of enterprise server: CPU utilization, memory utilization, and disk I / O utilization, after normalization. Using a set of interactive data as the target data for pressure monitoring, we analyze the corresponding short-term pressure parameters at N time points. We then arrange the short-term pressure parameters of N consecutive time points in chronological order to construct a multi-dimensional time series as the input to the LSTM model.

[0030] It should be noted that for a data interaction cycle, multiple time points are divided for short-term stress monitoring and assessment. The average response time, request success rate, and data traffic consumption are normalized and weighted with weights of 0.4, 0.3, and 0.3, respectively. CPU utilization, memory utilization, and disk I / O utilization are normalized and weighted with weights of 0.7, 0.2, and 0.1, respectively. Short-term stress values ​​range from 0 to 1, with higher values ​​indicating higher stress. The duration of a single time point is adjustable from 30 to 600 seconds, and can be adjusted based on instantaneous server stress. N ranges from 6 to 30, corresponding to the sliding window length of the prediction model. M ranges from 10 to 60.

[0031] According to an embodiment of the present invention, step S2 further includes: Construct an LSTM prediction model, which includes an input layer, two LSTM hidden layers, a fully connected layer, and an output layer. The input layer receives time series with an N×2 dimension, corresponding to two indicators: network pressure and server pressure at N time points. LSTM hidden layers are used to extract long-short-term dependency features from time-series data; The output layer outputs the pressure prediction values ​​for the next M time points; the arithmetic mean of the M prediction values ​​is taken as the long-term pressure parameter of a certain set of interactive data, including the long-term network pressure value and the long-term server pressure value.

[0032] It should be noted that the LSTM neural network model is pre-trained using historical stress datasets, and the Adam optimizer is used during the training process. The loss function is the mean squared error (MSE). After the model is trained, the prediction accuracy is verified using a test set. When the prediction error rate is below 5%, it is put into online use. During online operation, an incremental learning mechanism is adopted, and the model is fine-tuned daily using the latest stress data to ensure that the prediction accuracy remains stable as business changes.

[0033] Here, network stress monitoring mainly includes average data response time, request success rate, and data traffic consumption. Data interaction stress monitoring mainly includes the overall utilization of the server's CPU, memory, and disk, as well as stress assessment. Short-term stress parameters include short-term network stress and short-term server stress values, while long-term stress parameters include long-term network stress and long-term server stress values. All stress values ​​are comprehensive assessment values.

[0034] According to an embodiment of the present invention, step S3 specifically includes: The short-term and long-term pressure parameters are numerically vectorized to generate a real-time pressure feature vector. We introduce standard pressure eigenvalues ​​and real-time pressure eigenvectors to perform deviation analysis based on Manhattan distance, and characterize the overall pressure of real-time interactive data on enterprise servers based on the deviation. Based on the degree of deviation, the real-time interactive data is classified and categorized.

[0035] It should be noted that the real-time pressure feature vector includes four dimensions: two dimensions of short-term pressure and two dimensions of long-term pressure. The standard pressure feature value is the standard vector. The standard vector is the pressure feature vector of an enterprise server under ideal conditions.

[0036] Data prioritization refers to the classification of data priorities. Data with a large deviation indicates that it is frequently referenced and has a high probability of repeated access in the short to long term. Under high server pressure, it is necessary to perform repeated data verification and reuse. In this case, a lower priority can be set, and it can be stored in a cache library with a high reference frequency in the caching mechanism. Data with a small deviation indicates that it puts less pressure on the data end of real-time interaction. A higher priority can be set, and it can be considered to store it on the cloud data end to ensure the consistency of data on the user data end, enterprise end, and cloud data end under the requirement of efficient storage.

[0037] Real-time stress characteristics are used to comprehensively assess the processing pressure and instantaneous and long-term fluctuations of enterprise servers on real-time interactive data, and to subsequently perform tiered storage and tiered caching of the data.

[0038] The degree of deviation is represented by Manhattan distance. The threshold for classifying the degree of deviation can be determined based on the actual classification hierarchy. For example, if a three-level priority is set, two thresholds can be set based on Manhattan distance to filter and classify all multiple sets of interactive data.

[0039] According to an embodiment of the present invention, step S4 specifically includes: Based on the hierarchical classification of multiple sets of interactive data, group labels are associated with priorities. In the next data interaction cycle, real-time interactive data from different groups are classified and cached based on a multi-level caching mechanism, and the highest-level data is uploaded to the cloud data terminal.

[0040] It should be noted that the classification results are mapped to the priority field of the data tag. At the beginning of the next data interaction cycle, the system reads the priority field in the data tag and automatically matches the corresponding cache and storage path. The duration of the data interaction cycle is configurable. At the end of the cycle, the classification calculation of all data in this cycle is completed in a unified manner, and a new classification strategy can be loaded when the next cycle starts.

[0041] According to an embodiment of the present invention, the multi-level caching mechanism in S4 is specifically as follows: The multi-level caching mechanism adopts a three-level caching architecture. The first-level cache is a memory-level cache, deployed in the server memory, used to store high-priority data. The second-level cache is a local disk cache, deployed on the server's local disk array, with access latency in the sub-second range; Tier 3 storage is cloud data endpoint storage, deployed in cloud-based object storage services, enabling multi-copy data backup and cross-node sharing.

[0042] Here, the three-level caching mechanism is just an example. Based on the scale and complexity of the stored data, more levels of mechanisms can be set up, such as a five-level mechanism. In the multi-level setting, the higher the level of data storage, the greater the demand for storing data on the cloud data terminal.

[0043] According to embodiments of the present invention, a differentiated caching strategy is also included: For high-priority data, it is written to both the first-level memory cache and the cloud data storage. The memory cache is set with an expiration and eviction mechanism, and the LRU eviction algorithm is used to manage the cache space. The cloud storage uses an incremental synchronization method, only synchronizing the changed data fields. For medium-priority data, it is written to the secondary local disk cache and a timed batch synchronization mechanism is adopted. The accumulated medium-priority data is uploaded to the cloud data terminal in batches at set intervals to reduce network connection establishment overhead. For low-priority data, it is only written to the second-level local disk cache and is not actively synchronized to the cloud. An observation period is set. If the deviation continues to decrease and reaches the medium-priority standard within several consecutive interaction periods, it is upgraded to medium-priority and the corresponding synchronization strategy is executed.

[0044] Differentiated caching strategies can be directly added to multi-level caching mechanisms to achieve efficient storage and accurate data upload to the cloud.

[0045] In this invention, the differentiated caching strategy addresses the problems in traditional technologies where, under the indiscriminate caching and full cloud synchronization modes, core high-value data and low-value fluctuating data share storage and bandwidth resources, resulting in low memory cache hit rates, poor bandwidth utilization, and a mismatch between storage resource investment and business returns. It also solves the problems of redundant and insufficient timeliness in cloud synchronization of core data. Traditional technologies lack an effective hierarchical system and rely on fixed-period data uploads to cloud storage, leading to lag in cloud synchronization of core data and affecting the real-time performance of cloud risk control and operational decisions.

[0046] Meanwhile, the differentiated caching strategy enables high-priority data (core transaction full-link data, and stable user group behavior data) to reside in the first-level memory cache, with access latency controlled at the 1~10ms level. Compared with local disk caching, the access speed is significantly improved, which can support the processing needs of thousands of orders per second during peak promotion periods. The core link response timeout rate is reduced by more than 20%, achieving a dual improvement in cloud synchronization efficiency and timeliness, and significantly improving the accuracy and response speed of cloud data terminal decision-making.

[0047] Figure 3 The diagram shows a block diagram of an enterprise server data optimization platform applicable to multiple scenarios according to the present invention.

[0048] A second aspect of the present invention also provides an enterprise server data optimization platform applicable to multiple scenarios. The platform includes: a memory, a processor, a user data terminal, and a cloud data terminal. The memory includes an enterprise server data optimization program applicable to multiple scenarios. When executed by the processor 22, the enterprise server data optimization program applicable to multiple scenarios performs the following steps: S1. During a data interaction cycle on the enterprise server, monitor the data interaction between the user data terminal and the cloud data terminal, and classify and label the real-time interaction data to obtain multiple sets of interaction data. S2. During the interactive monitoring process, analyze the network pressure monitoring pressure and data interaction pressure monitoring of the user data terminal for each set of interactive data in a single time node to obtain short-term pressure parameters. Based on LSTM prediction, serialize and predict long-term features of the short-term pressure parameters corresponding to multiple time nodes to obtain long-term pressure parameters. S3. Construct a real-time pressure feature vector by using short-term and long-term pressure parameters, introduce standard pressure feature values ​​and real-time pressure features for deviation comparison, evaluate the overall pressure of real-time interactive data on enterprise servers, and classify the real-time interactive data. S4. Based on the hierarchical classification, map the data to the data tags. In the next data interaction cycle, set up a multi-level caching mechanism to classify and cache the real-time interactive data, and upload the highest-level data to the cloud data terminal.

[0049] The platform can perform one or more steps of the enterprise server data optimization method described above, which is applied to multiple scenarios.

[0050] A third aspect of the present invention also provides a computer-readable storage medium comprising an enterprise server data optimization program applicable to multiple scenarios, wherein when the enterprise server data optimization program applicable to multiple scenarios is executed by a processor, the program implements the steps of the enterprise server data optimization method applicable to multiple scenarios as described in any of the preceding claims.

[0051] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0052] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0053] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0054] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0055] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0056] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for optimizing enterprise server data in multiple scenarios, characterized in that, include: S1. During a data interaction cycle on the enterprise server, monitor the data interaction between the user data terminal and the cloud data terminal, and classify and label the real-time interaction data to obtain multiple sets of interaction data. S2. During the interactive monitoring process, analyze the network pressure monitoring pressure and data interaction pressure monitoring of the user data terminal for each set of interactive data in a single time node to obtain short-term pressure parameters. Based on LSTM prediction, serialize and predict long-term features of the short-term pressure parameters corresponding to multiple time nodes to obtain long-term pressure parameters. S3. Construct a real-time pressure feature vector by using short-term and long-term pressure parameters, introduce standard pressure feature values ​​and real-time pressure features for deviation comparison, evaluate the overall pressure of real-time interactive data on enterprise servers, and classify the real-time interactive data. S4. Based on the hierarchical classification, map the data to the data tags. In the next data interaction cycle, set up a multi-level caching mechanism to classify and cache the real-time interactive data, and upload the highest-level data to the cloud data terminal.

2. The enterprise server data optimization method applied in multiple scenarios according to claim 1, characterized in that, In S1, the real-time interactive data includes core transaction and order full-link data, real-time user behavior and interaction data, server and application performance operation and maintenance data, and real-time network traffic and security data.

3. The enterprise server data optimization method applied in multiple scenarios according to claim 2, characterized in that, The S1 further includes: The dimensions for classifying real-time interactive data include user geographic location, product category preference, access time period, and data volume per access. During a data interaction cycle, multiple sets of interaction data are obtained by classifying real-time interaction data from different users. The same set of interaction data has similar access frequency, data volume, and data pressure characteristics.

4. The enterprise server data optimization method applied in multiple scenarios according to claim 3, characterized in that, Specifically, S2 is: Short-term stress parameters include short-term network stress values ​​and short-term server stress values; The short-term network pressure value is obtained by weighted calculation of three indicators after normalization: the average response time of a set of interactive data at a single time node, the request success rate, and the data traffic consumption. The short-term server pressure value is obtained by weighted calculation of three indicators of enterprise server: CPU utilization, memory utilization, and disk I / O utilization, after normalization. Using a set of interactive data as the target data for pressure monitoring, we analyze the corresponding short-term pressure parameters at N time points. We then arrange the short-term pressure parameters of N consecutive time points in chronological order to construct a multi-dimensional time series as the input to the LSTM model.

5. The enterprise server data optimization method applied in multiple scenarios according to claim 4, characterized in that, The S2 further includes: Construct an LSTM prediction model, which includes an input layer, two LSTM hidden layers, a fully connected layer, and an output layer. The input layer receives time series with an N×2 dimension, corresponding to two indicators: network pressure and server pressure at N time points. LSTM hidden layers are used to extract long-short-term dependency features from time-series data; The output layer outputs the pressure prediction values ​​for the next M time points; the arithmetic mean of the M prediction values ​​is taken as the long-term pressure parameter of a certain set of interactive data, including the long-term network pressure value and the long-term server pressure value.

6. The enterprise server data optimization method applied in multiple scenarios according to claim 5, characterized in that, Specifically, S3 is: The short-term and long-term pressure parameters are numerically vectorized to generate a real-time pressure feature vector. We introduce standard pressure eigenvalues ​​and real-time pressure eigenvectors to perform deviation analysis based on Manhattan distance, and characterize the overall pressure of real-time interactive data on enterprise servers based on the deviation. Based on the degree of deviation, the real-time interactive data is classified and categorized.

7. The enterprise server data optimization method applied in multiple scenarios according to claim 6, characterized in that, Specifically, S4 is: Based on the hierarchical classification of multiple sets of interactive data, group labels are associated with priorities. In the next data interaction cycle, real-time interactive data from different groups are classified and cached based on a multi-level caching mechanism, and the highest-level data is uploaded to the cloud data terminal.

8. The enterprise server data optimization method applied in multiple scenarios according to claim 7, characterized in that, The multi-level caching mechanism in S4 is specifically as follows: The multi-level caching mechanism adopts a three-level caching architecture. The first-level cache is a memory-level cache, deployed in the server memory, used to store high-priority data. The second-level cache is a local disk cache, deployed on the server's local disk array, with access latency in the sub-second range; Tier 3 storage is cloud data endpoint storage, deployed in cloud-based object storage services, enabling multi-copy data backup and cross-node sharing.

9. A data optimization platform for enterprise servers applied in multiple scenarios, characterized in that, The platform includes: a memory, a processor, a user data terminal, and a cloud data terminal. The memory includes an enterprise server data optimization program applicable to multiple scenarios. When the processor executes the enterprise server data optimization program applicable to multiple scenarios, it performs the following steps: S1. During a data interaction cycle on the enterprise server, monitor the data interaction between the user data terminal and the cloud data terminal, and classify and label the real-time interaction data to obtain multiple sets of interaction data. S2. During the interactive monitoring process, analyze the network pressure monitoring pressure and data interaction pressure monitoring of the user data terminal for each set of interactive data in a single time node to obtain short-term pressure parameters. Based on LSTM prediction, serialize and predict long-term features of the short-term pressure parameters corresponding to multiple time nodes to obtain long-term pressure parameters. S3. Construct a real-time pressure feature vector by using short-term and long-term pressure parameters, introduce standard pressure feature values ​​and real-time pressure features for deviation comparison, evaluate the overall pressure of real-time interactive data on enterprise servers, and classify the real-time interactive data. S4. Based on the hierarchical classification, map the data to the data tags. In the next data interaction cycle, set up a multi-level caching mechanism to classify and cache the real-time interactive data, and upload the highest-level data to the cloud data terminal.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes an enterprise server data optimization program applicable to multiple scenarios. When the enterprise server data optimization program applicable to multiple scenarios is executed by a processor, it implements the steps of the enterprise server data optimization method applicable to multiple scenarios as described in any one of claims 1 to 8.

Images