Server resource operation risk early warning method and device and electronic equipment

Abstract

The invention discloses a server resource operation risk early warning method and device and electronic equipment, and relates to the technical field of server resources, and the method comprises the steps: obtaining resource parameters including a current state parameter and a historical state parameter, determining a prediction state parameter according to the historical state parameter, and providing an expected reference standard which accords with resource operation characteristics; by determining the state difference index between the current state parameter and the predicted state parameter, the deviation degree between the actual state and the expected state can be quantified, and a reliable basis is provided for risk degree determination, so that the risk state parameter is determined based on the state difference index, and the target early warning parameter is determined according to the risk state parameter. According to the invention, accurate matching of the early warning information and the actual risk can be realized, the technical problem of inaccurate server resource operation risk early warning in related technologies is effectively solved, and the accuracy of operation risk early warning is improved.

Description

Technical Field

[0001] This application relates to the field of server resource technology, and in particular to methods, devices and electronic equipment for early warning of operational risks of server resources. Background Technology

[0002] In related technologies, the operational status of server resources directly affects the stability and service continuity of the server or the system it corresponds to. Fluctuations in load and abnormal resource usage can lead to performance degradation and service interruptions. Therefore, it is necessary to provide early warnings of operational risks to server resources to identify potential problems in advance and take timely countermeasures. However, in related technologies, there are technical issues with the accuracy of these early warnings.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This application provides a method, apparatus, and electronic device for early warning of operational risks of server resources, in order to at least solve the technical problem in the related art that the early warning of operational risks of server resources is inaccurate.

[0005] This application provides a method for early warning of operational risks of server resources, comprising: acquiring resource parameters of the server resources, wherein the resource parameters include current state parameters and historical state parameters; determining a predicted state parameter corresponding to the server resources based on the historical state parameters; determining a state difference index between the current state parameters and the predicted state parameters; determining a risk state parameter corresponding to the server resources based on the state difference index; determining a target early warning parameter corresponding to the server resources based on the risk state parameter; and providing an operational risk warning for the server resources based on the target early warning parameter.

[0006] This application also provides a server resource operation risk early warning device, comprising: an acquisition module for acquiring resource parameters of the server resource, wherein the resource parameters include current state parameters and historical state parameters; a first determination module for determining a predicted state parameter corresponding to the server resource based on the historical state parameters; a second determination module for determining a state difference index between the current state parameter and the predicted state parameter; a third determination module for determining a risk state parameter corresponding to the server resource based on the state difference index; a fourth determination module for determining a target early warning parameter corresponding to the server resource based on the risk state parameter; and a fifth determination module for performing an operation risk early warning on the server resource based on the target early warning parameter.

[0007] This application also provides an electronic device, including: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the server resource operation risk warning method described above.

[0008] This application also provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the server resource operation risk warning method described above.

[0009] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the above-described server resource operation risk warning methods.

[0010] By acquiring the current and historical status parameters of server resources, comprehensive data support can be provided for subsequent accurate predictions. Determining the predicted status parameters based on historical parameters allows for a reasonable estimation of the future operating status of server resources based on past data patterns. Furthermore, by determining the state difference index between the current and predicted status parameters, the degree of deviation between the current operating status and the expected normal state can be quantified. Based on this, risk status parameters are determined according to the state difference index, accurately defining the risk level and impact range of server resources. Then, target warning parameters are determined based on the risk status parameters, ensuring that the warning information matches the actual risk. Finally, operational risk warnings are issued based on the target warning parameters, avoiding inaccurate warnings caused by information bias and achieving precise control over the operational risks of server resources. Attached Figure Description

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

[0012] Figure 1 This is a hardware structure block diagram of the server resource operation risk warning method according to an embodiment of this application;

[0013] Figure 2 This is a flowchart of a server resource operation risk warning method according to an embodiment of this application;

[0014] Figure 3 This is a flowchart of the server resource operation risk warning method in an optional implementation of this application;

[0015] Figure 4This is a schematic diagram of the anomaly determination and handling process in the optional implementation of this application;

[0016] Figure 5 This is a schematic diagram of the operational risk early warning system structure in an optional implementation of this application;

[0017] Figure 6 This is a structural block diagram of a server resource operation risk warning device according to an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0019] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0020] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] The specific application environment architecture or specific hardware architecture on which the execution of the risk warning method for server resources depends is described here.

[0022] The methods and embodiments provided in this application can be executed on a server device or a similar computing device. Taking running on a server device as an example, Figure 1 This is a hardware structure block diagram of the server resource operation risk warning method according to an embodiment of this application, such as... Figure 1 As shown, the server device may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. The server device may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1The structure shown is for illustrative purposes only and does not limit the structure of the server equipment described above. For example, the server equipment may also include components that are more... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0023] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the server resource operation risk warning method in this embodiment. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thus implementing the above-described method. The memory 104 may include high-speed random access memory and non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the server device via a network. Examples of the aforementioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof. The transmission device 106 is used to receive or send data via a network. Specific examples of the aforementioned networks may include wireless networks provided by the communication provider of the server device. In one instance, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0024] According to an embodiment of this application, a method embodiment for a server resource operation risk warning method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0025] Figure 2 This is a flowchart of a server resource operation risk warning method according to an embodiment of this application, such as... Figure 2 As shown, the method includes the following steps:

[0026] S202, Obtain the resource parameters of the server resources, including current status parameters and historical status parameters;

[0027] This involves server resources, which are the various hardware and software resources used to support the operation of the server or its corresponding system, including multiple resource items (i.e., multi-dimensional performance indicators). These multiple resource items are those whose state characterization indices are greater than a state characterization threshold. The state characterization index is used to represent the degree of characterization of the server resource's operational status. These multiple resource items include, but are not limited to, CPU utilization, memory usage, disk read / write speed, and network communication bandwidth. The operational status of these resource items directly affects the system's stability and service continuity. By filtering resource items whose state characterization indices are greater than the threshold and obtaining their current and historical state parameters, we can focus on the core indicators that have a critical impact on the server resource's operational status, avoiding irrelevant indicators from interfering with the analysis.

[0028] This involves resource parameters, which are data used to describe the operating status of server resources, including current status parameters and historical status parameters.

[0029] This involves current status parameters, which are real-time collected server resource performance data that reflect the actual operating status of the resources at the current moment. For example, the collected current multi-dimensional performance index data (including data corresponding to CPU utilization, memory usage, disk read / write speed, network communication bandwidth, etc.) are uniformly timestamped and standardized to form the current detection dataset (i.e., the current status parameters).

[0030] This involves historical status parameters, which are historical detection datasets of server resources collected and stored periodically or in real time up to the current moment. These datasets contain records of the server resources' operational status at different historical time periods.

[0031] By acquiring the current and historical status parameters of server resources, comprehensive real-time and historical data support can be provided for subsequent analysis, ensuring a sufficient data foundation for predicting future states based on historical patterns. At the same time, dynamic risk identification can be achieved through real-time data comparison, avoiding prediction bias caused by missing or incomplete data, and ultimately achieving accurate early warning of server resource operation risks.

[0032] S204, Based on historical status parameters, determine the predicted status parameters corresponding to the server resources;

[0033] This involves predicted state parameters, which are the expected resource state parameters of the server resources at a predetermined time (such as the current time) based on historical state parameters. In other words, these predicted state parameters reflect the theoretical state value of the server resources at the predetermined time, representing a reasonable estimate of the future operating state of the server resources. They reflect the state that the resources should exhibit under normal circumstances, providing a clear reference standard for subsequent comparison with current state parameters and determining whether there are any anomalies in the resources.

[0034] Historical state parameters can comprehensively reflect the historical operating patterns of server resources. Based on these historical state parameters, predicted state parameters can be determined, generating reasonable estimates that conform to the characteristics of server resource changes. This provides a clear reference for comparing current state parameters with expected states.

[0035] S206, Determine the state difference index between the current state parameters and the predicted state parameters;

[0036] This involves a state difference index, which quantifies the degree of deviation between current state parameters and predicted state parameters. By calculating the difference between the actual operating state and the theoretically expected state, it reflects the deviation of the server resources' actual state from the normal predicted value at a certain moment, providing direct numerical basis for subsequent risk status assessment.

[0037] By determining the state difference index between the current state parameters and the predicted state parameters, the deviation between the actual operating state and the theoretical expectation can be quantified, ensuring accurate identification of abnormal server resource states, and thus providing reliable numerical basis for subsequent risk assessment.

[0038] Furthermore, when server resources include multiple resource items, for any one of these resource items, the state difference index between the current state parameter and the predicted state parameter corresponding to that resource item can be determined using the following formula:

[0039]

[0040] in, Let be the state difference index of the i-th resource item of the server resources at time t, which is dimensionless; Let be the predicted state parameter of the i-th resource item of the server resource at time t, which is dimensionless; Let be the state stability index of the i-th resource item of the server resources at time t, which is dimensionless; This is a dimensionless index representing the state fluctuation of server resources. Let be the prediction deviation of the i-th resource item of the server resources at time t, which is dimensionless; The adjustment coefficient is dimensionless. The adjustment coefficient is dimensionless and satisfies... , This is the reference state parameter for the i-th resource item of the server resources at time t.

[0041] S208, based on the state difference index, determine the risk state parameters corresponding to the server resources;

[0042] This involves risk status parameters, which are parameters used to represent the risk characteristics (i.e., abnormal characteristics) of the server resource's operating status, including risk level (i.e., abnormality level) and risk type (i.e., abnormality type).

[0043] Based on the deviation of the actual and expected states of resources quantified by the state difference index, the corresponding risk state parameters are determined to quantify the risk level (including the degree and type of risk) and accurately reflect the risk level of server resources, providing a basis for subsequent analysis.

[0044] Furthermore, based on the aforementioned state difference index between the current state parameter and the predicted state parameter of any resource item, the state difference indices corresponding to multiple resource items can be combined to determine the risk state parameter corresponding to the server resource. Thus, by determining the risk state parameter based on the state difference indices of multiple core resource items, the overall operational risk of the server resource can be comprehensively assessed from multiple dimensions, avoiding misjudgment of the overall risk level due to data anomalies corresponding to a single resource item.

[0045] S210, Based on the risk status parameters, determine the target early warning parameters corresponding to the server resources;

[0046] This involves target early warning parameters, which are the final early warning information parameters determined based on risk status parameters. These parameters include alarm levels (such as Level 1 severe anomaly warning, Level 2 moderate anomaly warning, and Level 3 mild anomaly warning), corresponding processing instructions (such as automatic capacity expansion, resource scheduling, and increased detection frequency), and early warning push methods (such as local messages, emails, SMS, and API call interfaces), which are used to directly trigger operation and maintenance response operations.

[0047] S212 provides early warnings of operational risks to server resources based on target early warning parameters.

[0048] This includes risk warning, which involves the entire process of executing risk warning operations based on target warning parameters.

[0049] Since target warning parameters can accurately reflect the key risk status of server resources, operational risk warnings for server resources can be issued based on target warning parameters. Potential resource bottlenecks or anomalies can be identified in advance, avoiding service unavailability caused by resource overload or failure, and achieving efficient management and continuous availability of server resources.

[0050] Based on the above steps, by acquiring the current and historical state parameters of server resources, comprehensive data support can be provided for subsequent accurate predictions. Determining the predicted state parameters based on historical state parameters allows for a reasonable estimation of the future operating state of server resources based on past data patterns. Furthermore, by determining the state difference index between the current and predicted state parameters, the degree of deviation between the current operating state and the expected normal state can be quantified. On this basis, risk state parameters can be determined based on the state difference index, which can accurately define the risk level and scope of impact of server resources. Then, target warning parameters can be determined based on the risk state parameters, ensuring that the warning information matches the actual risk. Finally, operational risk warnings can be issued based on the target warning parameters, avoiding inaccurate warnings caused by information bias and achieving precise control over the operational risks of server resources.

[0051] As an optional embodiment, determining the predicted state parameters corresponding to the server resources based on historical state parameters includes: decomposing the historical state parameters to obtain a first state component, a second state component, and a third state component corresponding to the server resources, wherein the first state component represents the trend characteristics of the server resources' operating state, the second state component represents the stability characteristics of the server resources' operating state, and the third state component represents the fluctuation characteristics of the server resources' movement state; determining the temporal state characteristics corresponding to the server resources based on the first state component, the second state component, and the third state component, wherein the temporal state characteristics represent the resource state of the server resources from a time dimension; and determining the predicted state parameters corresponding to the server resources based on the temporal state characteristics.

[0052] This involves decomposition processing, which is the process of breaking down historical state parameters into different feature components. Specifically, this decomposition processing is the process of structurally splitting the historical state parameters of server resources from different analytical dimensions in order to extract different dimensional features of resource operation.

[0053] This involves a first state component, which is a feature quantity that reflects the long-term changing trend of the server resource operating status.

[0054] This involves a second state component, which is a characteristic quantity that characterizes the periodic stability of the server resource operating state (that is, the periodic repetitive change pattern).

[0055] This involves a third state component, which is a feature quantity that describes the characteristics of random fluctuations (e.g., sudden fluctuations) in the operating state of server resources.

[0056] This involves temporal state characteristics, which are determined by combining the first state component, the second state component, and the third state component. These characteristics comprehensively reflect the changing patterns of server resources over time, thus depicting the changing patterns of resource states over time from multiple dimensions.

[0057] By decomposing historical state parameters, we obtain a first state component reflecting long-term trends, a second state component characterizing periodic stability, and a third state component describing random fluctuations. This allows us to capture the time dependence of resource states from different dimensions. Based on these three components, we can determine the time-series state characteristics, comprehensively integrating the changing patterns of resources over time. Furthermore, based on these time-series state characteristics, we can determine the predicted state parameters, ensuring that the prediction results align with the actual operating logic of resources and avoiding prediction biases caused by single-dimensional feature analysis. This provides a more reliable predictive reference for subsequent state comparisons and risk assessments.

[0058] As an optional embodiment, determining the temporal state features corresponding to the server resource based on the first state component, the second state component, and the third state component includes: determining the target state component corresponding to the server resource based on the first state component, the second state component, and the third state component; determining the first temporal feature corresponding to the server resource based on the first target sub-feature according to the execution order of the multiple target sub-features included in the target state component; and determining the next temporal feature corresponding to the server resource based on the first temporal feature and the next target sub-feature, until the multiple target sub-features are executed, thereby obtaining the temporal state features corresponding to the server resource.

[0059] This involves a target state component, which is an integrated feature set formed by fusing the first state component (trend characteristics), the second state component (stability characteristics), and the third state component (fluctuation characteristics).

[0060] This involves multiple target sub-features, which are state features at multiple time points (including the first time point, the second time point, etc.) corresponding to the target state components. These include trend sub-features, stability sub-features, and fluctuation sub-features, which are used to analyze the temporal evolution of resource states layer by layer.

[0061] This involves the execution order, which is a preset order in which multiple target sub-features are processed sequentially. This execution order can be chronological.

[0062] This involves the first target sub-feature, which is the target sub-feature that is ranked first in the execution order.

[0063] This involves a first temporal feature, which is a resource status temporal feature of server resources at a first time point extracted based on the first target sub-feature.

[0064] This involves the next target sub-feature, which is the next target sub-feature arranged in the execution order after the first target sub-feature.

[0065] This involves the next time-series feature, which is generated by fusing the previous time-series feature (such as the first time-series feature) with the next target sub-feature, and is the resource status time-series feature of the server resources at the next time point.

[0066] By fusing the first, second, and third state components to obtain an integrated target state component, the trend, stability, and fluctuation attributes of resources can be comprehensively represented. Then, multiple target sub-features included in the target state component are processed sequentially. Based on the first target sub-feature, the first time-series feature at the first time point is determined. Subsequently, by fusing the previous time-series feature with the next target sub-feature, the next time-series feature at the next time point is generated, until all target sub-features are traversed. This allows for layer-by-layer analysis of the evolution of server resource states over time, ensuring that the time-series state features fully cover the details of resource changes at different time points. This avoids inaccurate characterization of time-series patterns due to feature omissions or logical confusion, providing a comprehensive and realistic time-series basis for accurately determining subsequent predicted state parameters.

[0067] As an optional embodiment, determining the temporal state features corresponding to the server resources based on the first state component, the second state component, and the third state component includes: retrieving a target model, wherein the target model includes target parameters, the target parameters are used for feature extraction from the time dimension, the target parameters are obtained by training initial parameters based on sample data, the sample data includes sample state components and reference temporal features corresponding to the sample state components; and determining the temporal state features corresponding to the server resources based on the first state component, the second state component, the third state component, and the target model.

[0068] This involves a target model, which is an algorithmic model used to extract server resource state features from a time dimension. This target model includes a Long Short-Term Memory (LSTM) network, which comprehensively analyzes the changing characteristics of server resources over time through target parameters, providing algorithmic support for determining the predicted state parameters. The LSTM network is a deep learning model that controls the transmission and updating of information flow through a gating mechanism, enabling it to learn long-term and short-term features in time-series data in a long-term, reliable manner.

[0069] This involves target parameters, which are model parameters optimized through training with sample data. The magnitude and combination of these target parameters determine the target model's ability to capture time-series patterns such as resource trends, stability, and fluctuations.

[0070] This involves initial parameters, which are the model parameters before the target model is trained. These initial parameters have not been optimized with sample data and only have basic feature extraction logic. They need to be continuously adjusted and updated through iterative learning with sample data to eventually form target parameters that can accurately capture temporal patterns.

[0071] This involves sample data, which is a dataset used to train the target model (essentially training the target parameters). It contains a large number of fully labeled sample instances, each of which includes sample state components and corresponding reference time-series features.

[0072] This involves sample state components, which are sample data used to simulate the actual state components of server resources. Their structure is consistent with the first state component, second state component, and third state component in the actual scenario (i.e., they include trend, stability, and fluctuation sample components), providing input samples for the target model to learn feature extraction logic.

[0073] This involves reference time-series features, which are standard time-series features in the sample data corresponding to the sample state components. These features represent the true time-series patterns that the sample state components should possess (such as the true changes in resources corresponding to the sample state components over time). They serve as the standard answer for model training and are used to measure the deviation between the model output features and the true features.

[0074] Because the target model, which contains target parameters, is retrieved and optimized based on initial parameters trained using sample data, and this sample data includes sample state components and corresponding reference temporal features, the target model can accurately capture the temporal patterns of server resources over time. Furthermore, by determining the temporal state features of server resources based on the first, second, and third state components and the target model, the output features can comprehensively represent the temporal evolution logic of the resources, avoiding feature extraction bias caused by inaccurate model parameters.

[0075] As an optional embodiment, before retrieving the target model, the method further includes: determining the predicted temporal features corresponding to the sample state component and the sub-predicted features corresponding to the multiple sample sub-features, based on multiple sample sub-features and a first arrangement order corresponding to the multiple sample sub-features, wherein the sample state component includes multiple sample sub-features; determining the feature deviation index between the reference temporal features and the predicted temporal features; determining the update index corresponding to the initial parameters based on the sub-predicted features corresponding to the multiple sample sub-features, a second arrangement order corresponding to the multiple sample sub-features, and the feature deviation index, wherein the update index is used to represent the degree of update of the initial parameters, wherein the second arrangement order and the first arrangement order have a time correlation; and updating the initial parameters according to the update index to obtain the target parameters.

[0076] This involves multiple sample sub-features, which are the state features of the sample state components at multiple sample time points and are the basic feature units that constitute the sample state components.

[0077] This involves a first permutation order, which represents the temporal order of multiple sample sub-features (e.g., from the first time point to the last time point). This first permutation order is used to standardize the input order of sample sub-features during forward training, ensuring that the model can learn the correlation between features according to the temporal evolution law, laying a logical foundation for accurately generating predicted temporal features. Forward training involves inputting multiple sample sub-features according to the first permutation order (the temporal order of the sample sub-features) based on initial parameters. The model's built-in algorithmic logic learns the temporal patterns corresponding to the sample state components, thereby generating predicted temporal features corresponding to the sample state components and sub-predicted features corresponding to each sample sub-feature. The core of this forward training is to use sample data to allow the model to initially learn the mapping relationship between sample sub-features and temporal features, providing a basic output for subsequent bias calculation and backpropagation.

[0078] This involves predicting temporal features, which are obtained by predicting the temporal features of each sample sub-feature in the first order of multiple sample sub-features, and finally obtaining the prediction output corresponding to the sample state component, so as to characterize the sample temporal pattern of the sample state component.

[0079] This involves sub-prediction features, which are the prediction outputs at the corresponding sample time points, such as the prediction outputs corresponding to the sample state components at sample time point 1.

[0080] This involves a second permutation order, which is a permutation order that has a temporal correlation with the first permutation order. The temporal correlation is that the time points are the same, but the traversal direction of the time points is opposite. For example, if the first permutation order is sample time point 1, sample time point 2, ..., sample time point T, then the second permutation order is sample time point T, sample time point T-1, ..., sample time point 1. This second permutation order is used during backpropagation training to match sample sub-features, sub-predicted features, and bias information across different time dimensions, ensuring that parameter updates align with the inherent logic of temporal evolution. The backpropagation training process uses a feature bias index as a basis, combined with the second permutation order, to match the sub-predicted features corresponding to each sample sub-feature to determine the update index used to adjust the model parameters. The initial parameters of the model are then corrected based on the update index until the deviation between the predicted temporal features output by the model and the reference temporal features reaches a preset threshold, ultimately yielding target parameters that accurately capture temporal patterns. The core of this backpropagation training is to trace the source of the deviation in reverse, gradually correct the model parameters, and make the model prediction results continuously approach the reference time series characteristics, so as to obtain the target parameters that can accurately capture the time series patterns.

[0081] This involves the feature deviation index, which is used to quantify the degree of deviation between the reference time-series features (true labels) and the predicted time-series features (positive output of the model). It directly reflects the gap between the current prediction effect of the target model and the true pattern, and clarifies the direction of the initial parameter update.

[0082] This involves an update index, which is calculated by combining sub-prediction features, the second permutation order, and the feature deviation index. This index represents the magnitude and direction of the initial parameter adjustment, so as to accurately map the gap between the initial parameters and the optimal target parameters, provide a clear adjustment basis for the iterative optimization of the initial parameters, and ensure that the parameter deviation is gradually corrected through backpropagation.

[0083] By inputting multiple sample sub-features in the first chronological order, the predicted temporal features corresponding to the sample state components and the sub-predicted features at each time point are determined. This allows the model to initially learn the mapping relationship between sample sub-features and temporal features based on initial parameters. Then, by calculating the feature deviation index between the reference temporal features and the predicted temporal features, the difference between the model's prediction results and the actual patterns can be accurately quantified. Subsequently, based on the second chronological order which is inversely related to the first chronological order, the sub-predicted features are matched and the update index is determined in conjunction with the feature deviation index. This clarifies the adjustment range and direction of the initial parameters. Finally, the initial parameters are iteratively updated based on the update index to obtain the target parameters. This ensures that the model parameters are gradually optimized to the optimal state, avoiding temporal feature extraction deviations caused by unoptimized parameters. This provides reliable algorithmic parameter support for the subsequent accurate determination of the temporal state features of server resources based on the target model.

[0084] As an optional embodiment, determining the state difference index between the current state parameters and the predicted state parameters includes: determining the state stability index and the state fluctuation index corresponding to the server resources under the current state parameters; determining the reference state parameters corresponding to the server resources based on the state stability index and the state fluctuation index; and determining the state difference index corresponding to the server resources based on the reference state parameters.

[0085] This includes a state stability index, which is used to quantify the stability of server resource operation (e.g., to reflect the stability of resource status). Specifically, the state stability index can be represented by the average state value of server resources.

[0086] This includes the state fluctuation index, which is used to quantify the degree of short-term fluctuation of server resources under current state parameters. Specifically, the state fluctuation index can be represented by the state standard deviation of server resources.

[0087] This involves a reference state parameter, which is determined based on a combination of the state stability index and the state fluctuation index. It represents the baseline state parameter of the server resources during the normal operation range at the time point corresponding to the current state parameter, providing a comparison benchmark for subsequent quantification of the state difference index.

[0088] By determining the state stability index (such as the state mean) and state fluctuation index (such as the state standard deviation) of server resources under the current state parameters, the smoothness of resource operation and the intensity of short-term fluctuations can be quantified respectively. By combining these two indices to determine the reference state parameters that characterize the normal operating range of resources at the current time point, a unified and reasonable comparison benchmark can be constructed. Then, based on the reference state parameters, the state difference index can be determined, which can accurately quantify the degree of deviation between the current state parameters and the predicted state parameters.

[0089] As an optional embodiment, determining the target early warning parameter corresponding to the server resource based on the risk status parameter includes: determining multiple reference risk parameters corresponding to the server resource, wherein the multiple reference risk parameters correspond to different risk status levels; comparing the risk status parameter with the multiple reference risk parameters to obtain multiple risk comparison results corresponding to the risk status parameter, wherein the multiple risk comparison results correspond one-to-one with the multiple reference risk parameters; and determining the target early warning parameter corresponding to the server resource based on the multiple risk comparison results corresponding to the risk status parameter.

[0090] This involves multiple reference risk parameters, which are pre-defined parameters used to classify server resources into different risk levels. For example, the reference risk levels include Level 1, Level 2, and Level 3, which are then combined with the anomaly determination results to generate early warning information (i.e., target early warning parameters). Level 1 corresponds to severe anomalies, Level 2 to moderate anomalies, and Level 3 to mild anomalies.

[0091] This involves multiple risk comparison results, which are a set of matching results generated by comparing the current risk status parameters with each reference risk parameter one by one. Each result reflects the degree of deviation between the current status and the corresponding risk level (such as whether it exceeds the threshold, the magnitude of the deviation, etc.), and is used to determine the triggering conditions of the target warning parameters.

[0092] By setting multiple reference risk parameters corresponding to different risk levels, clear risk level boundaries can be defined for server resources. By comparing the risk level parameters with each reference parameter one by one to generate a one-to-one risk comparison result, the deviation of the current state from each risk level can be accurately quantified. Finally, the target warning parameters are determined based on the comparison results, which can ensure that the warning information strictly matches the actual risk level, avoid false alarms or missed alarms caused by fuzzy thresholds, and achieve the accuracy and timeliness of risk warning.

[0093] Based on the above embodiments and optional embodiments, an optional implementation method is provided, which is described in detail below.

[0094] In related technologies, the operational status of server resources directly affects the stability and service continuity of the server or its corresponding system. Fluctuations in load and abnormal resource usage can lead to performance degradation and service interruptions. Therefore, early warning of operational risks to server resources is necessary to identify potential problems in advance and implement timely countermeasures. However, related technologies suffer from inaccurate early warnings of operational risks to server resources.

[0095] There is currently no effective solution to the above problems.

[0096] In view of this, the optional implementation of this application provides a method for early warning of operational risks of server resources, which can effectively solve the above-mentioned technical problems. Figure 3 This is a flowchart of a server resource operation risk warning method in an optional implementation of this application, such as... Figure 3 As shown, a detailed description follows.

[0097] S1, obtain the resource parameters of the server resources, including the current status parameters and historical status parameters;

[0098] Specifically, when server resources include multiple resource items, resource parameters corresponding to each of these items are obtained. These multiple resource items are those whose state characterization index is greater than a state characterization threshold. The state characterization index represents the degree of characterization of the server resource's operational state. For example, multiple resource items include CPU utilization, memory usage, disk read / write speed, and network communication bandwidth. Obtaining the server resource parameters involves: collecting multi-dimensional performance indicator data of the server's operating resources, and performing unified timestamp marking and standardization on the multi-dimensional performance indicator (i.e., multiple resource items) data to form a detection dataset. This detection dataset is then stored to form a historical detection dataset. The multi-dimensional performance indicator data includes CPU utilization, memory usage, disk read / write speed, and network communication bandwidth, resulting in the current detection dataset (i.e., current state parameters) and the historical detection dataset (i.e., historical state parameters).

[0099] S2, Based on historical state parameters, determine the predicted state parameters corresponding to the server resources;

[0100] Taking multiple resource items, including CPU utilization, memory usage, disk read / write speed, and network communication bandwidth, as an example, the corresponding server resources are predicted based on historical detection datasets to obtain the predicted state parameters of the corresponding server resources. Specifically, when server resources include multiple resource items, the historical state parameters of any one of the resource items are decomposed to obtain a first state component, a second state component, and a third state component corresponding to that resource item. For example, taking multiple resource items including CPU utilization, memory usage, disk read / write speed, and network communication bandwidth as an example, the historical detection dataset is subjected to time series decomposition processing to form decomposed performance indicators. The time series decomposition processing includes extracting a trend component (i.e., the first state component), a seasonal component (i.e., the second state component), and a random component (i.e., the third state component). Among them, the trend component is used to reflect the long-term and continuous change trend of the resource status of server resources (e.g., the resource status of the corresponding resource item); the seasonal component is used to reflect the periodic repetitive state fluctuation characteristics of the resource status of server resources (e.g., the resource status of the corresponding resource item); and the random component is used to reflect other fluctuation characteristics of the resource status of server resources (e.g., the resource status of the corresponding resource item) besides the trend component and the seasonal component, such as irregular and sudden fluctuation characteristics.

[0101] Specifically, determining the predicted state parameters corresponding to server resources based on historical state parameters can be achieved through a target model. This target model can be a trained time-series prediction model. The trained time-series prediction model performs trend prediction on multi-dimensional performance index data to generate predicted resource states (i.e., predicted state parameters). The first state component, the second state component, and the third state component are input into the long short-term memory network included in the target model to obtain the time-series state features corresponding to the server resources. For example, the time-series prediction model can be trained based on a sample detection dataset, and the trained time-series prediction model can be used to perform trend prediction on multi-dimensional performance index data to generate predicted resource states. The trained time-series prediction model includes a Long Short-Term Memory (LSTM) network, which is used to extract time-dependent features (i.e., time-series state features) from multi-dimensional performance index data. Specific steps include: performing time-series decomposition on the historical detection dataset to form decomposed performance indices; the time-series decomposition process includes extracting trend components, seasonal components, and random components; inputting the decomposed performance indices (i.e., trend components, seasonal components, and random components) into the LSM network, outputting a time-series feature vector (i.e., time-series state features); and predicting the time-series feature vector to generate predicted resource states.

[0102] Furthermore, before retrieving the target model, the process includes: determining the predicted temporal features corresponding to the sample state components and the sub-predicted features corresponding to each of the multiple sample sub-features, based on multiple sample sub-features and their corresponding first arrangement; determining the feature deviation index between the reference temporal features and the predicted temporal features; determining the update index corresponding to the initial parameters based on the sub-predicted features corresponding to each of the multiple sample sub-features, their corresponding second arrangement, and the feature deviation index, whereby the update index represents the degree of update to the initial parameters; and updating the initial parameters according to the update index to obtain the target parameters. Specifically, this can be achieved using the backpropagation algorithm. For example, a Long Short-Term Memory (LSTM) network model can be trained using historical data (i.e., sample data), and the model parameters (i.e., the target parameters) can be optimized using the backpropagation algorithm to minimize the prediction error of the LSM network model.

[0103] Further, based on the first state component, the second state component, and the third state component, the temporal state characteristics corresponding to the server resource are determined, including: determining the target state component corresponding to the server resource based on the first state component, the second state component, and the third state component; determining the first temporal characteristic corresponding to the server resource based on the first target sub-feature according to the execution order of the multiple target sub-features included in the target state component, and determining the next temporal characteristic corresponding to the server resource based on the first temporal characteristic and the next target sub-feature, until the multiple target sub-features have been executed, thus obtaining the temporal state characteristics corresponding to the server resource. Specifically, this can be achieved through the target model described above.

[0104] S3, determine the state difference index between the current state parameters and the predicted state parameters;

[0105] Further, S3 includes: determining the state stability index and state fluctuation index corresponding to the server resources under the current state parameters; determining the reference state parameters corresponding to the server resources based on the state stability index and the state fluctuation index; and determining the state difference index corresponding to the server resources based on the reference state parameters. Specifically, the state stability index can be the state mean of the server resources, and the state fluctuation index can be the state standard deviation of the server resources. For example, the predicted resource state and the current detection dataset are compared to obtain the dynamic threshold comparison result (i.e., the state difference index). This comparison process can use an adaptive threshold model. The adaptive threshold model is an algorithm model that dynamically adjusts the threshold, mainly used for anomaly detection, classification, or prediction tasks in data. By using the adaptive threshold model for dynamic threshold comparison and combining it with the dynamic threshold residual calculation method, the predicted resource state and the actual detection data are compared and analyzed to accurately assess the resource usage. For the obtained dynamic threshold comparison result, multi-objective comprehensive analysis is performed to obtain the anomaly judgment result (i.e., the risk state parameter), which can be specifically achieved by dynamic threshold residual calculation. When the server resources include at least one resource item, the calculation formula for the dynamic threshold residual is:

[0106]

[0107] in, Let be the residual value (i.e., the state difference index) of the i-th type of resource (i.e., the i-th resource item) at time t, which is dimensionless; Let be the predicted resource state (i.e., the predicted state parameter) of the i-th type of resource at time t, which is dimensionless; Let be the moving average (i.e., the state stability index) of the i-th type of resource at time t, which is dimensionless; Let be the sliding standard deviation (i.e., the state fluctuation index) of the i-th type of resource at time t, which is dimensionless; Let be the prediction deviation of the i-th type of resource at time t, which is dimensionless; The adjustment coefficient is dimensionless. The adjustment coefficient is dimensionless and satisfies... , The reference resource state (i.e., reference state parameter) of the i-th type of resource (i.e., the i-th resource item) at time t.

[0108] S4. Based on the state difference index, determine the risk state parameters corresponding to the server resources;

[0109] Specifically, the risk status parameters are determined based on the residual values ​​obtained from the aforementioned calculation formula of the dynamic threshold residual. These risk status parameters include Level 1 severe anomaly, Level 2 moderate anomaly, and Level 3 mild anomaly.

[0110] S5, based on the risk status parameters, determine the target early warning parameters corresponding to the server resources;

[0111] Furthermore, S5 includes: determining multiple reference risk parameters corresponding to server resources, where each reference risk parameter corresponds to a different level of risk status; comparing the risk status parameter with each of the multiple reference risk parameters to obtain multiple risk comparison results corresponding to the risk status parameter, where each risk comparison result corresponds one-to-one with the multiple reference risk parameters; and determining the target early warning parameter corresponding to the server resources based on the multiple risk comparison results corresponding to the risk status parameter. Specifically, the multiple reference risk parameters can be multiple reference risk levels. For example, the multiple reference risk levels include Level 1, Level 2, and Level 3, which are then combined with the anomaly judgment results to generate early warning information (i.e., target early warning parameters). Level 1 corresponds to severe anomalies, Level 2 corresponds to moderate anomalies, and Level 3 corresponds to mild anomalies. Through multi-target comprehensive analysis, anomaly judgment results are formed, which in turn generate early warning information containing alarm levels and corresponding processing instructions. This information is pushed to the operation and maintenance management terminal via local messages, emails, SMS, etc., and triggers instructions such as automatic expansion, resource scheduling, or increased detection frequency based on different anomaly levels, effectively improving resource management efficiency, ensuring the efficient and stable operation of the server, and reducing potential failure risks.

[0112] S6 provides operational risk warnings for server resources based on target warning parameters.

[0113] Figure 4 This is a schematic diagram of the anomaly determination and handling process in an optional implementation of this application, such as... Figure 4 As shown, the early warning information (i.e., the target early warning parameters) includes the alarm level and corresponding handling instructions, which are then pushed to the operation and maintenance management terminal. The corresponding handling instructions include automatic capacity expansion instructions, resource scheduling instructions, and detection frequency increase instructions. Level 3 minor anomalies trigger automatic capacity expansion instructions, Level 2 moderate anomalies trigger resource scheduling instructions, and Level 1 severe anomalies trigger detection frequency increase instructions. Early warning information is pushed via local messages, email, SMS, and application programming interface (API) calls, with the push method switching based on the anomaly determination result.

[0114] By collecting multi-dimensional performance index data of server operating resources, and performing unified timestamp marking and standardization processing, a detection dataset is formed. Based on this detection dataset, a time-series prediction model is constructed, and long short-term memory networks are used to predict the trends of multi-dimensional performance index data, generating predicted resource status. This enables real-time detection and prediction of server resource status, providing data support for subsequent early warning and resource management. An adaptive threshold model is used for dynamic threshold comparison, combined with a dynamic threshold residual calculation method, to compare and analyze the predicted resource status with the actual detection data, accurately assessing resource usage and achieving effective intelligent detection and early warning. This process achieves accurate resource prediction and dynamic threshold comparison, significantly improving server resource utilization efficiency, reducing manual intervention, enhancing operational response speed, and realizing an automated early warning mechanism and intelligent analysis. It effectively reduces system downtime and ensures system stability and reliability. Furthermore, the above steps have good scalability, adapting to server environments of different sizes, and provide flexible early warning responses through a customized alarm mechanism, ensuring timely handling of various anomalies.

[0115] Figure 5 This is a schematic diagram of the operational risk early warning system structure in an optional implementation of this application, such as... Figure 5 As shown, the server resource operation risk early warning system includes a data acquisition layer, a data processing layer, a storage layer, a prediction model layer, a threshold comparison layer, an analysis and judgment layer, and an early warning execution layer. Specifically, the data acquisition layer collects multi-dimensional indicator data of the server resources; the data processing layer standardizes the collected multi-dimensional indicator data and outputs the processed standardized data; the storage layer stores the processed standardized data, forming a historical dataset for subsequent use; the prediction model layer uses an LSTM model to calculate the predicted state parameters of the server resources based on the historical dataset; the threshold comparison layer compares the current state parameters with the predicted state parameters and outputs dynamic threshold comparison results; the analysis and judgment layer determines the anomaly level of the server resources based on the dynamic threshold comparison results; and the early warning execution layer generates early warning push instructions based on the anomaly level and sends the instructions to the operation and maintenance management terminal.

[0116] The following description, with specific examples, will further illustrate this point.

[0117] A specific example of building a time-series prediction model based on historical monitoring data to accurately predict the future state of server resources, thereby optimizing resource management and early warning mechanisms, is as follows:

[0118] 1) Hypothetical Data: Assume the following four server performance metrics were collected, with a sampling period of once per hour, and the data covers the past 30 days. Table 1 is a sample table of the collected data, as shown in Table 1:

[0119] Table 1

[0120]

[0121] 2) Time Series Decomposition Processing: First, extract the data for each performance metric from the detection dataset, ensuring the data is arranged in chronological order. Next, apply time series decomposition to each performance metric, decomposing it into three components: a trend component, a seasonal component, and a random component. The trend component reflects the long-term trend of data over time; for example, CPU utilization may show a gradual upward or downward trend. The seasonal component captures periodic changes in the data; server load fluctuates significantly between day and night. The random component represents random fluctuations that cannot be explained by the trend and seasonal components. Then, the decomposed trend, seasonal, and random components are stored separately for subsequent model training and analysis.

[0122] 3) Training using a Long Short-Term Memory (LSTM) network: First, the decomposed data is used as input to the LSM network model. To ensure training effectiveness, all data must be standardized to ensure consistency in the dimensions of each indicator. Next, an LSM network model is constructed to process time-series data. This model learns the time dependencies between different performance indicators based on historical data and sets an appropriate time window, using data from the past 24 hours to predict resource status for the next hour. Then, the LSM network model (hereinafter referred to as the model) is trained using historical data, and the model parameters are optimized through backpropagation to minimize prediction error. It is assumed that 50 epochs are performed during training, and cross-validation is used to adjust hyperparameters to improve the model's prediction accuracy. Here, an epoch represents the process of traversing the entire training dataset once. Finally, after training, the model is used to predict the test set data, generating resource status predictions for the next 24 hours. Assume the predicted CPU utilization is as follows: 63% at 00:00 on August 31, 2025; 64% at 01:00 on August 31, 2025; and 66% at 02:00 on August 31, 2025. Compare these predictions with actual data to verify the model's accuracy.

[0123] 4) Time Series Feature Vector Prediction: First, based on a trained Long Short-Term Memory (LSTM) network model, the resource status over a future period is predicted. Each predicted value represents an estimate of the resource status at a specific future moment; for example, the predicted CPU utilization for the next hour is 63%. Next, the model's predictions are compared with actual data. For example, if the actual CPU utilization at 00:00 on August 31, 2025 is 62%, while the predicted value is 63%, the model's prediction error is 1%. A small error indicates good predictive ability. Then, the model's accuracy is analyzed by comparing the predicted and actual values. If the prediction error is within an acceptable range, the model effectively reflects changes in resource status. If the error is large, the model needs adjustment and optimization.

[0124] Through the above steps, a Long Short-Term Memory (LSTM) network model was used to perform time-series predictions of various server resource indicators (i.e., individual resource items). By decomposing the data into time series sequences, the model effectively captured the changing patterns of resource usage, and the prediction results matched the actual data with minimal error. This model provides accurate predictive basis for resource management, enabling timely identification of potential resource bottlenecks or anomalies, thereby improving resource utilization efficiency and system stability.

[0125] Dynamic threshold residual calculation and multi-objective comprehensive analysis are used to accurately determine abnormal server resource states and implement corresponding strategies to ensure system stability. A specific example is as follows:

[0126] Assume a server resource operation risk early warning system (hereinafter referred to as the system) is responsible for monitoring the resources of a server. Based on historical monitoring data, the system has established a time-series prediction model. The system uses dynamic threshold residual calculation to determine anomalies by comparing the difference between the actual resource status and the predicted status. Specifically, the anomaly determination result is formed by multi-objective comprehensive analysis of the dynamic threshold comparison results. The dynamic threshold residual calculation formula is as follows:

[0127]

[0128] in, For the first Class resources at any time The residual value is dimensionless. For the first Class resources at any time Predicted resource status, dimensionless. For the first Class resources at any time The moving average, dimensionless. For the first Class resources at any time The sliding standard deviation, dimensionless. For the first Class resources at any time Prediction bias, dimensionless. The adjustment coefficient is dimensionless. The adjustment coefficient is dimensionless and satisfies... .

[0129] 1) Data Acquisition and Prediction Model: Assuming the system is at time point... At that time, data collection and prediction had been completed, and resource detection data and predicted values ​​had been obtained. Table 2 is a schematic table of resource detection data and predicted values. As shown in Table 2, the predicted values ​​for the corresponding resource types are displayed. Moving average Sliding standard deviation and prediction bias .

[0130] Table 2

[0131]

[0132] 2) Normalization: The detection data for each resource is normalized. The normalization formula is:

[0133]

[0134] in, The detection data is after normalization. The detection data before normalization; The minimum value of the detection data before normalization; the maximum value of the detection data before normalization.

[0135] For example, the maximum value for CPU utilization is 90%, and the minimum value is 2%; the maximum value for memory utilization is 70%, and the minimum value is 1%. The maximum value for network bandwidth is 1000%, and the minimum value is 1%. Table 3 shows the normalization results.

[0136] Table 3

[0137]

[0138] 3) Set the adjustment coefficient: Set the adjustment coefficient =0.6 indicates a high level of concern regarding the moving standard deviation. Set an adjustment factor. This indicates a relatively low level of concern regarding prediction bias.

[0139] 4) Dynamic threshold residual calculation: Using the above formula, calculate the residual for each resource at each time point. residual value .

[0140] CPU resources: Obtain the residual value of CPU resources. This indicates that the state of the resource deviates little from the predicted value and has not reached an abnormal level.

[0141] Memory resources: Obtain the residual value of memory resources. This also indicates that the resource is in a normal state and there are no obvious abnormalities.

[0142] Online resources: Obtain the residual value of network resources Although the deviation is small, it indicates that there is a certain degree of fluctuation, and the system should continue to monitor.

[0143] 5) Comprehensive Analysis and Anomaly Detection: A comprehensive analysis is performed based on the calculated residual values, including CPU residual values. This indicates that the resource status differs little from the predicted value, and the system status is normal. Memory resource residual value The resource status is normal. Network resource residual value. Although the deviation is small, fluctuations in network bandwidth still need to be monitored.

[0144] 6) Anomaly Judgment Criteria: Based on the set anomaly thresholds, the anomaly level is determined, including: Level 1 Severe Anomaly: Residual value exceeds 0.1, indicating an abnormal resource status, which may lead to system crashes or service interruptions. Level 2 Moderate Anomaly: Residual value between 0.05 and 0.1, indicating significant resource pressure, which may affect system performance, requiring scheduling or capacity expansion. Level 3 Mild Anomaly: Residual value less than 0.05, resource status is normal, but further observation can be conducted by increasing the detection frequency.

[0145] In the example above, the residual values ​​for all resources are less than 0.05, indicating that the system is in normal condition and has not reached an abnormal level. However, fluctuations in network resources should continue to be monitored.

[0146] 7) Generate early warning information: Based on the current calculation results, the system has not triggered any serious anomalies. If the residual value of network resources increases further when the system updates data subsequently, an early warning may be triggered.

[0147] By applying dynamic threshold residual calculation and multi-objective comprehensive analysis methods, abnormal conditions of server resources can be accurately identified, ensuring that different resources are compared under the same dimensions. This also helps the system to judge the status of each resource in a timely manner, ensuring the stability and performance of server resources. When resource anomalies occur, the system can automatically trigger early warnings and take appropriate adjustment measures to ensure stable system operation.

[0148] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0149] Embodiments of this application also provide a server resource operation risk early warning device. Figure 6 This is a structural block diagram of a server resource operation risk warning device according to an embodiment of this application, such as... Figure 6 As shown, the device includes: an acquisition module 602, a first determination module 604, a second determination module 606, a third determination module 608, a fourth determination module 610, and a fifth determination module 612. The device will be described in detail below.

[0150] The acquisition module 602 is used to acquire resource parameters of server resources, wherein the resource parameters include current status parameters and historical status parameters; the first determination module 604, connected to the acquisition module 602, is used to determine the predicted status parameters corresponding to the server resources based on the historical status parameters; the second determination module 606, connected to the first determination module 604, is used to determine the status difference index between the current status parameters and the predicted status parameters; the third determination module 608, connected to the second determination module 606, is used to determine the risk status parameters corresponding to the server resources based on the status difference index; the fourth determination module 610, connected to the third determination module 608, is used to determine the target warning parameters corresponding to the server resources based on the risk status parameters; and the fifth determination module 612, connected to the fourth determination module 610, is used to provide an operational risk warning for the server resources based on the target warning parameters.

[0151] It should be noted that the above-mentioned acquisition module 602, first determination module 604, second determination module 606, third determination module 608, fourth determination module 610, and fifth determination module 612 correspond to steps S202 to S212 in the method for early warning of operational risks of server resources. The instances and application scenarios implemented by multiple modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments.

[0152] As can be seen from the above, in the scheme described in the above embodiments of this application, the resource parameters of the server resources can be obtained by the acquisition module. The resource parameters include current state parameters and historical state parameters, which can provide comprehensive data support for subsequent state prediction and risk assessment. The first determination module determines the predicted state parameters corresponding to the server resources based on the historical state parameters to obtain an expected reference benchmark that conforms to the operating characteristics of the server resources. The second determination module determines the state difference index between the current state parameters and the predicted state parameters, which can accurately quantify the deviation between the actual state and the expected state, providing an objective basis for risk assessment. The third determination module determines the risk state parameters corresponding to the server resources based on the state difference index. The fourth determination module determines the target warning parameters corresponding to the server resources based on the risk state parameters, which can ensure that the warning information is accurately matched with the actual risk level. The fifth determination module performs operational risk warning for the server resources based on the target warning parameters, which can realize timely identification and alarm of server resource operation risks. This hierarchical and progressive module collaboration mechanism can form a closed loop in the risk warning process, improving the accuracy and timeliness of the warning.

[0153] Optionally, the first determining module includes: a first determining unit, configured to decompose historical state parameters to obtain a first state component, a second state component, and a third state component corresponding to the server resource, wherein the first state component represents the trend characteristics of the server resource's operating state, the second state component represents the stability characteristics of the server resource's operating state, and the third state component represents the fluctuation characteristics of the server resource's motion state; based on the first state component, the second state component, and the third state component, determine the temporal state characteristics corresponding to the server resource, wherein the temporal state characteristics represent the resource state of the server resource from a time dimension; and based on the temporal state characteristics, determine the predicted state parameters corresponding to the server resource.

[0154] Optionally, the first determining unit includes: a second determining unit, configured to determine a target state component corresponding to the server resource based on a first state component, a second state component, and a third state component; determine a first temporal feature corresponding to the server resource based on the first target sub-feature according to the execution order of multiple target sub-features included in the target state component; and determine a next temporal feature corresponding to the server resource based on the first temporal feature and the next target sub-feature, until multiple target sub-features are executed, thereby obtaining the temporal state feature corresponding to the server resource.

[0155] Optionally, the first determining unit includes: a third determining unit, used to retrieve a target model, wherein the target model includes target parameters, the target parameters are used for feature extraction from the time dimension, the target parameters are obtained by training initial parameters based on sample data, the sample data includes sample state components and reference time-series features corresponding to the sample state components; and the time-series state features corresponding to the server resources are determined based on the first state component, the second state component, the third state component, and the target model.

[0156] Optionally, the third determining unit includes: a fourth determining unit, configured to determine, based on multiple sample sub-features and a first arrangement order corresponding to the multiple sample sub-features, a predicted temporal feature corresponding to the sample state component and sub-predicted features corresponding to the multiple sample sub-features respectively, wherein the sample state component includes multiple sample sub-features; determine a feature deviation index between the reference temporal feature and the predicted temporal feature; determine an update index corresponding to the initial parameter based on the sub-predicted features corresponding to the multiple sample sub-features respectively, a second arrangement order corresponding to the multiple sample sub-features, and the feature deviation index, wherein the update index is used to represent the degree of update of the initial parameter, wherein the second arrangement order and the first arrangement order have a time correlation; and update the initial parameter based on the update index to obtain the target parameter.

[0157] Optionally, the second determining module includes: a fifth determining unit, used to determine the state stability index and state fluctuation index corresponding to the server resources under the current state parameters; determine the reference state parameters corresponding to the server resources based on the state stability index and the state fluctuation index; and determine the state difference index corresponding to the server resources based on the reference state parameters.

[0158] Optionally, the fourth determining module includes: a sixth determining unit, used to determine multiple reference risk parameters corresponding to server resources, wherein the multiple reference risk parameters correspond to different risk status levels; compare the risk status parameters with the multiple reference risk parameters respectively to obtain multiple risk comparison results corresponding to the risk status parameters, wherein the multiple risk comparison results correspond one-to-one with the multiple reference risk parameters; and determine the target early warning parameter corresponding to the server resources based on the multiple risk comparison results corresponding to the risk status parameters.

[0159] Embodiments of this application also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any of the above embodiments of the server resource operation risk warning method.

[0160] Embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above-described embodiments of the server resource operation risk warning method when running.

[0161] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0162] The embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above-described embodiments of the server resource operation risk warning method.

[0163] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application. The above provides a detailed description of a server resource operation risk warning provided by this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A method for early warning of operational risks of server resources, characterized in that, include: Obtain resource parameters of server resources, wherein the resource parameters include current status parameters and historical status parameters; Based on the historical state parameters, determine the predicted state parameters corresponding to the server resources; Determine the state difference index between the current state parameter and the predicted state parameter; Based on the state difference index, determine the risk state parameters corresponding to the server resources; Based on the risk status parameters, determine the target early warning parameters corresponding to the server resources; Based on the target warning parameters, an operational risk warning is issued for the server resources.

2. The method for early warning of operational risks of server resources according to claim 1, characterized in that, The step of determining the predicted state parameters corresponding to the server resources based on the historical state parameters includes: The historical state parameters are decomposed to obtain a first state component, a second state component, and a third state component corresponding to the server resource. The first state component is used to represent the trend characteristics of the server resource's operating state, the second state component is used to represent the stability characteristics of the server resource's operating state, and the third state component is used to represent the fluctuation characteristics of the server resource's motion state. Based on the first state component, the second state component, and the third state component, a temporal state feature corresponding to the server resource is determined, wherein the temporal state feature is used to represent the resource state of the server resource from a time dimension. Based on the time-series state characteristics, predictive state parameters corresponding to the server resources are determined.

3. The method for early warning of operational risks of server resources according to claim 2, characterized in that, The step of determining the temporal state characteristics corresponding to the server resource based on the first state component, the second state component, and the third state component includes: Based on the first state component, the second state component, and the third state component, a target state component corresponding to the server resource is determined. According to the execution order of multiple target sub-features included in the target state component, a first temporal feature corresponding to the server resource is determined based on the first target sub-feature, and a next temporal feature corresponding to the server resource is determined based on the first temporal feature and the next target sub-feature, until the multiple target sub-features are executed, and the temporal state feature corresponding to the server resource is obtained.

4. The method for early warning of operational risks of server resources according to claim 2, characterized in that, The step of determining the temporal state characteristics corresponding to the server resource based on the first state component, the second state component, and the third state component includes: The target model is retrieved, wherein the target model includes target parameters, the target parameters are used for feature extraction from the time dimension, the target parameters are obtained by training initial parameters based on sample data, and the sample data includes sample state components and reference time-series features corresponding to the sample state components; Based on the first state component, the second state component, the third state component, and the target model, the temporal state characteristics corresponding to the server resources are determined.

5. The method for early warning of operational risks of server resources according to claim 4, characterized in that, Before retrieving the target model, the following steps are also included: Based on multiple sample sub-features and a first arrangement order corresponding to the multiple sample sub-features, predictive temporal features corresponding to the sample state component and sub-predictive features corresponding to the multiple sample sub-features are determined, wherein the sample state component includes the multiple sample sub-features; Determine the feature deviation index between the reference time-series features and the predicted time-series features; Based on the sub-predicted features corresponding to the multiple sample sub-features, the second arrangement order corresponding to the multiple sample sub-features, and the feature deviation index, an update index corresponding to the initial parameter is determined. The update index is used to represent the degree of update of the initial parameter. The second arrangement order and the first arrangement order have a time correlation. The initial parameters are updated based on the update index to obtain the target parameters.

6. The method for early warning of operational risks of server resources according to claim 1, characterized in that, Determining the state difference index between the current state parameter and the predicted state parameter includes: Determine the state stability index and state fluctuation index corresponding to the server resources under the current state parameters; Based on the state stability index and the state fluctuation index, determine the reference state parameters corresponding to the server resources; Based on the reference state parameters, determine the state difference index corresponding to the server resources.

7. The method for early warning of operational risks of server resources according to any one of claims 1 to 6, characterized in that, The step of determining the target early warning parameter corresponding to the server resource based on the risk status parameter includes: Multiple reference risk parameters are determined for the server resources, wherein the multiple reference risk parameters correspond to different risk status levels; The risk status parameter is compared with the plurality of reference risk parameters to obtain a plurality of risk comparison results corresponding to the risk status parameter, wherein the plurality of risk comparison results correspond one-to-one with the plurality of reference risk parameters; Based on the risk status parameters and the results of multiple risk comparisons, the target early warning parameters corresponding to the server resources are determined.

8. A server resource operation risk early warning device, characterized in that, include: The acquisition module is used to acquire resource parameters of server resources, wherein the resource parameters include current status parameters and historical status parameters; The first determining module is used to determine the predicted state parameters corresponding to the server resources based on the historical state parameters. The second determining module is used to determine the state difference index between the current state parameter and the predicted state parameter; The third determining module is used to determine the risk status parameters corresponding to the server resources based on the status difference index. The fourth determining module is used to determine the target early warning parameters corresponding to the server resources based on the risk status parameters; The fifth determining module is used to provide operational risk warnings for the server resources based on the target warning parameters.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the server resource operation risk warning method as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, it implements the steps of the server resource operation risk warning method as described in any one of claims 1 to 7.

Images