Private cloud resource demand prediction method and system based on time sequence

Abstract

The invention discloses a private cloud resource demand prediction method and system based on a time sequence, and the method comprises the steps: analyzing a cloud resource configuration log of a private cloud management domain, collecting cloud resource configuration historical data, and carrying out the preprocessing; performing Kalman filtering smoothing processing on the preprocessed cloud resource configuration historical data to obtain change trend feature information of the cloud resource configuration historical data, and screening and extracting the cloud resource configuration historical data meeting the change trend corresponding to the change trend feature information as effective data; utilizing the effective data to train an Informer model, and predicting cloud resource demand information of a private cloud management domain needing to be expanded in a future set time period in a corresponding time period; and optimizing and adjusting the subordinate cloud resources of the private cloud management domain, determining all the cloud resources configured to the private cloud management domain which needs to be expanded, and adjusting the operation of configuring the cloud resources to the private cloud management domain.

Description

Technical Field

[0001] The present invention relates to the technical field of cloud computing resources, and in particular to a method and system for predicting private cloud resource demand based on time series. Background Art

[0002] As cloud service providers continue to expand the scale of cloud computing resources, scientifically and effectively predicting cloud resource demand is crucial for better planning and management of cloud computing resources. Existing methods usually collect cloud resource capacity data from cloud platforms, use it as historical sample data, and use classic time series models (ARIMA) or traditional machine learning models (LSTM) to predict cloud resource demand, but there are still the following problems:

[0003] (1) Existing technical solutions focus on short-sequence prediction. When processing long-sequence data, there are problems such as high computing resource consumption, long training time, and poor prediction effect. In addition, it is difficult to capture complex dynamic patterns, resulting in inaccurate prediction results.

[0004] (2) The performance of existing technical solutions degrades when processing multidimensional data involving multiple related variables. They are unable to fully utilize the correlation between multiple variables and the complex relationship between data, and are unable to accurately and timely provide cloud resource configuration for different users. Summary of the invention

[0005] The purpose of the present invention is to provide a method and system for predicting private cloud resource demand based on time series in order to solve the above problems.

[0006] To achieve the above purpose, the technical solution provided by the present invention is:

[0007] The first aspect of the present application provides a time series-based private cloud resource demand forecast, comprising the following steps:

[0008] Analyze the cloud resource configuration logs of the private cloud management domain, collect and pre-process the cloud resource configuration history data;

[0009] Perform Kalman filter smoothing on the preprocessed cloud resource configuration historical data to obtain the change trend characteristic information of the cloud resource configuration historical data, and filter and extract the cloud resource configuration historical data that meets the corresponding change trend of the change trend characteristic information as valid data;

[0010] The Informer model is trained using the valid data, and the trained Informer model is used to predict the cloud resource demand information of the private cloud management domain that needs to be expanded within a set time period in the future within a corresponding time period;

[0011] Based on the cloud resource demand information, optimize and adjust the cloud resources under the private cloud management domain, and determine all cloud resources to be allocated to the private cloud management domain that needs to be expanded;

[0012] All cloud resources are configured as needed for the private cloud management domain that needs to be expanded, and the operation of configuring cloud resources for the private cloud management domain is adjusted.

[0013] To optimize the above technical solutions, the specific measures taken also include:

[0014] Furthermore, the analysis of the cloud resource configuration log of the private cloud management domain, the collection of cloud resource configuration history data and pre-processing includes:

[0015] Analyze the cloud resource configuration logs of several private cloud management domains, determine all cloud resource configuration operations that occurred in all private cloud management domains within a preset historical time period, perform cloud resource flow tracking processing on all cloud resource configuration operations, and determine the destination terminal identity information of the cloud resource flow corresponding to all cloud resource operations, and use this as the identity information of all management domains that have received the cloud resource configuration operations;

[0016] Based on the identity information of all the management domains, extract the cloud resource configuration records of all the management domains; analyze all the cloud resource configuration records to determine the real-time working status information of the management domain corresponding to each cloud resource configuration record when it is generated;

[0017] Based on the real-time working status information, matching cloud resource configuration historical data are screened from all cloud resource configuration records; the matching cloud resource configuration historical data are sequentially subjected to interference noise data cleaning preprocessing, abnormal data elimination preprocessing and missing data filling preprocessing.

[0018] Furthermore, the real-time working status information includes at least one of the CPU computing load, memory usage ratio and storage space usage ratio of the management domain; the matching cloud resource configuration historical data includes the cloud resource configuration amount and the cloud resource configuration duration.

[0019] Furthermore, the interference noise data cleaning preprocessing is specifically as follows: according to the standard data frame header and data frame tail of the matched cloud resource configuration historical data, the interference noise data between the data frame tail and the data frame header is cleaned and preprocessed preferentially, so that the interference noise data that is easy to clear is cleaned and preprocessed preferentially; according to the interference noise data between the repeated standard data frame headers, cleaning preprocessing is performed, so that the repeated frame header data and the data between the repeated frame header data are removed, so as to avoid the repeated frame header data from being confused in the cloud resource configuration historical data in the form of interference noise data; according to the first standard data frame header and the last standard data frame tail, the head and tail interference noise data are cleaned and preprocessed to complete the overall interference noise data cleaning preprocessing.

[0020] Furthermore, the use of the valid data to train the Informer model, and using the trained Informer model to predict the cloud resource demand information of the private cloud management domain that needs to be expanded within a set time period in the future within a corresponding time period, includes:

[0021] The valid data is input into the encoder part of the Informer model for feature extraction and encoding processing to obtain the corresponding encoded output data; the encoded output data is then input into the decoder part of the Informer model for prediction processing to obtain the corresponding decoded output data; the decoded output data is mapped to the dimension of the target prediction, so as to predict the cloud resource demand information of all management domains under the private cloud management domain that needs to be expanded in the future set time period in the corresponding time period in the future; wherein the cloud resource demand information includes the cloud resource demand and the expected configuration time of cloud resources of all management domains under the private cloud management domain in the corresponding time period in the future.

[0022] Furthermore, the Informer model introduces a multi-scale attention mechanism and a global attention mechanism to capture the correlations and global long-term dependencies of different time scales in time series data.

[0023] Further, the step of configuring all cloud resources in the private cloud management domain that needs to be expanded as needed and adjusting the operation of configuring cloud resources in the private cloud management domain includes:

[0024] Based on the expected configuration time of cloud resources included in the cloud resource demand information, determine the actual cloud resource demand of the corresponding management domain at the corresponding time; based on the difference between the actual cloud resource demand and the cloud resource amount retrieved from the corresponding storage location, judge whether the retrieved cloud resource amount meets the actual cloud resource demand conditions of the corresponding management domain; if so, directly configure the cloud resources retrieved from the corresponding storage location to the corresponding management domain; if not, retrieve idle cloud resources from other storage locations adjacent to the corresponding storage location, and integrate the cloud resources retrieved from the corresponding storage locations and configure them to the corresponding management domain.

[0025] A second aspect of the present application provides a time series-based private cloud resource demand prediction system, comprising:

[0026] The cloud resource configuration data collection and preprocessing module is used to analyze the cloud resource configuration logs of the private cloud management domain, collect cloud resource configuration historical data and perform preprocessing;

[0027] The data extraction module is used to perform Kalman filtering and smoothing processing on the pre-processed cloud resource configuration historical data to obtain the change trend characteristic information of the cloud resource configuration historical data, and screen and extract the cloud resource configuration historical data that meets the corresponding change trend of the change trend characteristic information as valid data;

[0028] A cloud resource demand prediction module is used to train the Informer model using the valid data, and use the trained Informer model to predict the cloud resource demand information of the private cloud management domain that needs to be expanded within a set time period in the future within a corresponding time period;

[0029] A cloud resource screening and identification module is used to optimize and adjust the cloud resources under the private cloud management domain based on the cloud resource demand information, and determine all cloud resources to be configured to the private cloud management domain that needs to be expanded;

[0030] A cloud resource configuration adjustment module is used to configure all cloud resources to the private cloud management domain that needs to be expanded as needed, and adjust the operation of configuring cloud resources to the private cloud management domain;

[0031] The third aspect of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for predicting private cloud resource demand based on time series as described above is implemented.

[0032] A fourth aspect of the present application provides a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the time series-based private cloud resource demand prediction method as described above.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The prediction accuracy of the present invention is higher: the Informer model based on the sparse self-attention mechanism is used, which is significantly improved compared with the traditional time series prediction method in processing long time series and multivariate correlation, and can more accurately predict the long-term change trend of private cloud resource demand, thereby improving the prediction accuracy;

[0035] The prediction performance of the present invention is faster: the Informer model significantly reduces the computational complexity and memory consumption of deep neural networks through sparse attention mechanism and distillation operation, making the prediction task more efficient;

[0036] The present invention introduces Kalman filter smoothing technology in data processing to reduce data mutation and fluctuation, reduce the impact of noise, and provide consistent and coherent state estimation, thereby improving the accuracy and stability of subsequent predictions;

[0037] The present invention can predict multiple cloud resources at the same time, fully considering the correlation between different cloud resources, while improving the comprehensiveness and accuracy of the prediction, helping private cloud customers to flexibly adjust resource demand planning based on the prediction results;

[0038] The present invention is based on predicting cloud resource demand information, determines the cloud resources to be allocated to the private cloud management domain that needs to be expanded according to the proposed rules, and judges whether the retrieved cloud resources meet the actual cloud resource demand conditions of the corresponding management domain based on the difference between the actual cloud resource demand and the cloud resource amount retrieved from the corresponding storage location. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 : Schematic diagram of the flow of the private cloud resource demand prediction system based on time series of the present invention.

[0040] Figure 2 : Schematic diagram of the flow of the private cloud resource demand prediction method based on time series of the present invention.

[0041] Figure 3 : Schematic diagram of the framework of the private cloud resource demand prediction module based on time series of the present invention. DETAILED DESCRIPTION

[0042] The above contents of the present invention are further described in detail below in the form of embodiments, but this should not be understood as the scope of the above subject matter of the present invention being limited to the following embodiments, and all technologies realized based on the above contents of the present invention belong to the scope of the present invention.

[0043] Some of the terms involved in the solution of the present invention are explained as follows:

[0044] Time Series Forecasting: A time series is a sequence of data points arranged in chronological order, such as stock prices, temperature records, sales data, etc. Time series forecasting refers to the process of predicting future values ​​based on historical time series data through methods such as classical time series models or deep learning.

[0045] Private Cloud Resources: Private cloud resources refer to computing resources that are exclusively used by an organization or enterprise, including computing power, storage, network and other infrastructure. These resources are usually deployed in the internal data center of an organization or enterprise or hosted in a dedicated environment of a third-party service provider, providing services specifically for the organization or enterprise to ensure data security, privacy and control.

[0046] In one embodiment of the present invention, a private cloud resource demand prediction based on time series is proposed, and the flow chart is as follows: Figure 2 As shown, the following steps are included:

[0047] Analyze the cloud resource configuration logs of the private cloud management domain, collect and pre-process the cloud resource configuration history data;

[0048] Perform Kalman filter smoothing on the preprocessed cloud resource configuration historical data to obtain the change trend characteristic information of the cloud resource configuration historical data, and filter and extract the cloud resource configuration historical data that meets the corresponding change trend of the change trend characteristic information as valid data;

[0049] Use valid data to train the Informer model, and use the trained Informer model to predict the cloud resource demand information of the private cloud management domain that needs to be expanded within a set time period in the future.

[0050] Based on the cloud resource demand information, optimize and adjust the cloud resources under the private cloud management domain, and determine all cloud resources to be allocated to the private cloud management domain that needs to be expanded;

[0051] All cloud resources are configured as needed for the private cloud management domain that needs to be expanded, and the operation of configuring cloud resources for the private cloud management domain is adjusted.

[0052] The method of the present invention first analyzes the cloud resource configuration logs of several private cloud management domains to collect cloud resource configuration history data and pre-process the data; performs Kalman filter smoothing processing on the pre-processed cloud resource configuration history data to obtain change trend feature information of the cloud resource configuration history data; based on the change trend feature information, all cloud resource configuration history data that meet the corresponding change trend are screened and extracted from the pre-processed cloud resource configuration history data, and used as valid data for model use; based on the valid data, an informer model is trained, and based on the trained informer model, a private cloud management domain that needs to be expanded within a set time period in the future is predicted. The cloud resource demand information in the corresponding time period is input into the encoder part of the Informer model after preprocessing and Kalman filter smoothing technology. The encoder extracts features and encodes the input data through multiple encoder layers. The output of the encoder is passed to the decoder part, and the decoder makes predictions based on the output of the encoder and historical observations. The generator part maps the output of the decoder to the dimension of the target prediction to obtain the prediction result of the private cloud resource demand for a certain period of time in the future. Based on the cloud resource demand information, all cloud resources under the private cloud management domain are optimized and adjusted, and all cloud resources to be configured to the private cloud management domain are determined. The operation status of configuring cloud resources to the private cloud management domain is adjusted. Based on the predicted cloud resource demand information, all cloud resources under the private cloud management domain are optimized and adjusted, and all cloud resources to be configured to the private cloud management domain that needs to be expanded are determined, thereby adjusting the operation status of configuring cloud resources to the private cloud management domain.

[0053] The time series-based private cloud resource demand prediction method analyzes the cloud resource configuration log of the private cloud management domain, determines all the management domains in the private cloud management domain that have received cloud resource configuration operations, collects historical cloud resource configuration data corresponding to the management domain in different working states, preprocesses and extracts the historical cloud resource configuration data, obtains corresponding valid data, and uses this as training data for the Informer model, thereby predicting the cloud resource demand information of all management domains under the private cloud management domain in the corresponding time period in the future, and makes a relatively accurate and reliable prediction of the cloud resource demand of all management domains, providing a reliable basis for the subsequent retrieval of cloud resources; based on the cloud resource demand information, all cloud resources under the private cloud management domain are optimized and adjusted, all cloud resources configured to the private cloud management domain are determined, and combined with the current cloud resource configuration of the private cloud management domain, the operation status of configuring cloud resources to the private cloud management domain is adjusted, and appropriate and accurate suggestions are put forward for expanding the cloud resource capacity of the private cloud management domain, so as to meet the needs of cloud resource configuration business, improve the configuration and utilization efficiency of cloud resources, ensure the configuration reliability of cloud resources, and reduce the idle efficiency of cloud resources.

[0054] The technical solution of the present invention is further described in detail below in conjunction with specific implementation methods:

[0055] In some implementations, analyzing the cloud resource configuration logs of the private cloud management domain, collecting cloud resource configuration history data and preprocessing it includes:

[0056] Analyze the cloud resource configuration logs of several private cloud management domains, determine all cloud resource configuration operations that occurred in all private cloud management domains within a preset historical time period, perform cloud resource flow tracking processing on all cloud resource configuration operations, and determine the destination terminal identity information of the cloud resource flow corresponding to all cloud resource operations, and use this as the identity information of all management domains that have received the cloud resource configuration operations;

[0057] Based on the identity information of all the management domains, extract the cloud resource configuration records of all the management domains; analyze all the cloud resource configuration records to determine the real-time working status information of the management domain corresponding to each cloud resource configuration record when it is generated;

[0058] Based on the real-time working status information, matching cloud resource configuration historical data are screened from all cloud resource configuration records; the matching cloud resource configuration historical data are sequentially subjected to interference noise data cleaning preprocessing, abnormal data elimination preprocessing and missing data filling preprocessing.

[0059] The real-time working status information includes at least one of the CPU computing load, memory usage ratio and storage space usage ratio of the management domain; the matched cloud resource configuration history data includes the cloud resource configuration amount and the cloud resource configuration duration.

[0060] The private management domain contains a large number of cloud resources, which can be but not limited to cloud computing devices. When the management domain performs complex computing tasks, it needs to use the cloud resources under the private management domain to undertake the corresponding computing tasks. In order to ensure that accurate and sufficient cloud resource calculations can be performed for different management domains, it is necessary to make multi-dimensional predictions on the cloud resource requirements of different management domains under different working conditions. Accordingly, the cloud resource configuration log of the private cloud management domain is analyzed to determine all cloud resource configuration operations that occurred in the private cloud management domain within a preset historical time period; cloud resource flow tracking processing is performed on all cloud resource configuration operations to determine the destination terminal identity information of the cloud resource flow corresponding to all cloud resource operations, which is used as the identity information of all management domains that have received the cloud resource configuration operation, and then based on the identity information, the cloud resource configuration records of all management domains are extracted, so as to determine the real-time working status information of the management domain corresponding to each cloud resource configuration record when it is generated, and based on the real-time working status information, the matching cloud resource configuration historical data is screened from all cloud resource configuration records, so that the screened cloud resource configuration historical data can reflect the cloud resource configuration status of the management domain at a multi-dimensional level. The matching cloud resource configuration historical data is then subjected to interference noise data cleaning preprocessing, abnormal data elimination preprocessing and missing data filling preprocessing in sequence, which can improve the accuracy and credibility of the cloud resource configuration historical data.

[0061] In some embodiments, the interference noise data cleaning preprocessing is specifically as follows: based on the standard data frame header and data frame tail of the matched cloud resource configuration historical data, the interference noise data between the data frame tail and the data frame header is cleaned and preprocessed preferentially, so that the interference noise data that is easy to clear is cleaned and preprocessed preferentially; based on the interference noise data between repeated standard data frame headers, cleaning preprocessing is performed, so that the repeated frame header data and the data between the repeated frame header data are removed, so as to avoid the repeated frame header data from being confused in the cloud resource configuration historical data in the form of interference noise data; based on the first standard data frame header and the last standard data frame tail, the head and tail interference noise data are cleaned and preprocessed to complete the overall interference noise data cleaning preprocessing.

[0062] The following formula (1) is used to clean the interference noise data between the data frame tail and the data frame head according to the standard data frame head and data frame tail of the matched cloud resource configuration historical data:

[0063] Q1={{K_[E(a)]+1}→{C K _ [E(a)] (A)-1}}&&[1≤a≤m] (1)

[0064] In the above formula (1), Q1 represents the number of bits of the first interference noise data for cleaning preprocessing; K_[E(a)] represents the number of bits of the last data at the end of the a-th standard data frame in all matching cloud resource configuration history data; C K_[E(a)] (A) represents the number of bits of the first data of the standard data frame header that appears for the first time starting from the K_[E(a)]th bit in all the matching cloud resource configuration history data; m represents the total number of standard data frame tails detected; {{K_[E(a)]+1}→{C K _ [E(a)] (A)-1}}&&[1≤a≤m] means that under the condition of 1≤a≤m, all K_[E(a)]+1th to Cth K _[E(a)] (A) -1-bit number is combined into a set.

[0065] The data corresponding to the bit set is cleared, thus completing the first interference noise data cleaning preprocessing.

[0066] Step S2, using the following formula (2), performs cleaning preprocessing based on the interference noise data between repeated standard data frame headers:

[0067]

[0068] In the above formula (2), Q2 represents the set of bits for cleaning preprocessing of the second interference noise data; C_[A(b)] represents the number of bits of the first data of the b-th standard data frame header in all matching cloud resource configuration historical data; C_[A(b+1)] represents the number of bits of the first data of the b+1-th standard data frame header in all matching cloud resource configuration historical data; n represents the total number of detected standard data frame headers; It means that if there is a standard data frame tail E in the C_[A(b)]th to C_[A(b+1)]-1th bits in all matching cloud resource configuration historical data, the corresponding bit set is an empty set; if there is no standard data frame tail E in the C_[A(b)]th to C_[A(b+1)]-1th bits in all matching cloud resource configuration historical data, the corresponding bit set is the bit set of the C_[A(b)]th to C_[A(b+1)]-1th bits combined with &&[1≤b≤n], that is, the union of all corresponding bit sets under the condition of 1≤b≤n is taken.

[0069] Step S3, using the following formula (3), based on the first standard data frame header and the last standard data frame tail, the interference noise data at the beginning and the end are cleaned and preprocessed:

[0070] Q3={1→{C_[A(1)]-1}}&&{{K_[E(m)]+1}→R} (3)

[0071] In the above formula (3), Q3 represents the set of digits for the third cleaning preprocessing of the interference noise data; C_[A(1)] represents the number of digits of the first data of the first standard data frame header in all matching cloud resource configuration historical data; K_[E(m)] represents the number of digits of the last data of the last standard data frame tail in all matching cloud resource configuration historical data; R represents the total number of digits of all matching cloud resource configuration historical data; {1→{C_[A(1)]-1}}&&{{K_[E(m)]+1}→R} represents the set of digits consisting of the digits from the 1st to the C_[A(1)]-1th and the digits from the K_[E(m)]+1th to the Rth. Clear the data corresponding to the set of digits Q3, and the last cleaning preprocessing of the interference noise data is completed.

[0072] In some implementations, the informer model is trained using valid data, and the trained informer model is used to predict the cloud resource demand information of the private cloud management domain that needs to be expanded within a set time period in the future within a corresponding time period, including:

[0073] The valid data is input into the encoder part of the Informer model for feature extraction and encoding processing to obtain the corresponding encoded output data; the encoded output data is then input into the decoder part of the Informer model for prediction processing to obtain the corresponding decoded output data; the decoded output data is mapped to the dimension of the target prediction, so as to predict the cloud resource demand information of all management domains under the private cloud management domain that needs to be expanded in the future set time period in the corresponding time period in the future; wherein the cloud resource demand information includes the cloud resource demand and the expected configuration time of cloud resources of all management domains under the private cloud management domain in the corresponding time period in the future.

[0074] The cloud resource configuration history data that has completed the preprocessing is subjected to Kalman filtering and smoothing processing to obtain the change trend feature information of the cloud resource configuration history data that has completed the preprocessing, and all cloud resource configuration history data that meet the corresponding change trend are screened and extracted from the cloud resource configuration history data that has completed the preprocessing, so as to ensure that the cloud resource configuration history data that is finally screened truly matches the cloud resource demand of the management domain. In addition, the Kalman filtering and smoothing processing technology and the informer model based on generative AI can more accurately predict the long-term change trend of the private cloud resource demand and improve the prediction accuracy. Then the valid data is input into the encoder part of the informer model for feature extraction and encoding processing to obtain the corresponding encoded output data; then the encoded output data is input into the decoder part of the informer model for prediction processing to obtain the corresponding decoded output data; the decoded output data is mapped to the target prediction dimension, so as to predict the cloud resource demand information of all management domains under the private cloud management domain in the corresponding time period in the future.

[0075] In some embodiments, a multi-scale attention mechanism and a global attention mechanism are introduced into the Informer model to capture the correlations and global long-term dependencies of different time scales in time series data.

[0076] By introducing multi-scale attention mechanism and global attention mechanism into the Informer model, it can effectively capture the correlation of different time scales and global long-term dependencies in time series data, and improve the accuracy and stability of prediction; and by processing long-sequence data, according to the prediction of cloud resource demand in the next year, resource demand planning can be carried out early, helping private cloud users to flexibly adjust resource procurement strategies based on the prediction results, saving money while ensuring service quality. At the same time, it can also fully consider the complexity of private cloud resource demand under the premise of multiple variables such as CPU computing load, memory usage ratio and storage space usage ratio, and improve the comprehensiveness and accuracy of prediction; and adopt an automated way to generate predictions and suggestions, reducing the need for manual intervention and improving efficiency and accuracy.

[0077] In some implementations, all cloud resources in a private cloud management domain that needs to be expanded are configured as needed, and the operation of configuring cloud resources in the private cloud management domain is adjusted, and all cloud resources in a private cloud management domain that needs to be expanded are configured as needed, and the operation of configuring cloud resources in the private cloud management domain is adjusted, including:

[0078] Based on the expected configuration time of cloud resources included in the cloud resource demand information, determine the actual cloud resource demand of the corresponding management domain at the corresponding time; based on the difference between the actual cloud resource demand and the cloud resource amount retrieved from the corresponding storage location, judge whether the retrieved cloud resource amount meets the actual cloud resource demand conditions of the corresponding management domain; if so, directly configure the cloud resources retrieved from the corresponding storage location to the corresponding management domain; if not, retrieve idle cloud resources from other storage locations adjacent to the corresponding storage location, and integrate the cloud resources retrieved from the corresponding storage locations and configure them to the corresponding management domain.

[0079] Based on the cloud resource demand contained in the cloud resource demand information, all cloud resources under the private cloud management domain are screened and identified, and the storage location of the cloud resources configured to the corresponding management domain within the private cloud management domain is determined, so that all cloud resources under the private cloud management domain can be locked in a timely and accurate manner, which is convenient for the subsequent rapid configuration of the corresponding cloud resources for the corresponding management domain. Also based on the expected configuration time of the cloud resources contained in the cloud resource demand information, the actual cloud resource demand of the corresponding management domain at the corresponding time is determined; based on the difference between the actual cloud resource demand and the cloud resource amount retrieved from the corresponding storage location, it is judged whether the retrieved cloud resource amount meets the actual cloud resource demand condition of the corresponding management domain. When the actual cloud resource demand is less than or equal to the cloud resource amount retrieved from the corresponding storage location, it is judged that the retrieved cloud resource amount meets the actual cloud resource demand condition of the corresponding management domain; otherwise, it is judged that the retrieved cloud resource amount does not meet the actual cloud resource demand condition of the corresponding management domain, which is convenient for the subsequent targeted adjustment of the operation state of configuring cloud resources to the corresponding management domain, and ensures that the corresponding management domain can obtain sufficient cloud resource configuration. By comparing and analyzing the predicted results with the pre-set resource utilization threshold, appropriate scaling or expansion suggestions can be provided early to help users make reasonable resource allocation decisions.

[0080] In another embodiment, the present invention proposes a time series-based private cloud resource demand forecasting system, comprising:

[0081] The cloud resource configuration data collection and preprocessing module is used to analyze the cloud resource configuration logs of the private cloud management domain, collect cloud resource configuration historical data and perform preprocessing;

[0082] The data extraction module is used to perform Kalman filtering and smoothing processing on the pre-processed cloud resource configuration historical data to obtain the change trend characteristic information of the cloud resource configuration historical data, and screen and extract the cloud resource configuration historical data that meets the corresponding change trend of the change trend characteristic information as valid data;

[0083] The cloud resource demand prediction module is used to train the Informer model with valid data, and use the trained Informer model to predict the cloud resource demand information of the private cloud management domain that needs to be expanded within a set time period in the future.

[0084] A cloud resource screening and identification module is used to optimize and adjust the cloud resources under the private cloud management domain based on the cloud resource demand information, and determine all cloud resources to be allocated to the private cloud management domain that needs to be expanded;

[0085] A cloud resource configuration adjustment module is used to configure all cloud resources to the private cloud management domain that needs to be expanded as needed, and adjust the operation of configuring cloud resources to the private cloud management domain;

[0086] In another embodiment, the present invention proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the time series-based private cloud resource demand prediction method as described above is implemented.

[0087] In another embodiment of the present invention, a computer-readable storage medium is provided, which stores a computer program, wherein the computer program enables a computer to execute the private cloud resource demand prediction method based on time series as described above.

[0088] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the above. More specific examples of computer storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0089] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any technician familiar with the profession, without departing from the scope of the technical solution of the present invention, according to the technical essence of the present invention, any simple modification, equivalent replacement and improvement made to the above embodiment still falls within the protection scope of the technical solution of the present invention.

Claims

1. A time series-based private cloud resource demand forecast, characterized in that: The following steps are involved: Analyze the cloud resource configuration logs of the private cloud management domain, collect and pre-process the cloud resource configuration history data; Perform Kalman filter smoothing on the preprocessed cloud resource configuration historical data to obtain the change trend characteristic information of the cloud resource configuration historical data, and filter and extract the cloud resource configuration historical data that meets the corresponding change trend of the change trend characteristic information as valid data; The Informer model is trained using the valid data, and the trained Informer model is used to predict the cloud resource demand information of the private cloud management domain that needs to be expanded within a set time period in the future within a corresponding time period; Based on the cloud resource demand information, optimize and adjust the cloud resources under the private cloud management domain, and determine all cloud resources to be allocated to the private cloud management domain that needs to be expanded; All cloud resources are configured as needed for the private cloud management domain that needs to be expanded, and the operation of configuring cloud resources for the private cloud management domain is adjusted.

2. The time series-based private cloud resource demand forecasting method according to claim 1, characterized in that: The analysis of the cloud resource configuration log of the private cloud management domain, the collection of cloud resource configuration history data and pre-processing includes: Analyze the cloud resource configuration logs of several private cloud management domains, determine all cloud resource configuration operations that occurred in all private cloud management domains within a preset historical time period, perform cloud resource flow tracking processing on all cloud resource configuration operations, and determine the destination terminal identity information of the cloud resource flow corresponding to all cloud resource operations, and use this as the identity information of all management domains that have received the cloud resource configuration operations; Based on the identity information of all the management domains, extract the cloud resource configuration records of all the management domains; analyze all the cloud resource configuration records to determine the real-time working status information of the management domain corresponding to each cloud resource configuration record when it is generated; Based on the real-time working status information, matching cloud resource configuration historical data are screened from all cloud resource configuration records; the matching cloud resource configuration historical data are sequentially subjected to interference noise data cleaning preprocessing, abnormal data elimination preprocessing and missing data filling preprocessing.

3. The time series-based private cloud resource demand forecasting method according to claim 2 is characterized in that: The real-time working status information includes at least one of the CPU computing load, memory usage ratio and storage space usage ratio of the management domain; the matching cloud resource configuration history data includes the cloud resource configuration amount and the cloud resource configuration duration.

4. The time series-based private cloud resource demand forecasting method according to claim 2, characterized in that: The interference noise data cleaning preprocessing is specifically as follows: according to the standard data frame header and data frame tail of the matched cloud resource configuration historical data, the interference noise data between the data frame tail and the data frame header is cleaned and preprocessed first, so that the interference noise data that is easy to clear is cleaned and preprocessed first; According to the interference noise data between repeated standard data frame headers, cleaning preprocessing is performed to remove the repeated frame header data and the data between the repeated frame header data, so as to avoid the repeated frame header data being confused in the cloud resource configuration history data in the form of interference noise data; According to the first standard data frame header and the last standard data frame tail, the head and tail interference noise data are cleaned and preprocessed to complete the overall interference noise data cleaning preprocessing.

5. The time series-based private cloud resource demand forecasting method according to claim 1, characterized in that: The using of the valid data to train the Informer model, and using the trained Informer model to predict the cloud resource demand information of the private cloud management domain that needs to be expanded within a set time period in the future within a corresponding time period, includes: The valid data is input into the encoder part of the Informer model for feature extraction and encoding processing to obtain the corresponding encoded output data; the encoded output data is then input into the decoder part of the Informer model for prediction processing to obtain the corresponding decoded output data; the decoded output data is mapped to the dimension of the target prediction, so as to predict the cloud resource demand information of all management domains under the private cloud management domain that needs to be expanded in the future set time period in the corresponding time period in the future; wherein the cloud resource demand information includes the cloud resource demand and the expected configuration time of cloud resources of all management domains under the private cloud management domain in the corresponding time period in the future.

6. The time series-based private cloud resource demand forecasting method according to claim 5, characterized in that: The Informer model introduces multi-scale attention mechanism and global attention mechanism to capture the correlation and global long-term dependency of different time scales in time series data.

7. The time series-based private cloud resource demand forecasting method according to claim 1, characterized in that: The step of configuring all cloud resources in the private cloud management domain that needs to be expanded as needed and adjusting the operation of configuring cloud resources in the private cloud management domain includes: Based on the expected configuration time of cloud resources included in the cloud resource demand information, determine the actual cloud resource demand of the corresponding management domain at the corresponding time; based on the difference between the actual cloud resource demand and the cloud resource amount retrieved from the corresponding storage location, judge whether the retrieved cloud resource amount meets the actual cloud resource demand conditions of the corresponding management domain; if so, directly configure the cloud resources retrieved from the corresponding storage location to the corresponding management domain; if not, retrieve idle cloud resources from other storage locations adjacent to the corresponding storage location, and integrate the cloud resources retrieved from the corresponding storage locations and configure them to the corresponding management domain.

8. A private cloud resource demand prediction system based on time series, characterized in that: include: The cloud resource configuration data collection and preprocessing module is used to analyze the cloud resource configuration logs of the private cloud management domain, collect cloud resource configuration historical data and perform preprocessing; The data extraction module is used to perform Kalman filtering and smoothing processing on the pre-processed cloud resource configuration historical data to obtain the change trend characteristic information of the cloud resource configuration historical data, and screen and extract the cloud resource configuration historical data that meets the corresponding change trend of the change trend characteristic information as valid data; A cloud resource demand prediction module is used to train the Informer model using the valid data, and use the trained Informer model to predict the cloud resource demand information of the private cloud management domain that needs to be expanded within a set time period in the future within a corresponding time period; A cloud resource screening and identification module is used to optimize and adjust the cloud resources under the private cloud management domain based on the cloud resource demand information, and determine all cloud resources to be configured to the private cloud management domain that needs to be expanded; A cloud resource configuration adjustment module is used to configure all cloud resources to the private cloud management domain that needs to be expanded as needed, and adjust the operation of configuring cloud resources to the private cloud management domain; The system is used to execute the private cloud resource demand prediction method based on time series as described in any one of claims 1 to 7.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for predicting private cloud resource demand based on time series as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the time series-based private cloud resource demand prediction method according to any one of claims 1 to 7.

Images