WO2021051529A1

WO2021051529A1 - Method, apparatus and device for estimating cloud host resources, and storage medium

Info

Publication number: WO2021051529A1
Application number: PCT/CN2019/117048
Authority: WO
Inventors: 徐锐杰
Original assignee: 平安科技（深圳）有限公司
Priority date: 2019-09-19
Filing date: 2019-11-11
Publication date: 2021-03-25
Also published as: CN110806954B; CN110806954A

Abstract

The present application relates to the field of cloud services, and provides a method, apparatus and device for estimating cloud host resources, and a storage medium. The method comprises: analyzing target item information, and used resource types and resource usage situations of a plurality of cloud hosts by means of a resource monitoring model obtained through training, so as to acquire a cloud host to be subjected to estimation and a resource to be estimated of said cloud host; acquiring first sampling data and second sampling data of said resource; acquiring prediction information according to the first sampling data and the second sampling data; acquiring estimated information according to the first sampling data, the second sampling data and the prediction information; and generating, according to the estimated information, an optimization configuration policy corresponding to the estimated information. By means of the solution, the resource utilization rate of a cloud host can be increased.

Description

Method, device, equipment and storage medium for evaluating cloud host resources

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on September 19, 2019, the application number is 201910885428.1, and the invention title is "Methods, Apparatus, Equipment, and Storage Medium for Evaluating Cloud Host Resources". The entire content of the application is approved. The reference is incorporated in the application.

Technical field

This application relates to the field of cloud monitoring, in particular to methods, devices, equipment, and storage media for evaluating cloud host resources.

Background technique

As a new generation of hosting services, cloud hosting integrates high-performance servers and high-quality network bandwidth, effectively solving the shortcomings of traditional hosting rental prices and uneven service quality, and can fully meet the needs of small and medium-sized enterprises and individual webmaster users. Host rental services require low cost, high reliability, and easy management. Therefore, cloud hosting is widely used. However, with the development of time, personnel changes, or the shelving of projects, etc., the unreasonable use and waste of cloud host resources, such as the cloud host to which the resigned personnel belong, may not be involved in the resignation process due to possible project associations. It has not been released, resulting in many servers with extremely low resource usage but equipped with high-configuration cloud hosts, which in turn causes a waste of cloud host resources.

In the current cloud monitoring, by obtaining the current resource usage information in the cloud host, the resource usage information is analyzed and evaluated to obtain the evaluation information, the usage status of the cloud host is determined according to the evaluation information, and the usage status is determined according to the usage status Optimize the configuration plan.

Since the resource usage of the cloud host will increase with the increase in the number of users and usage time, the number of users and usage time will affect the resource usage of the cloud host. The inventor realizes that in this case, if you continue to use the cloud host The current resource usage analysis and evaluation plan will result in the single analysis and evaluation of resource usage information, resulting in the determined optimization configuration strategy not being rigorous and comprehensive, resulting in low resource utilization of the cloud host.

Summary of the invention

This application provides a method, device, equipment, and storage medium for evaluating cloud host resources, which can solve the problem of low resource utilization of cloud hosts.

In a first aspect, this application provides a method for evaluating cloud host resources, including: obtaining training data, inputting the training data to a neural network model, and training the neural network model to obtain a resource monitoring model, wherein , The training data includes resource sampling data of multiple cloud hosts for multiple periods; acquiring the input target project information, and analyzing the target project information to obtain the project data type and project operation of the target project information Demand information; obtain the resource type and resource usage of the cloud host, and determine the cloud host to be evaluated and the resource usage condition according to the resource usage type, the resource usage condition, the project data type, and the project operation requirement information The resource to be evaluated of the cloud host to be evaluated; obtaining the first sampling data of the resource to be evaluated in the first time period through the resource monitoring model, and obtaining the second sampling data of the resource to be evaluated in the second time period, Wherein, the start time of the first time period is later than the end time of the second time period; the cloud host to be evaluated is predicted according to the first sampling data and the second sampling data to obtain the first A prediction data, a second prediction data, and a third prediction data, and the third prediction data is used as main information, and the first prediction data and the second prediction data are used as auxiliary information to obtain prediction information, Wherein, the prediction includes a prediction of the resource usage status and the trend of each resource usage of the cloud host to be evaluated in a third period, and the start time of the third period is later than the end of the first period Time; feature extraction of the first sampling data, the second sampling data, and the prediction information to obtain the first key information, the second key information, and the third key information, respectively, and to obtain the first The first alternative resource of the resource to be evaluated in the sampling data, and the second alternative resource of the resource to be evaluated in the second sampling data is acquired; according to the first key information, the second key information, and the first Three key information, the first alternative resource and the second alternative resource evaluate the cloud host to be evaluated to obtain evaluation information; generate an optimized configuration strategy corresponding to the evaluation information according to the evaluation information , And output the evaluation information and the optimal configuration strategy.

In a second aspect, the present application provides an apparatus for evaluating cloud host resources, including: an input and output module for obtaining training data for obtaining input target item information; and a processing module for converting the input and output module The acquired training data is input to a neural network model, and the neural network model is trained to obtain a resource monitoring model, wherein the training data includes resource sampling data of multiple cloud hosts for multiple periods; and The input target project information obtained by the input and output module is analyzed to obtain the project data type and project operation requirement information of the target project information; the used resource type and the used resource situation of the cloud host are obtained, and according to the used resource type, The resource usage, the project data type, and the project operation requirement information are used to determine the cloud host to be assessed and the resource to be assessed of the cloud host to be assessed; and the resource to be assessed is acquired through the resource monitoring model. The first sampling data in a period of time, and the second sampling data of acquiring the resource to be evaluated in a second period of time, wherein the start time of the first period is later than the end time of the second period; according to The first sampling data and the second sampling data predict the cloud host to be evaluated to obtain first prediction data, second prediction data, and third prediction data, respectively, and use the third prediction data as Main information, and use the first prediction data and the second prediction data as auxiliary information to obtain prediction information, where the prediction includes the use of each resource status of the cloud host to be evaluated in the third period and For the prediction of the trend of each resource usage, the start time of the third period is later than the end time of the first period; the first sampling data, the second sampling data, and the prediction information are respectively performed Feature extraction to obtain the first key information, the second key information, and the third key information respectively, and to obtain the first alternative resource of the resource to be evaluated in the first sampling data, and to obtain the resource to be evaluated in the second sampling data. Evaluate the second alternative resource of the resource; according to the first key information, the second key information, the third key information, the first alternative resource and the second alternative resource Evaluate the cloud host for evaluation to obtain evaluation information; generate an optimized configuration strategy corresponding to the evaluation information according to the evaluation information, send the evaluation information and the optimized configuration strategy to the display module, and output through the display module The evaluation information and the optimal configuration strategy; the display module is used to display the evaluation information and the optimal configuration strategy.

In a third aspect, the present application provides a computer device, which includes at least one connected processor, a memory, a display, and an input and output unit, wherein the memory is used to store program code, and the processor is used to call the memory To execute the method described in the first aspect above.

In a fourth aspect, the present application provides a computer-readable storage medium having computer instructions stored in the computer-readable storage medium. When the computer instructions run on a computer, the computer executes the above-mentioned first aspect. method.

In the technical solution provided by this application, the resource monitoring model obtained through training analyzes the target project information, the types of resources used and the resource usage conditions of multiple cloud hosts to obtain the cloud host to be evaluated and the cloud host to be evaluated. Evaluate resources, obtain first sampling data and second sampling data of the resource to be evaluated, obtain prediction information according to the first sampling data and the second sampling data, and obtain prediction information according to the first sampling data and the first sampling data. Second, the sampling data and the prediction information obtain evaluation evaluation information, and based on the evaluation information, an optimized configuration strategy corresponding to the evaluation information is generated. Since the cloud host to be evaluated and the resources to be evaluated of the cloud host to be evaluated are determined based on the target project information, so that the evaluation requirements for the target project information can be met on the basis of the resource usage evaluation of the cloud host, Thereby improving the resource utilization of the cloud host; by evaluating the first sampling data, second sampling data and forecast information of the cloud host, the evaluated data information is multi-faceted, so that the evaluation results are more rigorous and accurate. Obtain forecast data from multiple angles to improve the accuracy of the assessment; in summary, on the one hand, obtain data on resource usage from multiple angles, so that users can quickly and comprehensively learn about the resource usage of the cloud host to be evaluated, on the other hand, By evaluating data from multiple angles to improve the rigor and accuracy of the evaluation, this application can improve the resource utilization of the cloud host.

Description of the drawings

FIG. 1 is a schematic flowchart of a method for evaluating cloud host resources in an embodiment of this application;

2 is a schematic structural diagram of an apparatus for evaluating cloud host resources in an embodiment of the application;

FIG. 3 is a schematic structural diagram of a computer device in an embodiment of the application.

detailed description

This application provides a method, device, device, and storage medium for evaluating cloud host resources, which can be used for the configuration and use of cloud hosts, for the release of cloud hosts that do not use resources, and for those who need to perform configuration reduction operations due to minimal resource usage. Cloud hosting provides reference. Among them, for cloud computing platform vendors, it can be used to provide quantitative standards for resource recovery to provide reference support for reducing the waste of internal computing resources; for cloud computing platform tenants, it can be used to provide information on the resource usage and rational use of business cloud hosts reference.

In order to solve the above technical problems, this application mainly provides the following technical solutions:

Please refer to FIG. 1, the following is an example of a method for evaluating cloud host resources provided by the present application. The method is executed by a computer device. The computer device may be a server or a terminal. When the device 20 shown in FIG. 2 is an application or an executing program , The terminal is a terminal on which the device 20 shown in FIG. 2 is installed. This application does not limit the type of execution subject. The method includes:

101. Obtain training data, input the training data into a neural network model, and train the neural network model to obtain a resource monitoring model.

Wherein, the training data includes resource sampling data of multiple cloud hosts for multiple periods; the training data is data of multiple sampling points that are sampled at a preset sampling interval and arranged in a forward time sequence, including the first data in the first period And the second data in the second time period, the start time of the first time period is later than the end time of the second time period. In this embodiment, the existing monitoring model can be trained through a migration learning algorithm to obtain a resource monitoring model for monitoring, evaluating, and optimizing the configuration of the resources used by the cloud host. The resource monitoring model in this embodiment is a backward propagation neural network model. By combining the nonlinear mapping of the input and output of the backward propagation neural network model, performing gradient descent calculations, having a certain generalization ability and the characteristics of using different transfer functions, on the one hand, it enables it to monitor, evaluate and evaluate the use of cloud host resources. Data mining such as optimized configuration for multi-dimensional feature construction, and when new data enters the neural network model network for training, the neural network can adjust the weight to adapt to more data; on the other hand, make its output value But any value, and reduce the error of the resource monitoring model, in order to obtain better evaluation information and optimize the allocation strategy.

Optionally, in some embodiments of the present application, the above-mentioned acquiring training data, inputting the training data to the neural network model, and training the neural network model to obtain the resource monitoring model include: acquiring training data, Perform data preprocessing; store the smoothed and abnormal data processed training data in the training database, and set up configuration files, where the configuration files include network structure, training duration, training and test ratio arrangements, output content, and optimized learning Rate setting, optimization parameters and archiving rules setting; according to the configuration file, the training data is smoothed to obtain the prediction information; the prediction information is evaluated according to the preset comprehensive evaluation rules to obtain the evaluation information, where the evaluation The information includes the scoring range and analysis information; according to the evaluation information, an optimized configuration strategy corresponding to the evaluation information is generated to obtain the resource monitoring model; the resource monitoring model is tested for accuracy and performance through the created test script; if the accuracy of the test results When the first preset threshold is reached and the result of the performance test reaches the second preset threshold, the resource monitoring model is used as the final resource monitoring model; if the accuracy check result does not reach the first preset threshold and/or the result of the performance test If the second preset threshold is not reached, the training data is continuously updated and the preset comprehensive evaluation rules are modified, and the resource monitoring model is retrained until the accuracy detection result reaches the first preset threshold and the performance test result reaches The second preset threshold.

Training the neural network model so that the acquired resource monitoring model has the ability to analyze the target project information, the resource type and resource usage of multiple cloud hosts, to obtain the cloud host to be assessed and the resource to be assessed of the cloud host to be assessed , Obtain the first sampling data and the second sampling data of the resource to be evaluated, obtain the prediction information according to the first sampling data and the second sampling data, and obtain the prediction information according to the first sampling data and the second sampling data The data and the prediction information obtain the evaluation evaluation information, and the function of generating an optimized configuration strategy corresponding to the evaluation information according to the evaluation information, so as to better evaluate the resource utilization of the cloud host, thereby improving the resource utilization of the cloud host rate.

Optionally, in some embodiments of the present application, the resource monitoring model includes a data processing sub-model, a prediction sub-model, an evaluation sub-model, and a strategy generation sub-model. The prediction sub-model and the data processing sub-model are connected in series, and the evaluation sub-model and prediction are connected in series. The sub-models are connected in series, and the strategy generation sub-models are connected in series with the evaluation sub-models. The data processing sub-model is used to smooth the sampled data obtained by the resource monitoring model and abnormal data processing; the prediction sub-model is used to predict the sampled data output by the data processing sub-model to obtain prediction information; the evaluation sub-model is used to predict The prediction information output by the sub-model is evaluated to obtain the evaluation information; the strategy generation sub-model is used to generate the corresponding optimal configuration strategy according to the evaluation information output by the evaluation sub-model.

102. Obtain the input target project information, and analyze the target project information to obtain the project data type and project operation requirement information of the target project information.

After obtaining the target project information input by the user, the target project information is analyzed to obtain the data type of the operation required to complete the target project and the operation requirement information of the target project. Among them, the project operation requirement information includes, but is not limited to, the planned time for project completion, deadlines, and project completion functions. For example: the target project is a management system, by analyzing the target project information input by the user, to obtain the planned time, deadlines, amount of cloud host resources required for the completion of the management system research and development, the performance of the management system and the requirements of the management system Function and so on. By obtaining the project data type and project operation requirement information in the target project information, in order to provide a better basis and analysis basis for the evaluation of the use of resources and resource allocation of the cloud host.

Optionally, in some embodiments of the present application, the above-mentioned obtaining the input target project information and analyzing the target project information to obtain the project data type and project operation requirement information of the target project information includes: creating a project operation The demand table, where the project operation demand table includes the preset completion time of the project, the host resource demand and the optimal host resource allocation corresponding to the host resource demand; obtain the target project information input by the user, and perform data on the target project information Preprocessing, where data preprocessing includes missing value filling processing, denoising processing, and data standardization processing; divide the target item information after data preprocessing into N groups, and perform multiple iterations on the target item information divided into N groups Regroup to obtain the optimal grouping plan; obtain the project data type of each group of target project information in the optimal grouping plan; analyze the target project information to obtain the fourth key information; traverse the project operation requirement table according to the fourth key information, To obtain the project operation requirement information corresponding to the fourth key information.

Through the cluster analysis of the target project information, a concise conclusion form can be obtained, so that the resource monitoring model can intuitively and easily obtain the project data type and project operation requirement information of the target project. By traversing the project operation requirement table to obtain the project operation requirement information corresponding to the key information, the project operation requirement information can be obtained from multiple angles, so that the cloud host to be assessed and the resource to be assessed of the cloud host to be assessed can be determined more quickly and accurately.

103. Analyze the resource types and usage conditions of the cloud host, and determine the cloud host to be assessed and the resource to be assessed based on the project data type and project operation requirement information.

Among them, cloud hosts include all cloud hosts used by connections; resources to be evaluated include resource usage and remaining resources, and the remaining resources are the total resources of the cloud hosts to be evaluated minus used resources. By determining the cloud host to be evaluated and the resource to be evaluated according to the target project information, on the one hand, the cloud host resources are evaluated in a purposeful and targeted manner, and on the other hand, unnecessary operation operations are reduced to reduce System pressure and increase operating speed. For example: Analyze the resource types and resource usage conditions of all cloud hosts, match the resource usage types corresponding to the project data type to obtain the first cloud host, and match the resource usage conditions corresponding to the project operation requirement information to obtain The second cloud host, judges the cloud host that meets the project data type and meets the project operation demand information among the first cloud host and the second cloud host, as the cloud host to be evaluated, and obtains the resource usage and the resource usage of the cloud host to be evaluated The remaining resources.

104. Obtain the first sampling data of the resource to be evaluated in the first time period through the resource monitoring model, and acquire the second sampling data of the resource to be evaluated in the second time period.

Wherein, the start time of the first time period is later than the end time of the second time period. Both the first sampling data and the second sampling data are data of multiple sampling points that are sampled at a preset sampling interval and arranged in a forward time sequence. Both the first sampling data and the second sampling data include the occupancy rate information of the central processing unit CPU , Memory occupancy information and input/output IO occupancy information. The occupancy information includes the proportion of physical resources and total resources, the proportion of functionally described information resources and total resources, and the non-functionally described information resources and total resources. The proportion of network resources and the proportion of total resources, the proportion of used resources and the total resources, and the proportion of unused resources and the total resources. By collecting sampling data in two different time periods, the diversity of evaluation resources is increased, so as to improve the reference and accuracy of cloud host resource evaluation.

Optionally, in some embodiments of the present application, the aforementioned resource to be evaluated includes a first task in a first period and a second task in a second period. The aforementioned resource monitoring model is used to obtain the resource to be evaluated. The first sampling data in the first time period and obtaining the second sampling data of the resource to be evaluated in the second time period include: obtaining the first occupancy rate information and the first priority of the first task through the resource monitoring model, And obtain the second occupancy rate information and the second priority of the second task; classify the first task according to the first priority, and identify the first category label, and perform the second task according to the second priority Classify and identify the second category label; according to the first occupancy rate information, classify the first task that identifies the first category label, and identify the third category label, and the second occupancy rate information, to identify the second category label Classify the second task and identify the fourth category label; determine whether the first priority meets the first preset sampling condition and/or whether the first occupancy rate information meets the second preset sampling condition, and determine whether the second priority Whether the third preset sampling condition and/or the second occupancy rate information meets the fourth preset sampling condition; if the first priority meets the first preset sampling condition and/or the first occupancy rate information meets the second preset sampling Condition, according to the preset first sampling frequency, the tasks of the same category label that meet the first preset sampling condition and/or the second preset sampling condition are sampled to obtain the first sampling data; if the second priority meets If the third preset sampling condition and/or the second occupancy rate information meets the fourth preset sampling condition, then according to the preset second sampling frequency, the same sampling condition that meets the third preset sampling condition and/or the fourth preset sampling condition is The task of the category label performs sampling to obtain the second sampling data.

Through the foregoing implementation manners, it is possible to facilitate the understanding of the use status of resources in the cloud host and the degree of influence on the proportion of resources, thereby supporting the accuracy and rigor of the subsequent evaluation information to a certain extent. Since each task submitted by the user (that is, an application request) is submitted with a specific priority, any relatively high priority task can preempt resources used by low priority, and is at a different level of high priority The tasks have corresponding characteristics. For example, high-priority tasks in the normal production stage are sensitive to delay, and generally will not be rejected, suspended or suspended due to excessive use of resources. Therefore, in order to facilitate a clear understanding of the use of each occupancy rate information The status and the degree of influence on the proportion of resources, the task is an application request initiated by the user, and the above operations can be used.

105. Perform predictions on the cloud host to be evaluated based on the first sampling data and the second sampling data to obtain the first prediction data, the second prediction data, and the third prediction data respectively, using the third prediction data as the main information, and the first The first prediction data and the second prediction data are used as auxiliary information to obtain prediction information.

Wherein, the prediction includes the prediction of the resource usage status and the trend of each resource usage of the cloud host to be evaluated in the third period, and the start time of the third period is later than the end time of the first period. The first prediction data includes the prediction data obtained by predicting the usage status of each resource and the trend of each resource usage of the cloud host in the third period based on the first sampling data, and the second prediction data includes the prediction data to be evaluated based on the second sampling data. The prediction data obtained by predicting the usage status of each resource and the trend of each resource usage of the cloud host in the third period. The third prediction data includes the cloud host to be evaluated in the third period based on the first sampling data and the second sampling data. The prediction data obtained by predicting the usage status of each resource and the trend of each resource usage, the start time of the third period is later than the end time of the first period. The main information and auxiliary information refer to the processing of the main information first, and then combined with the auxiliary information for processing. By using multiple prediction data and using the third prediction data as the main information, and using the first prediction data and the second prediction data as auxiliary information, the accuracy of the evaluation can be improved.

Optionally, in some embodiments of the present application, after the foregoing forecast information is obtained, the method further includes: acquiring first time series data of the forecast information; performing sliding window processing on the first time series data to generate a preset number The time series subsequence with a preset length; analyze the statistical indicators of the time series subsequence to obtain statistical characteristic information, and use the statistical characteristic information as the updated forecast information, where the statistical characteristic information includes the maximum value, the minimum value, and the median , First quartile, third quartile, variance and standard deviation.

By extracting the statistical features of the forecast information, the data of the forecast information is more concentrated, more systematic, and more clearly reflecting the objective reality, so that the subsequent future forecast values are more biased towards the actual value, and the accuracy of the forecast is promoted.

Optionally, in some embodiments of the present application, before predicting the cloud host to be evaluated based on the first sampling data and the second sampling data, the method further includes: acquiring second time series data of the first sampling data, and Obtain the third time series data of the second sampling data; use the exponentially weighted moving average method EWMA in the data processing sub-model to evaluate the second time series data and the third time series data respectively to obtain the first smoothed data And the first smoothing data, the second time series data and the third time series data are evaluated as follows:

Among them, x _t is the actual second time series data or the actual third time series data at time t, the coefficient α is the rate of weighted decline, and V _t is the EWMA value at time t; through the Mahalanobis distance in the data processing sub-model The algorithm detects and identifies the extreme abnormal points in the first smoothing data and the second smoothing data respectively, and deletes the marked extreme abnormal points. The calculation for detecting and identifying the extreme abnormal points is as follows:

among them,

Is b and

the distance,

Is the mean vector of the smoothed data, b is the other objects in the smoothed data, and S is the covariance matrix.

Smooth processing and abnormal data processing are performed on the first sampled data and the second sampled data to obtain relatively balanced and stable time series data, reduce errors, and provide further support for the subsequent accurate values of future predicted values.

106. Perform feature extraction on the first sampling data, the second sampling data, and the prediction information to obtain the first key information, the second key information, and the third key information, respectively, and obtain the first sampling data of the resource to be evaluated. An alternative resource, and a second alternative resource for obtaining the resource to be evaluated in the second sampling data.

Among them, the analysis includes the identification and acquisition of the type, use time, operation rate, and resource proportion of each resource of the first sampling data, the second sampling data, and the prediction information. The first key information includes the type, usage time, operation rate, and resource proportion of each resource obtained by analyzing the first sampled data, and the second key information includes the type of each resource obtained by analyzing the second sampled data, Use time, operation rate, and resource proportion. The third key information includes the type, use time, operation rate, and resource proportion of each resource obtained by analyzing the forecast information. On the one hand, by obtaining multi-angle resource usage data, users can quickly and comprehensively learn the resource usage of the cloud host to be evaluated. On the other hand, by combining diversified data for evaluation, the accuracy of the evaluation can be improved.

Optionally, in some embodiments of the present application, the method includes a resource database. The resource database includes collected resource data corresponding to multiple data types. The first alternative resource of the resource to be evaluated in the first sampling data is obtained as described above. , And acquiring the second alternative resource of the resource to be evaluated in the second sampling data, including: acquiring first characteristic information of the resource to be evaluated in the first sampling data, and acquiring second characteristic information of the resource to be evaluated in the second sampling data , Where the first feature information and the second feature information both include the data type, the total capacity of the resource, the usage proportion of the resource corresponding to each data type, the performance and characteristics of the resource; the first feature information is obtained from the resource database according to the first feature information. The first resource data corresponding to one feature information, and the second resource data corresponding to the second feature information is obtained in the resource database according to the second feature information; the first resource data is calculated and filtered by preset replacement conditions to obtain the first resource data A replaceable resource, and the second resource data is calculated and filtered through preset replacement conditions to obtain the second replaceable resource; based on the first replaceable resource, analysis and matching are performed in cloud hosts other than the cloud host to be evaluated , To obtain the first alternative cloud host, and obtain the first resource usage information of the first alternative cloud host, and mark the first alternative cloud host and the first resource usage information on the first alternative resource to obtain The final first alternative resource, and according to the second alternative resource, perform analysis and matching among cloud hosts other than the cloud host to be evaluated to obtain the second alternative cloud host, and obtain the second alternative cloud host Second resource usage information, and mark the second replaceable cloud host and the second resource usage information on the second replaceable resource to obtain the final second replaceable resource.

By marking the resource usage information of alternative cloud hosts and alternative cloud hosts, on the one hand, users can learn about alternative cloud hosts to save resources, and on the other hand, provide multi-angle and multi-faceted data for evaluation , Thereby improving the accuracy and feasibility of the assessment.

107. Evaluate the cloud host to be evaluated based on the first key information, the second key information, the third key information, the first replaceable resource, and the second replaceable resource to obtain evaluation information.

According to the first key information, the second key information, and the third key information, the current use resources of the cloud host to be evaluated and the estimated use resources for a certain period of time in the future are evaluated, and based on the first alternative resource and the second alternative resource The evaluation can replace other cloud hosts of the cloud host to be evaluated in order to rationally use the cloud host and improve the resource utilization rate of the cloud host. For example: A cloud host is a cloud host to be evaluated, and a cloud host B is a cloud host other than the cloud host to be evaluated. According to the evaluation of the first key information, the second key information and the third key information, the available resources of the A Cloud host are 10% left, and the indicators do not meet the requirements of continued use and the requirements of the project. On the other hand, there are still 90% of the available resources of Cloud B, and it has been vacant for a long time. According to the first key information, the second key information, the third key information, the first alternative resource and the second alternative resource, Cloud host and B cloud host analysis, the resource type of B cloud host is the same and/or similar to A cloud host, which can be replaced, and all indicators meet the requirements of continued use and the requirements of the project, then the input goals are Project information, Cloud B host can be used as a replaceable cloud host for Cloud A host. Among them, the process information and result information generated by the analysis process are used as evaluation information.

108. According to the evaluation information, generate an optimized configuration strategy corresponding to the evaluation information, and output the evaluation information and optimized configuration strategy.

Among them, the optimal configuration strategy includes the resource usage status of the cloud host, the occupancy rate information of each physical resource of the cloud host, whether the cloud host needs to be configured to upgrade or reduce or release resources, and the cloud host's preset resources for each time period in the future Usage, user behavior matching, and semantic matching.

Compared with the existing mechanism, in the embodiments of the present application, on the one hand, data on resource usage from multiple perspectives is obtained, so that users can quickly and comprehensively learn about the resource usage of the cloud host to be evaluated. On the other hand, through multiple perspectives The data is evaluated to improve the rigor and accuracy of the evaluation. Therefore, this application can improve the resource utilization rate of the cloud host.

The technical features mentioned in any optional embodiment or optional implementation in the embodiment corresponding to FIG. 1 or the embodiment corresponding to FIG. 1 are also applicable to those corresponding to FIG. 2 and FIG. 3 in this application. In the embodiment, the similarities will not be repeated in the following.

The foregoing describes a method for evaluating cloud host resources in the present application, and the following describes an apparatus for executing the foregoing method for evaluating cloud host resources.

As shown in FIG. 2 is a schematic structural diagram of a device 20 for evaluating cloud host resources, which can be applied to the configuration and use of cloud hosts, for the release of cloud hosts that do not use resources, and the need to perform due to the minimal use of resources. The cloud host that reduces the configuration operation provides a reference. Among them, for cloud computing platform vendors, it can be used to provide quantitative standards for resource recovery to provide reference support for reducing the waste of internal computing resources; for cloud computing platform tenants, it can be used to provide information on the resource usage and rational use of business cloud hosts reference. The apparatus 20 in the embodiment of the present application can implement the method corresponding to the embodiment corresponding to FIG. 1 or any one of the optional embodiments or the optional implementation manners in the embodiment corresponding to FIG. 1 for evaluating cloud host resources. step. The functions implemented by the device 20 can be implemented by hardware, or can be implemented by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above-mentioned functions, and the modules may be software and/or hardware. The device 20 may include an input/output module 201, a processing module 202, and a display module 203. The functional realization of the input/output module 201, the processing module 202, and the display module 203 may refer to the embodiment corresponding to FIG. 1 or the implementation corresponding to FIG. 1 The operations performed in any optional embodiment or optional implementation manner in the example will not be repeated here. The processing module 202 can be used to control the receiving and sending operations of the input/output module 201, and the display module 203 can be used to display the processing operations of the processing module 202.

In some embodiments, the input and output module 201 is used to obtain training data and is used to obtain input target item information; the processing module 202 is used to input the training data obtained by the input and output module 201 into the neural network model, and the neural network model Train to obtain the resource monitoring model; analyze the target project information obtained by the input and output module 201 to obtain the project data type and project operation requirement information of the target project information; obtain the resource type and resource usage of the cloud host, and Determine the cloud host to be assessed and the resource to be assessed of the cloud host to be assessed according to the resource usage type, the resource usage situation, the project data type, and the project operation requirement information; obtain through the resource monitoring model The first sampling data of the resource to be evaluated in the first time period, and the second sampling data of the resource to be evaluated in the second time period are acquired; The cloud host to be evaluated performs prediction to obtain first prediction data, second prediction data, and third prediction data, respectively, using the third prediction data as main information, and combining the first prediction data and the first prediction data The second prediction data is used as auxiliary information to obtain prediction information; feature extraction is performed on the first sampling data, the second sampling data, and the prediction information, respectively, to obtain the first key information, the second key information, and the first key information, respectively. Three key information, and acquiring the first alternative resource of the resource to be evaluated in the first sampling data, and acquiring the second alternative resource of the resource to be evaluated in the second sampling data; according to the first key information, The second key information, the third key information, the first replaceable resource, and the second replaceable resource evaluate the cloud host to be evaluated to obtain evaluation information; generate and evaluate based on the evaluation information The optimized configuration strategy corresponding to the information, the evaluation information and the optimized configuration strategy are sent to the display module 203, and the evaluation information and the optimized configuration strategy are output through the display module 203; the display module 203 is used to display the evaluation information and the optimized configuration strategy.

Among them, the training data includes resource sampling data of multiple cloud hosts in multiple periods; the start time of the first period is later than the end time of the second period; the prediction includes the resource usage status of the cloud host to be evaluated in the third period and For the prediction of the trend of each resource usage, the start time of the third time period is later than the end time of the first time period.

Optionally, the aforementioned processing module 202 is further configured to: create a project operation demand table, where the project operation demand table includes the preset completion time of the project, the host resource demand, and the optimal host resource allocation corresponding to the host resource demand; Obtain the target project information input by the user, and perform data preprocessing on the target project information. The data preprocessing includes missing value filling processing, denoising processing and data standardization processing; the target project information after data preprocessing is divided into N groups, The target project information divided into N groups is regrouped many times by iterative method to obtain the optimal grouping plan; the project data type of each group of target project information in the optimal grouping plan is obtained; the target project information is analyzed to obtain the first Four key information: traverse the project operation demand table according to the fourth key information to obtain the project operation demand information corresponding to the fourth key information.

Optionally, the above-mentioned processing module 202 is further configured to: obtain the first occupancy rate information and the first priority in the first task, and obtain the second occupancy rate information and the second priority in the second task through the resource monitoring model According to the first priority, the first task is classified and the first category label is identified, and the second task is classified according to the second priority, and the second category label is identified; according to the first occupancy rate information, Classify the first task that identifies the first category label, identify the third category label, and second occupancy rate information, classify the second task that identifies the second category label, and identify the fourth category label; determine the first Whether the priority meets the first preset sampling condition and/or whether the first occupancy rate information meets the second preset sampling condition, and whether the second priority meets the third preset sampling condition and/or whether the second occupancy rate information Meet the fourth preset sampling condition; if the first priority meets the first preset sampling condition and/or the first occupancy rate information meets the second preset sampling condition, then according to the preset first sampling frequency, the Set the sampling condition and/or the task of the same category label under the second preset sampling condition to sample to obtain the first sampling data; if the second priority meets the third preset sampling condition and/or the second occupancy rate information meets the first Four preset sampling conditions, according to the preset second sampling frequency, sampling tasks of the same category label that meet the third preset sampling condition and/or the fourth preset sampling condition are sampled to obtain second sampling data.

Optionally, after the above-mentioned processing module 202 is executed to obtain the prediction information, it is further used to: obtain the first time series data of the prediction information; perform sliding window processing on the first time series data to generate a preset number of preset lengths. The time series subsequence; analyze the statistical indicators of the time series subsequence to obtain statistical feature information, and use the statistical feature information as the updated forecast information, where the statistical feature information includes the maximum value, the minimum value, the median, and the first quartile Digits, third quartile, variance and standard deviation.

Optionally, the processing module 202 described above is further configured to: obtain the second time series data of the first sampled data, and obtain the second sampled data before performing the prediction based on the first sampled data and the second sampled data of the cloud host to be evaluated. The third time series data; the second time series data and the third time series data are evaluated by the exponentially weighted moving average method EWMA in the data processing sub-model to obtain the first smoothing data and the first smoothing Data, the evaluation of the second time series data and the third time series data is calculated as follows:

among them,

Is b and

the distance,

Optionally, the above-mentioned processing module 202 is further configured to: obtain first characteristic information of the resource to be assessed in the first sampled data, and obtain second characteristic information of the resource to be assessed in the second sampled data, where the first characteristic information and The second feature information includes the data type, the total resource capacity, the proportion of resource usage corresponding to each data type, the performance and characteristics of the resource; the first resource corresponding to the first feature information is obtained from the resource database according to the first feature information Data, and obtain the second resource data corresponding to the second characteristic information in the resource database according to the second characteristic information; calculate and filter the first resource data through preset replacement conditions to obtain the first alternative resource, and obtain the first alternative resource through preset replacement conditions; Set replacement conditions to calculate and filter the second resource data to obtain the second alternative resource; according to the first alternative resource, perform analysis and matching in the cloud host other than the cloud host to be evaluated to obtain the first alternative cloud Host, and obtain the first resource usage information of the first alternative cloud host, and mark the first alternative cloud host and the first resource usage information on the first alternative resource to obtain the final first alternative resource, And according to the second alternative resource, perform analysis and matching among cloud hosts other than the cloud host to be evaluated to obtain the second alternative cloud host, and obtain the second resource usage information of the second alternative cloud host, and The second replaceable cloud host and the second resource usage information are marked on the second replaceable resource to obtain the final second replaceable resource.

Optionally, the above-mentioned processing module 202 is further used to: obtain training data, perform data preprocessing on the training data; store the training data that has undergone smoothing and abnormal data processing in a training database, and set a configuration file, where the configuration file Including network structure, training duration, training and testing ratio arrangement, output content, optimization learning rate setting, optimization parameter and archiving rule setting; according to the configuration file, the training data is smoothed to obtain prediction information; The comprehensive evaluation rules set up evaluate the forecast information to obtain the evaluation information. The evaluation information includes the scoring range and analysis information; according to the evaluation information, the optimized configuration strategy corresponding to the evaluation information is generated to obtain the resource monitoring model; and the created detection The script performs accuracy detection and performance testing on the resource monitoring model; if the accuracy detection result reaches the first preset threshold and the performance test result reaches the second preset threshold, the resource monitoring model is used as the final resource monitoring model; if If the result of the accuracy detection does not reach the first preset threshold and/or the result of the performance test does not reach the second preset threshold, the training data is continuously updated and the preset comprehensive evaluation rules are modified, and the resource monitoring model is retrained , Until the result of the accuracy detection reaches the first preset threshold and the result of the performance test reaches the second preset threshold.

In the embodiments of this application, on the one hand, multi-angle resource usage data is obtained so that users can quickly and comprehensively learn the resource usage of the cloud host to be evaluated; on the other hand, multi-angle data is evaluated to improve the evaluation Because of its rigor and accuracy, this application can improve the resource utilization of the cloud host.

Optionally, in some embodiments of this application, the technical features mentioned in any embodiment or implementation of the method for evaluating cloud host resources are also applicable to the evaluation of cloud host resources in this application. For the device 20 of the method, the similarities will not be repeated here.

The device 20 in the embodiment of the present application is described above from the perspective of modular functional entities. The following describes a computer device from the perspective of hardware, as shown in FIG. 3, which includes: a processor, a memory, a display, and an input and output unit ( It may also be a transceiver (not identified in FIG. 3) and a computer program stored in the memory and running on the processor. For example, the computer program may be a program corresponding to the method for evaluating cloud host resources in the embodiment corresponding to FIG. 1 or any optional embodiment in the embodiment corresponding to FIG. 1 or the optional implementation manner. For example, when a computer device implements the function of the device 20 shown in FIG. 2, the processor executes the computer program to implement the method for evaluating cloud host resources executed by the device 20 in the embodiment corresponding to FIG. 2 Or, when the processor executes the computer program, the function of each module in the apparatus 20 of the embodiment corresponding to FIG. 2 is realized. For another example, the computer program may be a program corresponding to the method in the embodiment corresponding to FIG. 1 or any optional embodiment in the embodiment corresponding to FIG. 1 or the optional implementation manner.

The so-called processor can be a central processing unit (Central Processing Unit, CPU), other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), off-the-shelf Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the computer device, and various interfaces and lines are used to connect various parts of the entire computer device.

The memory may be used to store the computer program and/or module, and the processor implements the computer by running or executing the computer program and/or module stored in the memory and calling data stored in the memory. Various functions of the device. The memory may mainly include a storage program area and a storage data area, where the storage program area can store an operating system and at least one application program required by a function (such as determining the cloud host to be evaluated and the resource to be evaluated of the cloud host to be evaluated, etc.) Etc.; the data storage area can store data created based on the use of the mobile phone (such as obtaining training data, etc.), etc. In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as hard disks, memory, plug-in hard disks, smart media cards (SMC), and secure digital (SD) cards. , Flash Card, at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage device.

The input and output units can also be replaced by receivers and transmitters, and they can be the same or different physical entities. When they are the same physical entity, they can be collectively referred to as input and output units. The input and output unit may be a transceiver.

The memory may be integrated in the processor, or may be provided separately from the processor.

The present application also provides a computer-readable storage medium. The computer-readable storage medium may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores computer instructions, and when the computer instructions are executed on the computer, the computer executes the following steps:

Obtain training data, input the training data into the neural network model, and train the neural network model to obtain a resource monitoring model, where the training data includes resource sampling data of multiple cloud hosts for multiple periods; obtain the input target item Information, and analyze the target project information to obtain the project data type and project operation requirement information of the target project information; obtain the resource type and resource usage of the cloud host, and according to the resource type, resource usage, and project data type And project operation requirements information, determine the cloud host to be evaluated and the resource to be evaluated; obtain the first sampling data of the resource to be evaluated in the first period through the resource monitoring model, and obtain the resource to be evaluated in the second period The second sampling data of the first time period is later than the end time of the second time period; the cloud host to be evaluated is predicted based on the first sampling data and the second sampling data to obtain the first prediction data and the first prediction data, respectively. The second prediction data and the third prediction data, and the third prediction data is used as the main information, and the first prediction data and the second prediction data are used as auxiliary information to obtain the prediction information, where the prediction includes the cloud host to be evaluated in the third To predict the usage status of each resource and the trend of each resource usage in a time period, the start time of the third time period is later than the end time of the first time period; feature extraction is performed on the first sampling data, the second sampling data and the prediction information respectively, To obtain the first key information, the second key information, and the third key information respectively, and obtain the first alternative resource of the resource to be evaluated in the first sampling data, and obtain the second alternative resource of the resource to be evaluated in the second sampling data Resources; evaluate the cloud host to be evaluated based on the first key information, second key information, third key information, first alternative resource, and second alternative resource to obtain the evaluation information; generate the corresponding evaluation information based on the evaluation information Optimize the configuration strategy, and output the evaluation information and optimize the configuration strategy.

Those skilled in the art can clearly understand that, for the convenience and conciseness of the description, the specific working process of the above-described system, device, and unit can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

Claims

A method for evaluating cloud host resources, the method comprising:

Obtain training data, input the training data into a neural network model, and train the neural network model to obtain a resource monitoring model, wherein the training data includes resource sampling data of multiple cloud hosts for multiple periods ；

Obtaining the input target project information, and analyzing the target project information to obtain the project data type and project operation requirement information of the target project information;

Obtain the used resource type and used resource situation of the cloud host, and determine the cloud host to be assessed and the cloud to be assessed based on the used resource type, the used resource situation, the project data type, and the project operation requirement information The host's resources to be evaluated;

Obtain the first sampling data of the resource to be evaluated in the first time period through the resource monitoring model, and acquire the second sampling data of the resource to be evaluated in the second time period, where the beginning of the first time period The start time is later than the end time of the second time period;

According to the first sampling data and the second sampling data, the cloud host to be evaluated is predicted to obtain the first prediction data, the second prediction data, and the third prediction data, respectively, and the third prediction data As main information, and using the first prediction data and the second prediction data as auxiliary information to obtain prediction information, wherein the prediction includes the status of each resource usage of the cloud host to be evaluated in the third period And prediction of the trend of each resource usage, the start time of the third time period is later than the end time of the first time period;

Perform feature extraction on the first sampling data, the second sampling data, and the prediction information to obtain the first key information, the second key information, and the third key information, respectively, and obtain the first sampling data The first alternative resource of the resource to be evaluated in the second sample data, and the second alternative resource of the resource to be evaluated in the second sampling data;

Evaluate the cloud host to be evaluated according to the first key information, the second key information, the third key information, the first replaceable resource, and the second replaceable resource to obtain the evaluation information;

According to the evaluation information, an optimized configuration strategy corresponding to the evaluation information is generated, and the evaluation information and the optimized configuration strategy are output.
The method according to claim 1, said obtaining the input target project information and analyzing the target project information to obtain the project data type and project operation requirement information of the target project information, comprising:

Creating a project operation requirement table, where the project operation requirement table includes the preset completion time of the project, the host resource demand, and the optimal host resource allocation corresponding to the host resource demand;

Acquiring target item information input by the user, and performing data preprocessing on the target item information, wherein the data preprocessing includes missing value filling processing, denoising processing, and data standardization processing;

Divide the target item information that has undergone data preprocessing into N groups, and regroup the target item information divided into N groups through repeated iterations to obtain the optimal grouping scheme;

Acquiring the project data type of each group of target project information in the optimal grouping scheme;

Analyze the target item information to obtain fourth key information;

Traverse the project operation requirement table according to the fourth key information to obtain project operation requirement information corresponding to the fourth key information.
The method according to claim 1, wherein the resource to be evaluated includes a first task in a first time period and a second task in a second time period, and the resource to be evaluated is acquired through the resource monitoring model. The first sampling data in the first time period and obtaining the second sampling data of the resource to be evaluated in the second time period include:

Obtaining the first occupancy rate information and the first priority in the first task, and obtaining the second occupancy rate information and the second priority in the second task through the resource monitoring model;

Classify the first task according to the first priority, and identify a first category label, and classify the second task according to the second priority, and identify a second category label;

According to the first occupancy rate information, the first task that identifies the first category label is classified, and the third category label is identified, and the second occupancy rate information is used to identify the first task that identifies the second category label. Two tasks are classified and the fourth category label is identified;

Determine whether the first priority meets a first preset sampling condition and/or whether the first occupancy rate information meets a second preset sampling condition, and determine whether the second priority meets a third preset sampling condition And/or whether the second occupancy rate information meets the fourth preset sampling condition;

If the first priority meets the first preset sampling condition and/or the first occupancy rate information meets the second preset sampling condition, then according to the preset first sampling frequency, the pair meets the first preset sampling condition and / Or the tasks of the same category label under the second preset sampling conditions are sampled to obtain the first sampling data;

If the second priority level meets the third preset sampling condition and/or the second occupancy rate information meets the fourth preset sampling condition, then according to the preset second sampling frequency, the pair meets the third preset sampling condition and /Or the tasks of the same category label under the fourth preset sampling condition are sampled to obtain the second sampling data.
The method according to claim 1, after said obtaining the prediction information, the method further comprises:

Acquiring the first time series data of the forecast information;

Performing sliding window processing on the first time series data to generate a preset number of time series subsequences with a preset length;

Analyze the statistical indicators of the time series subsequence to obtain statistical feature information, and use the statistical feature information as updated forecast information, where the statistical feature information includes the maximum value, the minimum value, the median, the first four Quantile, third quartile, variance and standard deviation.
The method according to claim 1, before predicting the cloud host to be evaluated based on the first sampling data and the second sampling data, the method further comprises:

Acquiring second time series data of the first sampling data, and acquiring third time series data of the second sampling data;

Through the exponentially weighted moving average method EWMA in the data processing sub-model, the second time series data and the third time series data are evaluated respectively to obtain the first smoothed data and the first smoothed data , The evaluation of the second time series data and the third time series data is calculated as follows:

Where, x t is the actual second time series data or the actual third time series data at time t, the coefficient α is the rate of weighted decline, and V t is the EWMA value at time t;

Through the Mahalanobis algorithm in the data processing sub-model, the extreme abnormal points in the first smoothing data and the second smoothing data are detected and identified, and the identified extreme abnormal points are deleted , The calculation for detecting and identifying the extreme abnormal point is as follows:

among them,
Is b and
the distance,
Is the mean vector of the smoothed data, b is other objects in the smoothed data, and S is the covariance matrix.
The method according to claim 1, wherein the method includes a resource database, the resource database includes resource data corresponding to multiple data types collected, and the acquiring of the first alternative of the resource to be evaluated in the first sampling data The resource, and the second alternative resource that obtains the resource to be evaluated in the second sampling data includes:

Acquire first characteristic information of the resource to be evaluated in the first sampled data, and acquire second characteristic information of the resource to be evaluated in the second sampled data, wherein the first characteristic information and the second characteristic information All include data type, total resource capacity, usage proportion of resources corresponding to each data type, performance and characteristics of resources;

Acquire first resource data corresponding to the first feature information in the resource database according to the first feature information, and obtain the second feature information from the resource database according to the second feature information Corresponding second resource data;

Calculate and filter the first resource data based on a preset replacement condition to obtain a first alternative resource, and perform calculation and screening on the second resource data based on the preset replacement condition to obtain a second alternative resource ；

According to the first replaceable resource, perform analysis and matching among cloud hosts other than the cloud host to be evaluated, to obtain a first replaceable cloud host, and obtain the first resource of the first replaceable cloud host Use information, and mark the first replaceable cloud host and the first resource use information on the first replaceable resource to obtain the final first replaceable resource, and according to the second replaceable resource Resources, perform analysis and matching in cloud hosts other than the cloud host to be evaluated to obtain a second alternative cloud host, and obtain second resource usage information of the second alternative cloud host, and compare the The second replaceable cloud host and the second resource usage information are marked on the second replaceable resource to obtain the final second replaceable resource.
The method according to claim 1, wherein the acquiring training data, inputting the training data to a neural network model, and training the neural network model to obtain a resource monitoring model includes:

Acquiring training data, and performing data preprocessing on the training data;

The training data that has been smoothed and processed with abnormal data is stored in the training database, and a configuration file is set. The configuration file includes the network structure, the training duration, the ratio of training and testing, the output content, and the settings for optimizing the learning rate. Setting and optimizing parameters and archiving rule settings;

Smoothing the training data according to the configuration file to obtain prediction information;

Evaluate the prediction information according to a preset comprehensive evaluation rule to obtain evaluation information, where the evaluation information includes a scoring range and analysis information;

Generating an optimized configuration strategy corresponding to the evaluation information according to the evaluation information to obtain a resource monitoring model;

Perform accuracy detection and performance testing on the resource monitoring model through the created detection script;

If the result of the accuracy detection reaches the first preset threshold and the result of the performance test reaches the second preset threshold, the resource monitoring model is used as the final resource monitoring model;

If the result of the accuracy detection does not reach the first preset threshold and/or the result of the performance test does not reach the second preset threshold, then by continuously updating the training data and modifying the preset comprehensive evaluation rules, The resource monitoring model is retrained until the result of the accuracy detection reaches the first preset threshold and the result of the performance test reaches the second preset threshold.
A device for evaluating resources of a cloud host, the device comprising:

Input and output module, used to obtain training data, used to obtain input target item information;

The processing module is used to input the training data obtained by the input and output module into a neural network model, and train the neural network model to obtain a resource monitoring model, wherein the training data includes multiple cloud hosts Resource sampling data for a period of time; analyze the input target project information obtained by the input and output module to obtain the project data type and project operation requirement information of the target project information; obtain the resource type and usage of the cloud host Resource situation, and determine the cloud host to be evaluated and the resource to be evaluated of the cloud host to be evaluated according to the resource usage type, the resource usage situation, the project data type, and the project operation requirement information; The resource monitoring model acquires the first sampling data of the resource to be evaluated in the first time period, and acquires the second sampling data of the resource to be evaluated in the second time period, wherein the start time of the first time period is later At the end of the second time period; predict the cloud host to be evaluated based on the first sampling data and the second sampling data to obtain the first prediction data, the second prediction data, and the third prediction, respectively Data, the third prediction data is used as the main information, and the first prediction data and the second prediction data are used as auxiliary information to obtain prediction information, wherein the prediction includes The host uses the state of each resource and the forecast of the trend of each resource usage in the third time period, the start time of the third time period is later than the end time of the first time period; Perform feature extraction on the second sampling data and the prediction information to obtain the first key information, the second key information, and the third key information, respectively, and obtain the first alternative resource of the resource to be evaluated in the first sampling data , And acquiring the second alternative resource of the resource to be evaluated in the second sampling data; according to the first key information, the second key information, the third key information, the first alternative resource and The second alternative resource evaluates the cloud host to be evaluated to obtain evaluation information; generates an optimized configuration strategy corresponding to the evaluation information according to the evaluation information, and compares the evaluation information with the optimized configuration strategy To a display module, and output the evaluation information and the optimal configuration strategy through the display module;

The display module is used to display the evaluation information and the optimal configuration strategy.
According to the device of claim 8, the processing module is further configured to:

Creating a project operation requirement table, where the project operation requirement table includes the preset completion time of the project, the host resource demand, and the optimal host resource allocation corresponding to the host resource demand;

Acquiring target item information input by the user, and performing data preprocessing on the target item information, wherein the data preprocessing includes missing value filling processing, denoising processing, and data standardization processing;

Divide the target item information that has undergone data preprocessing into N groups, and regroup the target item information divided into N groups through repeated iterations to obtain the optimal grouping scheme;

Acquiring the project data type of each group of target project information in the optimal grouping scheme;

Analyze the target item information to obtain fourth key information;

Traverse the project operation requirement table according to the fourth key information to obtain project operation requirement information corresponding to the fourth key information.
According to the device of claim 8, the processing module is further configured to:

Obtaining the first occupancy rate information and the first priority in the first task, and obtaining the second occupancy rate information and the second priority in the second task through the resource monitoring model;

Classify the first task according to the first priority, and identify a first category label, and classify the second task according to the second priority, and identify a second category label;

According to the first occupancy rate information, the first task that identifies the first category label is classified, and the third category label is identified, and the second occupancy rate information is used to identify the first task that identifies the second category label. Two tasks are classified and the fourth category label is identified;

Determine whether the first priority meets a first preset sampling condition and/or whether the first occupancy rate information meets a second preset sampling condition, and determine whether the second priority meets a third preset sampling condition And/or whether the second occupancy rate information meets the fourth preset sampling condition;

If the first priority meets the first preset sampling condition and/or the first occupancy rate information meets the second preset sampling condition, then according to the preset first sampling frequency, the pair meets the first preset sampling condition and / Or the tasks of the same category label under the second preset sampling conditions are sampled to obtain the first sampling data;

If the second priority level meets the third preset sampling condition and/or the second occupancy rate information meets the fourth preset sampling condition, then according to the preset second sampling frequency, the pair meets the third preset sampling condition and /Or the tasks of the same category label under the fourth preset sampling condition are sampled to obtain the second sampling data.
The device according to claim 8, after the processing module executes the obtaining prediction information, it is further configured to:

Acquiring the first time series data of the forecast information;

Performing sliding window processing on the first time series data to generate a preset number of time series subsequences with a preset length;

Analyze the statistical indicators of the time series subsequence to obtain statistical feature information, and use the statistical feature information as updated forecast information, where the statistical feature information includes the maximum value, the minimum value, the median, the first four Quantile, third quartile, variance and standard deviation.
The device according to claim 8, before executing the prediction of the cloud host to be evaluated based on the first sampling data and the second sampling data, the processing module is further configured to:

Acquiring second time series data of the first sampling data, and acquiring third time series data of the second sampling data;

Through the exponentially weighted moving average method EWMA in the data processing sub-model, the second time series data and the third time series data are evaluated respectively to obtain the first smoothed data and the first smoothed data , The evaluation of the second time series data and the third time series data is calculated as follows:

Where, x t is the actual second time series data or the actual third time series data at time t, the coefficient α is the rate of weighted decline, and V t is the EWMA value at time t;

Through the Mahalanobis algorithm in the data processing sub-model, the extreme abnormal points in the first smoothing data and the second smoothing data are detected and identified, and the identified extreme abnormal points are deleted , The calculation for detecting and identifying the extreme abnormal point is as follows:

among them,
Is b and
the distance,
Is the mean vector of the smoothed data, b is other objects in the smoothed data, and S is the covariance matrix.
According to the device of claim 8, the processing module is further configured to:

Acquire first characteristic information of the resource to be evaluated in the first sampled data, and acquire second characteristic information of the resource to be evaluated in the second sampled data, wherein the first characteristic information and the second characteristic information All include data type, total resource capacity, usage proportion of resources corresponding to each data type, performance and characteristics of resources;

Acquire first resource data corresponding to the first feature information in the resource database according to the first feature information, and obtain the second feature information from the resource database according to the second feature information Corresponding second resource data;

Calculate and filter the first resource data based on a preset replacement condition to obtain a first alternative resource, and perform calculation and screening on the second resource data based on the preset replacement condition to obtain a second alternative resource ；

According to the first replaceable resource, perform analysis and matching among cloud hosts other than the cloud host to be evaluated, to obtain a first replaceable cloud host, and obtain the first resource of the first replaceable cloud host Use information, and mark the first replaceable cloud host and the first resource use information on the first replaceable resource to obtain the final first replaceable resource, and according to the second replaceable resource Resources, perform analysis and matching in cloud hosts other than the cloud host to be evaluated to obtain a second alternative cloud host, and obtain second resource usage information of the second alternative cloud host, and compare the The second replaceable cloud host and the second resource usage information are marked on the second replaceable resource to obtain the final second replaceable resource.
According to the device of claim 8, the processing module is further configured to:

Acquiring training data, and performing data preprocessing on the training data;

The training data that has been smoothed and processed with abnormal data is stored in the training database, and a configuration file is set. The configuration file includes the network structure, the training duration, the ratio of training and testing, the output content, and the settings for optimizing the learning rate. Setting and optimizing parameters and archiving rule settings;

Smoothing the training data according to the configuration file to obtain prediction information;

Evaluate the prediction information according to a preset comprehensive evaluation rule to obtain evaluation information, where the evaluation information includes a scoring range and analysis information;

Generating an optimized configuration strategy corresponding to the evaluation information according to the evaluation information to obtain a resource monitoring model;

Perform accuracy detection and performance testing on the resource monitoring model through the created detection script;

If the result of the accuracy detection reaches the first preset threshold and the result of the performance test reaches the second preset threshold, the resource monitoring model is used as the final resource monitoring model;

If the result of the accuracy detection does not reach the first preset threshold and/or the result of the performance test does not reach the second preset threshold, then by continuously updating the training data and modifying the preset comprehensive evaluation rules, The resource monitoring model is retrained until the result of the accuracy detection reaches the first preset threshold and the result of the performance test reaches the second preset threshold.
A device for evaluating cloud host resources includes a memory, a processor, and a computer program stored on the memory and running on the processor, and the processor implements the following steps when the processor executes the computer program:

Obtain training data, input the training data into a neural network model, and train the neural network model to obtain a resource monitoring model, wherein the training data includes resource sampling data of multiple cloud hosts for multiple periods ；

Obtaining the input target project information, and analyzing the target project information to obtain the project data type and project operation requirement information of the target project information;

Obtain the used resource type and used resource situation of the cloud host, and determine the cloud host to be assessed and the cloud to be assessed based on the used resource type, the used resource situation, the project data type, and the project operation requirement information The host's resources to be evaluated;

Obtain the first sampling data of the resource to be evaluated in the first time period through the resource monitoring model, and acquire the second sampling data of the resource to be evaluated in the second time period, where the beginning of the first time period The start time is later than the end time of the second time period;

According to the first sampling data and the second sampling data, the cloud host to be evaluated is predicted to obtain the first prediction data, the second prediction data, and the third prediction data, respectively, and the third prediction data As main information, and using the first prediction data and the second prediction data as auxiliary information to obtain prediction information, wherein the prediction includes the status of each resource usage of the cloud host to be evaluated in the third period And prediction of the trend of each resource usage, the start time of the third time period is later than the end time of the first time period;

Perform feature extraction on the first sampling data, the second sampling data, and the prediction information to obtain the first key information, the second key information, and the third key information, respectively, and obtain the first sampling data The first alternative resource of the resource to be evaluated in the second sample data, and the second alternative resource of the resource to be evaluated in the second sampling data;

Evaluate the cloud host to be evaluated according to the first key information, the second key information, the third key information, the first replaceable resource, and the second replaceable resource to obtain the evaluation information;

According to the evaluation information, an optimized configuration strategy corresponding to the evaluation information is generated, and the evaluation information and the optimized configuration strategy are output.
The device according to claim 15, wherein the processor executes the computer program to realize the acquisition of the input target item information, and analyzes the target item information to obtain the item data type and the item data type of the target item information. When project operation requires information, it includes the following steps:

Creating a project operation requirement table, where the project operation requirement table includes the preset completion time of the project, the host resource demand, and the optimal host resource allocation corresponding to the host resource demand;

Acquiring target item information input by the user, and performing data preprocessing on the target item information, wherein the data preprocessing includes missing value filling processing, denoising processing, and data standardization processing;

Divide the target item information that has undergone data preprocessing into N groups, and regroup the target item information divided into N groups through repeated iterations to obtain the optimal grouping scheme;

Acquiring the project data type of each group of target project information in the optimal grouping scheme;

Analyze the target item information to obtain fourth key information;

Traverse the project operation requirement table according to the fourth key information to obtain project operation requirement information corresponding to the fourth key information.
The device according to claim 15, wherein the processor executes the computer program to realize that the resource to be evaluated includes a first task in a first time period and a second task in a second time period. When the resource monitoring model acquires the first sampling data of the resource to be evaluated in the first time period and acquiring the second sampling data of the resource to be evaluated in the second time period, it includes the following steps:

Obtaining the first occupancy rate information and the first priority in the first task, and obtaining the second occupancy rate information and the second priority in the second task through the resource monitoring model;

Classify the first task according to the first priority, and identify a first category label, and classify the second task according to the second priority, and identify a second category label;

According to the first occupancy rate information, the first task that identifies the first category label is classified, and the third category label is identified, and the second occupancy rate information is used to identify the first task that identifies the second category label. Two tasks are classified and the fourth category label is identified;

Determine whether the first priority meets a first preset sampling condition and/or whether the first occupancy rate information meets a second preset sampling condition, and determine whether the second priority meets a third preset sampling condition And/or whether the second occupancy rate information meets the fourth preset sampling condition;

If the first priority meets the first preset sampling condition and/or the first occupancy rate information meets the second preset sampling condition, then according to the preset first sampling frequency, the pair meets the first preset sampling condition and / Or the tasks of the same category label under the second preset sampling conditions are sampled to obtain the first sampling data;

If the second priority level meets the third preset sampling condition and/or the second occupancy rate information meets the fourth preset sampling condition, then according to the preset second sampling frequency, the pair meets the third preset sampling condition and /Or the tasks of the same category label under the fourth preset sampling condition are sampled to obtain the second sampling data.
The device according to claim 15, after the processor executes the computer program to realize the acquisition of prediction information, further comprising the following steps:

Acquiring the first time series data of the forecast information;

Performing sliding window processing on the first time series data to generate a preset number of time series subsequences with a preset length;

Analyze the statistical indicators of the time series subsequence to obtain statistical feature information, and use the statistical feature information as updated forecast information, where the statistical feature information includes the maximum value, the minimum value, the median, the first four Quantile, third quartile, variance and standard deviation.
The device according to claim 15, before the processor executes the computer program to implement the prediction of the cloud host to be evaluated based on the first sampling data and the second sampling data, further comprising the following steps :

Acquiring second time series data of the first sampling data, and acquiring third time series data of the second sampling data;

Through the exponentially weighted moving average method EWMA in the data processing sub-model, the second time series data and the third time series data are evaluated respectively to obtain the first smoothed data and the first smoothed data , The evaluation of the second time series data and the third time series data is calculated as follows:

Where, x t is the actual second time series data or the actual third time series data at time t, the coefficient α is the rate of weighted decline, and V t is the EWMA value at time t;

Through the Mahalanobis algorithm in the data processing sub-model, the extreme abnormal points in the first smoothing data and the second smoothing data are detected and identified, and the identified extreme abnormal points are deleted , The calculation for detecting and identifying the extreme abnormal point is as follows:

among them,
Is b and
the distance,
Is the mean vector of the smoothed data, b is other objects in the smoothed data, and S is the covariance matrix.
A computer-readable storage medium stores computer instructions in the computer-readable storage medium, and when the computer instructions are executed on a computer, the computer executes the following steps:

Obtain training data, input the training data into a neural network model, and train the neural network model to obtain a resource monitoring model, wherein the training data includes resource sampling data of multiple cloud hosts for multiple periods ；

Obtaining the input target project information, and analyzing the target project information to obtain the project data type and project operation requirement information of the target project information;

Obtain the used resource type and used resource situation of the cloud host, and determine the cloud host to be assessed and the cloud to be assessed based on the used resource type, the used resource situation, the project data type, and the project operation requirement information The host's resources to be evaluated;

Obtain the first sampling data of the resource to be evaluated in the first time period through the resource monitoring model, and acquire the second sampling data of the resource to be evaluated in the second time period, where the beginning of the first time period The start time is later than the end time of the second time period;

According to the first sampling data and the second sampling data, the cloud host to be evaluated is predicted to obtain the first prediction data, the second prediction data, and the third prediction data, respectively, and the third prediction data As main information, and using the first prediction data and the second prediction data as auxiliary information to obtain prediction information, wherein the prediction includes the status of each resource usage of the cloud host to be evaluated in the third period And prediction of the trend of each resource usage, the start time of the third time period is later than the end time of the first time period;

Perform feature extraction on the first sampling data, the second sampling data, and the prediction information to obtain the first key information, the second key information, and the third key information, respectively, and obtain the first sampling data The first alternative resource of the resource to be evaluated in the second sample data, and the second alternative resource of the resource to be evaluated in the second sampling data;

Evaluate the cloud host to be evaluated according to the first key information, the second key information, the third key information, the first replaceable resource, and the second replaceable resource to obtain the evaluation information;

According to the evaluation information, an optimized configuration strategy corresponding to the evaluation information is generated, and the evaluation information and the optimized configuration strategy are output.