Computing power resource scheduling method and device and electronic equipment

Abstract

The invention provides a computing power resource scheduling method and device based on regional computing power resource efficiency evaluation, and electronic equipment, relates to the technical field of computing power resource scheduling, and proposes that a multi-dimensional evaluation strategy is utilized to determine the computing power resource type of each region; according to the data calling frequency and the real-time calling demand, classifying the computing power demand of each region; when the computing power resource demand type region proposes a computing power demand request and the request belongs to a temperature data demand class or a cold data demand class, matching the computing power demand with a plurality of computing power resource supply type regions in a preset range to obtain a computing power scheduling scheme; therefore, the problems of unbalanced computing power resources, no demand in centralized supply, no response in centralized demand and the like at present are solved to the greatest extent. According to the invention, the blank of matching the target supply region and assisting the computing power resource scheduling in the set range to realize the globally optimal solution based on the regional computing power demand in the industry is filled up.

Description

Technical Field

[0001] The present disclosure relates to the technical field of computing power resource scheduling, and in particular to a computing power resource scheduling method, device and electronic equipment based on regional computing power resource efficiency evaluation. Background Art

[0002] As the digital economy era is fully opened, computing power, as the core resource of the digital era, is injecting new momentum into the digital transformation of all walks of life in a new form of productivity, and has become an important driving force for high-quality economic and social development. In the process of vigorously promoting the construction and development of computing power-related infrastructure, technology and industrial applications, it is necessary to build a computing power resource efficiency evaluation method from a global perspective, track the development level of computing power resources in multiple administrative regions and regions in real time, and formulate cross-regional resource scheduling plans in a timely and effective manner for regions with strong demand and tight resources, so as to achieve coordinated scheduling and full utilization of computing power resources.

[0003] However, due to the large differences in the actual construction of computing resources, some regions have large data demands, high computing application requirements, high energy consumption costs, and tight computing resources, while other regions have relatively small data demands, fewer energy consumption restrictions, and relatively sufficient computing resources. How to dynamically monitor the computing power resources in the region and promptly dispatch the computing power from areas with sufficient resources to areas in need to maximize the overall utilization of computing power resources is an urgent problem to be solved.

[0004] To elaborate, some of the current solutions are entropy weight evaluation systems, which do not take into account objective and subjective factors in a coordinated manner. That is, they tend to ignore the importance of the indicator itself and lack consideration of the actual significance of the indicator and the subjective intention of the decision maker. Although some evaluations combine the subjective analytic hierarchy process with the objective entropy weight method to obtain a comprehensive weight, the addition calculation method is relatively direct and does not fully consider the conflict between subjective and objective weights, and does not seek coordination and compromise between weights. Some are computing power scheduling systems, which cannot solve the formulation of the optimal resource allocation and scheduling solution between multiple demands and multiple supplies. That is, no model method has been proposed from a macro perspective to solve the global optimal scheduling solution for the demand and supply of computing power resources between regions, and it is impossible to achieve dynamic evaluation and coordinated scheduling of regional computing power resource construction and application. Summary of the invention

[0005] The present disclosure is proposed in view of the above problems. The present disclosure provides a computing resource scheduling method, device and electronic device based on regional computing resource efficiency evaluation, and correspondingly provides a computer-readable storage medium.

[0006] According to one aspect of the present disclosure, a computing resource scheduling method based on regional computing resource efficiency evaluation is provided, comprising:

[0007] Using a multi-dimensional evaluation strategy, determine the computing power resource type of each region; wherein the computing power resource type includes at least: computing power resource demand type and computing power resource supply type;

[0008] The computing power requirements of each region are classified according to the frequency of data retrieval and the real-time retrieval requirements. The computing power requirement types include at least the preset warm data requirement category and cold data requirement category.

[0009] When a computing power resource demand area makes a computing power demand request and the request belongs to the warm data demand type or the cold data demand type, its computing power demand is matched with several computing power resource supply type areas within a preset range to obtain a computing power scheduling plan.

[0010] In addition, according to a computing power resource scheduling method based on regional computing power resource efficiency evaluation in one aspect of the present disclosure, obtaining a computing power scheduling solution includes:

[0011] If at least two regions initiate computing resource requests at the same time, the computing resource supply range is determined based on the computing resource type;

[0012] Select target areas that provide computing resources within the computing resource supply range;

[0013] Combine the request area and the target area with the bipartite graph matching algorithm to determine the computing power scheduling plan.

[0014] In addition, according to an aspect of the present disclosure, a computing power resource scheduling method based on regional computing power resource efficiency evaluation, the determining of the computing power scheduling plan includes: based on the bipartite graph constructed by matching the computing power resource supply-type area with the computing power resource demand-type area, according to the established latency and cost requirements, the edge weight is calculated by the distance between regions and the computing power capacity value, and an augmented path query is performed within the computing power demand and supply range to determine the optimal scheduling plan.

[0015] In addition, according to a computing resource scheduling method based on regional computing resource efficiency evaluation in one aspect of the present disclosure, determining the computing resource type of each region includes:

[0016] Determine multi-dimensional evaluation indicators and combine and weight them, where the multi-dimensional indicators include at least computing power infrastructure resources and computing power network application capabilities;

[0017] Use the weighted weights to score the health of regional computing resources;

[0018] Conduct multi-dimensional collaborative correlation analysis based on the scoring results and build a computing resource efficiency evaluation model;

[0019] The computing power resource efficiency evaluation model is used to classify the computing power resources in each region.

[0020] In addition, according to a computing resource scheduling method based on regional computing resource efficiency evaluation in one aspect of the present disclosure, the combination weighting method includes:

[0021] Determine the objective weight of each evaluation indicator based on the entropy weight method;

[0022] Determine the subjective weight of each evaluation indicator based on the analytic hierarchy process;

[0023] A game theory mechanism is adopted to determine the final weight corresponding to each evaluation indicator based on the objective weight and the subjective weight.

[0024] In addition, according to a computing power resource scheduling method based on regional computing power resource efficiency evaluation in one aspect of the present disclosure, the computing power demand type further includes: a hot data demand type;

[0025] If a computing resource demanding area makes a request of the aforementioned hot data demand type, a computing resource scheduling solution that uses local or nearby resources and meets preset latency and cost requirements is adopted.

[0026] According to another aspect of the present disclosure, a computing resource scheduling device based on regional computing resource efficiency evaluation is provided, comprising:

[0027] A computing resource evaluation module is used to determine the computing resource type of each region using a multi-dimensional evaluation strategy; wherein the computing resource type includes at least: computing resource demand type and computing resource supply type;

[0028] The computing power demand classification module is used to classify the computing power demand of each region according to the frequency of data retrieval and the real-time retrieval demand. The computing power demand types include at least the preset warm data demand class and cold data demand class.

[0029] The computing power resource scheduling module is used to match the computing power demand with several computing power resource supply areas within a preset range to obtain a computing power scheduling plan when a computing power resource demand area makes a computing power demand request and the request belongs to the warm data demand type or the cold data demand type.

[0030] In addition, according to a computing power resource scheduling device based on regional computing power resource efficiency evaluation in one aspect of the present disclosure, the computing power resource scheduling module specifically includes a computing power scheduling unit, which is specifically used to:

[0031] If at least two regions initiate computing resource requests at the same time, the computing resource supply range is determined based on the computing resource type;

[0032] Select target areas that provide computing resources within the computing resource supply range;

[0033] Combine the request area and the target area with the bipartite graph matching algorithm to determine the computing power scheduling plan.

[0034] In addition, according to an aspect of the present disclosure, in a computing power resource scheduling device based on regional computing power resource efficiency evaluation, the computing power scheduling unit specifically includes a scheduling scheme determination component, which is specifically used to: on the basis of the bipartite graph constructed by matching the computing power resource supply-type area with the computing power resource demand-type area, according to the established latency and cost requirements, calculate the edge weight through the distance between regions and the computing power capacity value, and perform augmented path query within the computing power demand and supply range to determine the optimal scheduling scheme.

[0035] In addition, according to a computing resource scheduling device based on regional computing resource efficiency evaluation according to one aspect of the present disclosure, the computing resource evaluation module includes:

[0036] A combined weighting unit, used to determine multi-dimensional evaluation indicators and perform combined weighting, wherein the multi-dimensional indicators at least include computing power infrastructure resources and computing power network application capabilities;

[0037] The scoring unit is used to score the health of regional computing resources using the weighted weights;

[0038] An evaluation model building unit, which is used to perform multi-dimensional collaborative correlation analysis based on the scoring results and build a computing resource efficiency evaluation model;

[0039] The resource classification unit is used to classify the computing resources of each region using the computing resource efficiency evaluation model.

[0040] In addition, according to the computing power resource scheduling device based on regional computing power resource efficiency evaluation in one aspect of the present disclosure, the combined weighting unit includes:

[0041] The objective weight setting component is used to determine the objective weight of each evaluation indicator based on the entropy weight method;

[0042] The subjective weight setting component is used to determine the subjective weight of each evaluation indicator based on the hierarchical analysis method;

[0043] The final weight determination component is used to determine the final weight corresponding to each evaluation indicator based on the objective weight and the subjective weight by adopting a game theory mechanism.

[0044] In addition, according to one aspect of the present disclosure, the computing power resource scheduling device based on regional computing power resource efficiency evaluation, the computing power demand type also includes: thermal data demand class; the computing power resource scheduling module is also used for: if the computing power resource demand type area makes a request for the thermal data demand class, then adopt a computing power scheduling solution that uses local or nearby resources and meets preset latency and cost requirements.

[0045] According to another aspect of the present disclosure, an electronic device is provided, including: a memory for storing computer-readable instructions; and a processor for executing the computer-readable instructions, so that the electronic device executes the computing power resource scheduling method based on regional computing power resource efficiency evaluation as described above.

[0046] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium is provided for storing computer-readable instructions. When the computer-readable instructions are executed by a processor, the processor executes the computing power resource scheduling method based on regional computing power resource efficiency evaluation as described above.

[0047] According to another aspect of the present disclosure, a computer program product is provided. When the computer program product is executed by a computer, it is used to execute the method in the first aspect or any possible implementation of the first aspect. In a possible design of the fifth aspect, the relevant program involved in the product can be stored in whole or in part on a memory packaged with the processor, or in part or in whole on a storage medium not packaged with the processor.

[0048] As will be described in detail below, the computing resource scheduling method, device, electronic device, computer-readable storage medium and computer program product based on regional computing resource efficiency evaluation according to the embodiments of the present disclosure mainly solve the following problems:

[0049] First, from a global perspective, a regional computing resource efficiency evaluation system is constructed, which mainly evaluates the current basic computing resource capabilities and computing network application capabilities of each region, improves entropy weights through game theory, combines subjective and objective factors to determine the optimal weights, and outputs regional resource capacity values ​​to provide effective input for computing scheduling methods;

[0050] The second is to divide the types of computing resources within a larger administrative area based on collaborative analysis, and dynamically change the "computing resource supply type, demand type and balance type" according to the capacity value;

[0051] The third is to implement the optimal scheduling plan for computing power resources within the region based on the KM algorithm (Kuhn-Munkres), respond to the needs of multiple demand-oriented areas in a timely manner, consider factors such as latency, distance, resource capacity, and formulate a one-to-one "demand-supply" scientific scheduling plan to fully activate the computing power resources within the region, improve utilization and solve the situation of highly tight computing power resources in some areas.

[0052] Compared with the existing technology, this disclosure takes a global perspective and constructs a regional computing power performance evaluation scheme for the needs of the digital industry. It effectively evaluates the development level of computing power basic resource construction and computing power resource application effects in each region, and uses a subjective and objective combination weighting method to clarify the indicator weights and obtain the computing power development capacity value of the region. On this basis, the corresponding evaluation results of regional computing power basic resources and resource application effects are obtained, and a linear regression model is constructed using two-dimensional synergy analysis to quantitatively analyze the average application effect of unit computing power basic resources in the current region, further identify the type of regional computing power development, and provide effective input for subsequent cross-regional computing power resource scheduling. Through the above evaluation and classification, the optimal scheduling of computing resources is achieved for computing power demand areas through the KM bipartite graph algorithm: specifically, by dynamically identifying computing power resource demand areas and supply areas in the region, for multiple demand areas at the same time, the demand-supply matrix is ​​constructed for warm and cold computing power demand respectively, and the edge weight is calculated by the distance between regions and the computing power capacity value from the perspective of latency requirements and cost control, and the augmented path is found to obtain a solution that fully matches demand and supply, so as to maximize the total weight and complete the optimal scheduling, thus solving the current problems of unbalanced computing power resources, concentrated supply without demand, and concentrated demand without response to the greatest extent. It can be seen that the present disclosure fills the gap that does not exist in the industry, starting from the regional computing power demand, matching the target supply area, and assisting in the scheduling of computing power resources within the set range to achieve the global optimal solution.

[0053] The disclosed adaptation industry computing power map construction and real-time monitoring can fully grasp the efficiency level of computing power resources in the entire region, and determine the regional computing power resource type by synergistically analyzing the computing power resource capacity and the computing power resource application level, and output the global optimal resource scheduling plan according to the hot-warm-cold computing power demand. Based on the resource scheduling plan provided by the disclosed invention, idle computing power resources can be activated and the efficiency level of computing power resources in each region can be further improved. In addition, it can be considered that when expanding the application of the disclosed plan, work can be carried out around the three aspects of computing power network infrastructure construction, computing power core technology research, and computing power industry application prosperity, so as to jointly promote the maturity and growth of the computing power network industry and provide strong technical support for the digital construction of the industry and the sustainable development of the digital economy.

[0054] It is to be understood that both the foregoing general description and the following detailed description are exemplary, and are intended to provide further explanation of the technology as claimed. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The above and other purposes, features and advantages of the present disclosure will become more apparent by describing the embodiments of the present disclosure in more detail in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and do not constitute a limitation of the present disclosure. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0056] Figure 1 It is a schematic diagram of the main process of the computing power resource scheduling method based on regional computing power resource efficiency evaluation according to an embodiment of the present disclosure.

[0057] Figure 2 It is a schematic diagram illustrating the determination of computing power resource types based on fitting results according to an embodiment of the present disclosure.

[0058] Figure 3 It is a flowchart illustrating a method for determining a scheduling scheme according to an embodiment of the present disclosure.

[0059] Figure 4 It is a flowchart illustrating a computing resource scheduling method based on a KM algorithm according to an embodiment of the present disclosure.

[0060] Figure 5 It is a functional block diagram of a computing power resource scheduling device based on regional computing power resource efficiency evaluation according to an embodiment of the present disclosure.

[0061] Figure 6 is a hardware block diagram illustrating an electronic device according to an embodiment of the present disclosure.

[0062] Figure 7 is a schematic diagram illustrating a computer-readable storage medium according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solution and advantages of the present disclosure more obvious, the exemplary embodiments according to the present disclosure will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all the embodiments of the present disclosure, and it should be understood that the present disclosure is not limited to the exemplary embodiments described here.

[0064] First, refer to Figures 1 to 4 A computing power resource scheduling method based on regional computing power resource efficiency evaluation according to an embodiment of the present disclosure is described. Figure 1 It is a schematic diagram of the main process of the computing power resource scheduling method based on regional computing power resource efficiency evaluation according to an embodiment of the present disclosure. Figure 2 It is a schematic diagram further illustrating the determination of computing power resource types based on fitting results according to an embodiment of the present disclosure. Figure 3 It is a flowchart further illustrating the method for determining a scheduling scheme according to an embodiment of the present disclosure. Figure 4 It is a flowchart further illustrating the computing power resource scheduling method based on the KM algorithm according to an embodiment of the present disclosure.

[0065] Will refer to Figure 1 The method for scheduling computing resources based on regional computing resource efficiency evaluation according to an embodiment of the present disclosure is described in detail, including:

[0066] Step S1: using a multi-dimensional evaluation strategy to determine the computing power resource type of each region; wherein the computing power resource type includes at least: computing power resource demand type and computing power resource supply type;

[0067] Step S2: Classify the computing power requirements of each region according to the data retrieval frequency and the real-time retrieval requirements, where the computing power requirement types include at least a preset warm data requirement class and a cold data requirement class;

[0068] Step S3: When a computing power resource demand type area makes a computing power demand request, and the request belongs to the warm data demand type or the cold data demand type, its computing power demand is matched with several computing power resource supply type areas within a preset range to obtain a computing power scheduling plan.

[0069] In actual operation, the first link (computing power resource evaluation, CPRE) can construct a computing power resource efficiency evaluation index system from at least two dimensions: computing power basic resources and resource application (Arithmetic network) effects, and improve the entropy weight method based on the principles of game theory to comprehensively determine the index weights, thereby obtaining an objective evaluation result of regional computing power resource efficiency; the second link is to construct a computing power resource classification model based on synergy analysis, and divide the computing power resources of each region into two types: demand and supply according to the results of synergy analysis, and of course, equilibrium types can also be included; the third link (resource scheduling, RS) can be based on but not limited to the Kuhn-Munkres (KM) bipartite graph algorithm to achieve the optimal allocation and scheduling between demand-type and supply-type regional computing power resources, thereby significantly improving the utilization rate of computing power resources and effectively solving the problem of unbalanced computing power resources.

[0070] The following will combine Figure 1 The three steps are schematically explained in detail for each of the three links.

[0071] 1. Regional computing resource efficiency evaluation based on combined weighting

[0072] It can provide regional computing resource efficiency evaluation models for various cities, objectively evaluate the development of computing resources in various cities, grasp the overall layout of computing resources in real time, and provide effective input for the optimal scheduling method of computing resources in the subsequent region. The specific method of establishing the regional computing resource efficiency evaluation model is as follows:

[0073] Step 1: Build a regional computing resource efficiency evaluation index system

[0074] Based on the current development of regional computing resources, a regional computing resource efficiency evaluation system is constructed to comprehensively evaluate the use of computing resources and their application effects from the two aspects of infrastructure resources and network application capabilities. Among them, the infrastructure resource evaluation can be divided into six secondary dimensions: computing power, storage capacity, transportation capacity, energy consumption, safety, and investment, and after detailed division, a total of 23 evaluation indicators (three-level indicators) can be formed, which mainly evaluate the construction and development of regional infrastructure resources in various cities; the network application capability evaluation can be divided into five secondary dimensions: usage efficiency, cloud service status, industrial application, green energy efficiency and innovation capability, which are refined into 24 evaluation indicators (three-level indicators) in total, which mainly evaluate the regional resource utilization and network application process in various cities. It can be understood by those skilled in the art that the specific system composition such as the division of dimensions, indicator definition and calculation method can be set as needed in combination with the actual application scenario, and this disclosure does not limit this.

[0075] Step 2: Data input and standardization

[0076] First: Data Acquisition and Input

[0077] Obtain indicator data for various cities and regions. Some indicator data can be obtained directly from public materials, and some indicator data are obtained through indirect calculations, such as using established indicator definition formulas; other undisclosed data can be obtained through reporting by various cities and regions.

[0078] Second: Data standardization

[0079] Since the evaluation system selected in this example has 47 indicators, there are many indicators, and different data have different dimensions. In order to eliminate the impact of the dimension on the data, the indicator data needs to be standardized. The selected indicators are directional, so they can be standardized through Min-Max standardization:

[0080] Comparison method (+): The larger the value, the higher the score, the highest score is 100 points, the lowest score is 0 points, and the middle score is linear;

[0081] The lower of the two methods (-): The smaller the value, the higher the score, the highest score is 100, the lowest score is 0, and the middle score is linear.

[0082] For the positive impact (+) indicator, the standardized processing results are:

[0083]

[0084] For the negative impact (+) indicators, the standardized processing results are:

[0085]

[0086] Where i,j represents the jth indicator of the i-th city area, i=1,2.3...m,j=1,2.3...n; x ij Represented as the original data of the jth indicator in the i-th prefecture-level city; X ij It is expressed as the standardized value of the jth indicator in the i-th prefecture-level city; max(x ij ) and min(x ij ) represent the maximum and minimum values ​​of the jth indicator in the ith prefecture-level city.

[0087] Step 3: Improve the entropy weight method based on game theory to determine the weight of the grading indicators

[0088] This step uses the game theory to improve the entropy weight method, weighing the synergy and consistency of the hierarchical analysis and the entropy weight method, and comprehensively considering the data information and the actual significance of the indicators through a combination of subjective and objective methods, so as to obtain the optimal weighting result. Therefore, the specific steps of the regional computing resource efficiency model based on combined weighting are as follows:

[0089] First: Determine the objective weight of indicators based on the entropy weight method

[0090] Based on the standardized data, the information entropy of each indicator is determined, which is expressed as:

[0091]

[0092] Among them, E j is the information entropy of the jth indicator, n is the number of evaluation areas, p ij Expressed as the weight of the j-th indicator in the ith region.

[0093] Secondly: Based on the information entropy of each indicator, the weights of the three-level indicators are calculated, expressed as:

[0094]

[0095] Among them, m is the number of indicators, w j is the weight of the jth indicator.

[0096] Next: Determine the subjective weight of indicators based on the hierarchical analysis method

[0097] In actual operation, combined with the previous examples, the parameters can be set as follows:

[0098] First-level dimension: Infrastructure resources are set as A1, network application capabilities are set as A2, and the corresponding first-level weight is set as W a ;

[0099] Secondary dimensions: computing power, transportation capacity, storage capacity, energy consumption, security, investment, utilization efficiency, cloud service status, industrial application, green energy efficiency, and innovation capability are set as B1-B respectively. 11 , the corresponding secondary weight is W b ;

[0100] Level 3 indicator: Set level 3 indicator X j (j=1,2,...,n) corresponds to weight W c .

[0101] Furthermore, it is preferred to use Delphi method to classify the three-level index X j (j=1,2,...,n) to evaluate and score, and get the score M jj (j=1,2,...,n), construct a pairwise judgment matrix based on the 1-47 score pair criteria indicator layer, as shown below:

[0102] Evaluation indicators <![CDATA[X1]]> <![CDATA[X2]]> <![CDATA[X3]]> <![CDATA[X4]]> ... <![CDATA[X n ]]> <![CDATA[X1]]> 1 <![CDATA[1 / M 21 ]]> <![CDATA[1 / M 31 ]]> <![CDATA[1 / M 41 ]]> <![CDATA[1 / M ...1 ]]> <![CDATA[1 / M n1 ]]> <![CDATA[X2]]> <![CDATA[M 21 ]]> 1 <![CDATA[1 / M 32 ]]> <![CDATA[1 / M 42 ]]> <![CDATA[1 / M ...2 ]]> <![CDATA[1 / M n2 ]]> <![CDATA[X3]]> <![CDATA[M 31 ]]> <![CDATA[M 32 ]]> 1 <![CDATA[1 / M 43 ]]> <![CDATA[1 / M ...3 ]]> <![CDATA[1 / M n3 ]]> <![CDATA[X4]]> <![CDATA[M 41 ]]> <![CDATA[M 42 ]]> <![CDATA[M 43 ]]> 1 <![CDATA[1 / M (n-1)4 ]]> <![CDATA[1 / M n4 ]]> ... ... ... ... <![CDATA[M (n-1)4 ]]> 1 <![CDATA[1 / M n(n-1) ]]> <![CDATA[X n ]]> <![CDATA[M n1 ]]> <![CDATA[M n2 ]]> <![CDATA[M n3 ]]> <![CDATA[M n4 ]]> <![CDATA[M n(n-1) ]]> 1

[0103] Normalize the index by column to get the criterion index layer X j The weight of each indicator is W cj , and calculate the maximum eigenvalue λ of the judgment matrix max , and consistency indicators: By subjecting CI to consistency testing, it passes when the consistency ratio is as follows:

[0104] If the opposite is true, it means that the judgment matrix fails and should be discarded. The test results can be referred to as follows:

[0105] Degree 1 2 3 4 5 6 7 8 9 10 ... RI 0 0 0.52 0.89 1.12 1.26 1.36 1.41 1.46 0.49 ...

[0106] After the consistency test, according to the correspondence between the 47 indicators at the indicator level and the 11 dimensions at the second level in the evaluation system in the previous example, the third-level indicators and the 11 second-level dimensions W are calculated. b The weight of , in a paradigm example:

[0107] W b1 =W c1 +W c2 +W c3 +W c4 +W c6

[0108] According to the corresponding relationship between the 11 secondary dimensions and the 2 primary dimensions in the evaluation system, the weight of the secondary dimension to the primary dimension is calculated as follows:

[0109] W a1 =W b1 +W b2 +W b3 +Wb4 +W b5 +W b6

[0110] W a2 =W b7 +W b8 +W b9 +W b10 +W b11

[0111] Then: Output regional capability scores based on game theory combined weighting

[0112] In order to take into account both subjective and objective weighting methods and improve the rationality of weights, the game theory method is used to calculate the final weight results as follows: α = α1, α2, ..., α L Represents the combined weight coefficient in the index weighting method L, and the normalized optimized combined coefficient is obtained Game theory method combination weighting W * ,in:

[0113]

[0114] In order to seek consistency and compromise between different weights, the calculation is aimed at minimizing the deviation of and, and the L linear weight combination coefficients are optimized to obtain the optimal weights. According to the differential properties of the matrix, the linear equation system of the first-order derivative condition is optimized:

[0115]

[0116] The final combined weight W * for:

[0117]

[0118] Finally: Output the scores of infrastructure resources and network application capabilities of cities and regions

[0119] The final weight W of the indicator layer is calculated according to the combined weighting method. * , calculate the infrastructure resource score A1 and network application capability score A2 of all cities and regions:

[0120]

[0121] (II) Classification of regional computing resources based on synergy analysis

[0122] Step 4: Construct a collaborative analysis model of “computing power basic resources-resource application effects”

[0123] Based on the scoring results (A1, A2) in the areas of regional computing power basic resources and resource application effects in various cities, a two-dimensional collaborative correlation analysis was conducted, and a linear regression model was constructed to quantitatively explore the application effect level of unit computing power resources in various cities.

[0124] First: Correlation Analysis

[0125] Create a scatter plot with "resource application effect score" as the Y-axis and "computing power basic resource score" as the X-axis to observe whether there is a linear correlation trend between the resource application effect score and the computing power basic resource score.

[0126] The Pearson correlation coefficient is calculated to analyze the linear correlation between the resource application effect score and the computing power basic resource score, as follows:

[0127]

[0128] Among them, r represents the Pearson correlation coefficient, and its value range is [-1,1]. The larger the absolute value, the stronger the correlation. The closer it is to 1, the more positive linear correlation there is between the two groups of variables. The closer it is to -1, the more negative linear correlation there is between the two groups of variables. k Indicates the basic resource score of regional computing power. Represents the average score of regional computing power basic resources, Y k Indicates the regional resource application effect score, represents the average value of regional resource application effect, and n represents the sample size.

[0129] Second: Linear Regression

[0130] Through correlation analysis and testing, we finally take "resource application effect score" as the dependent variable x and "computing power basic resource score" as the independent variable y, and build a linear regression model based on the least squares method. For example, it can be set as:

[0131]

[0132] The least squares method is used to estimate the parameters a and b in the linear regression equation. The formula is as follows:

[0133]

[0134] By R 2 The value measures the goodness of fit of the regression model, R 2 The value range is [0,1]. The larger the value, the better the regression model fits the data object. 2 The calculation formula is as follows:

[0135]

[0136] Goodness of fit R2 ≥0.7, the fitting effect is considered qualified. The linear regression model is obtained The average application effect level corresponding to the basic computing resources per unit of computing power in each region is b.

[0137] Step 5: Determine the regional computing resource type based on the results of the synergy analysis model

[0138] The regional computing power development type is determined based on the linear fitting results. According to the classification judgment rules, the application effect level corresponding to the basic computing power resources of each region is compared with the average value to divide the regional computing power development type. In this step, the regional type is divided into the input basis of the computing power resource scheduling model, such as Figure 2 As shown in the figure, the sample points can be divided into three types through the fitting line, as follows:

[0139] (1) R area (Area above the fitting line), set as "computing resource demand type";

[0140] (2) S area (the area below the fitted line), set to "computing resource supply type";

[0141] (3) Area B (on the fitting line), set to "computing resource balance type".

[0142] By evaluating the development level of computing power basic resources and computing power application effects in each urban area, the type of computing power resources in each city is determined based on the synergy analysis model, and a list of computing power-demanding cities and computing power-supplying cities is output by administrative region. This serves as the basis for the subsequent scheduling of computing power resources for warm and cold computing power needs, effectively revitalizing idle and inefficient computing power resources, and further improving the utilization efficiency of unit computing power resources.

[0143] (III) Optimal Scheduling of Computing Resources Based on Bipartite Graph Algorithm

[0144] Step 6: Optimal Scheduling Method for Computing Resources Based on Kuhn-Munkres Bipartite Graph Algorithm

[0145] In order to solve the problems of mismatch between computing power demand and resources in various regions, low utilization rate of computing power resources in some computing power supply-type cities, and limited computing power resources in some computing power demand-type cities with strong demand, this step constructs a "hot-warm-cold" computing power resource scheduling method. Based on the frequency level of data retrieval, the data types of urban computing power demand are divided into three categories of data. The first-level data is data with frequent retrieval and real-time retrieval requirements, the second-level data is data with average retrieval and no real-time retrieval requirements, and the third-level data is data with low retrieval and no real-time retrieval requirements. The first, second, and third categories of data are also called hot data, warm data, and cold data.

[0146] (1) For the first type of computing power demand (hot data), a scheduling solution that meets the latency level requirements locally is adopted.

[0147] (2) For the second type of computing power demand (warm data), the KM algorithm is used to calculate the matching of computing power resources and demand in the administrative area.

[0148] (3) For the third type of computing power demand (cold data), the KM algorithm is used to calculate the matching of computing power resources and demand in all cities.

[0149] Here combined Figure 3 As an example, let's first give the specific implementation logic for determining the scheduling solution:

[0150] Step S31, using a bipartite graph constructed by matching computing resource supply-type regions with computing resource demand-type regions as a basis;

[0151] Step S32: Calculate edge weights based on the distance between regions and computing power according to the established latency and cost requirements;

[0152] Step S33: perform augmented path query within the computing power demand and supply range to determine the optimal scheduling solution.

[0153] In the preferred embodiment of the present disclosure, the optimal scheduling process of computing resources based on the KM algorithm can be referred to Figure 4 As shown, in Figure 4 The conditions used in the examples are first described as follows:

[0154] (1) It is assumed that at least two computing power demanding cities R1, R2, R3, etc. propose computing power demands at the same time;

[0155] (2) Computing power demand-type cities and computing power supply-type cities change dynamically based on the health assessment of computing power resources.

[0156] First: Screen the computing power demand and supply area range

[0157] When at least two computing resource demanding cities (R) in step 5 simultaneously put forward warm or cold computing power demands, first count their computing resource demand floating point numbers, and select supplying cities S1, S2, S3... that meet the demands (computing resource floating point number of supplying cities ≥ computing resource demand floating point number).

[0158] For warm data, the supply-type city selection range is the administrative region to which the demand-type city belongs. When the demand-type city is the main city, the supply-type city selection range is the two administrative regions closest to it. For cold data, the supply-type city selection range can be all cities.

[0159] Second: Create a bipartite graph

[0160] The matching between computing resource supply cities (S) and computing resource demand cities (R) is considered as a bipartite graph G = (S, R, E). S represents the set of computing resource supply areas, R represents the set of computing resource demand areas, and E represents the set of edges connecting computing resource supply areas and demand areas. The weight of each edge represents the matching degree between computing resource supply areas and demand areas.

[0161] Next: Initialization

[0162] According to the number of nodes in the bipartite graph, establish an S*R matrix M, and use Q tp represents the weight between the tth, t=1,2,3...,Rth computing resource demand city and the pth (p=1,2,3...,S)th supply city, that is, the edge weight between the tth computing resource demand city and the pth supply city. Set all initial values ​​to 0, for each left r t ∈R sets a top value θ p =max(Q tp ) is the tth R pair t1-t t The maximum matching score of the right side s p ∈S sets the top mark β t =0.

[0163] Then: calculate the edge weights

[0164] Based on the latency requirements and cost control of computing power scheduling, each element Q in the matrix M tp According to the distance between cities and the infrastructure resource score A1 (obtained in step 3), the corresponding edge weights are calculated, and the supply cities are matched to specific demand cities through the best matching of the bipartite graph, so that the total cost of the scheduling plan is minimized.

[0165] Dis represents the distance between the tth computing resource demand city and the pth computing resource supply city. The smaller the distance, the smaller the latency and the lower the transmission cost.

[0166] A1 represents the resource capacity score of the tth computing resource supply city. The higher the score, the more abundant its computing resources are and the lower its resource cost is.

[0167] The Dis indicator is normalized by the negative (+) indicator in step 2 to obtain Dis'.

[0168] Assuming the coefficients of the inter-city location distance Dis' and infrastructure resource score A1 are both 50%, then Q tp for:

[0169] Q tp=50%*Dis+50%*A1

[0170] After: Matching the goals of demand-oriented cities and supply-oriented cities

[0171] Find the maximum match: Use recursive vertical search for augmenting paths to perform the maximum match of G = (S, R, E).

[0172] Before the maximum matching is found, each time from s p ∈S starts looking for augmenting paths; if s p Already matched, then from s p Start looking for augmenting paths from the matched R vertices. For the augmenting paths found, replace the unmatched points s p and r t Match, for the original match, r t0 Cancel the match with s, and replace s p With r t0 This process is repeated until no augmenting path can be found, which is the maximum match.

[0173] If the maximum match achieves the maximum weight match, the computing power scheduling plan is output. Otherwise, the top mark value is modified for each s p ∈S and r t ∈R, update the label according to the formula:

[0174] θ p '=θ p -d

[0175] β t '=β t +d

[0176] Where d = min{θ p +β t -Q tp}.

[0177] Repeat the above steps until the optimal match with the largest sum of edge weights is found.

[0178] Finally: Output scheduling plan

[0179] Through the above steps, we can get the computing resource demand city r in the region. t The only optimal supply city s that matches p , to achieve the optimal scheduling plan for computing resource demands at the same time and solve the problem of resource shortage to the greatest extent.

[0180] The above describes the computing resource scheduling method based on regional computing resource efficiency evaluation according to the embodiment of the present disclosure. The following further describes the computing resource scheduling device based on regional computing resource efficiency evaluation for implementing the above control method. Figure 5It is a functional block diagram of a computing power resource scheduling device based on regional computing power resource efficiency evaluation according to an embodiment of the present disclosure.

[0181] like Figure 5 As shown, the computing power resource scheduling device 500 based on regional computing power resource efficiency evaluation according to an embodiment of the present disclosure includes:

[0182] The computing resource evaluation module 501 is used to determine the computing resource type of each region using a multi-dimensional evaluation strategy; wherein the computing resource type includes at least: computing resource demand type and computing resource supply type;

[0183] The computing power demand classification module 502 is used to classify the computing power demand of each region according to the frequency of data retrieval and the real-time retrieval demand, and the computing power demand type includes at least a preset warm data demand class and a cold data demand class;

[0184] The computing power resource scheduling module 503 is used to match the computing power demand with several computing power resource supply areas within a preset range to obtain a computing power scheduling plan when a computing power resource demand area makes a computing power demand request and the request belongs to the warm data demand type or the cold data demand type.

[0185] Furthermore, the computing resource scheduling module specifically includes a computing resource scheduling unit, which is specifically used to:

[0186] If at least two regions initiate computing resource requests at the same time, the computing resource supply range is determined based on the computing resource type;

[0187] Select target areas that provide computing resources within the computing resource supply range;

[0188] Combine the request area and the target area with the bipartite graph matching algorithm to determine the computing power scheduling plan.

[0189] Furthermore, the computing power scheduling unit specifically includes a scheduling scheme determination component, which is specifically used to: on the basis of the bipartite graph constructed by matching the computing power resource supply-type area with the computing power resource demand-type area, according to the established latency and cost requirements, calculate the edge weight through the distance between areas and the computing power capacity value, and perform augmented path query within the computing power demand and supply range to determine the optimal scheduling scheme.

[0190] Furthermore, the computing resource evaluation module includes:

[0191] A combined weighting unit, used to determine multi-dimensional evaluation indicators and perform combined weighting, wherein the multi-dimensional indicators at least include computing power infrastructure resources and computing power network application capabilities;

[0192] The scoring unit is used to score the health of regional computing resources using the weighted weights;

[0193] An evaluation model building unit, which is used to perform multi-dimensional collaborative correlation analysis based on the scoring results and build a computing resource efficiency evaluation model;

[0194] The resource classification unit is used to classify the computing resources of each region using the computing resource efficiency evaluation model.

[0195] Furthermore, the combined weighting unit includes:

[0196] The objective weight setting component is used to determine the objective weight of each evaluation indicator based on the entropy weight method;

[0197] The subjective weight setting component is used to determine the subjective weight of each evaluation indicator based on the hierarchical analysis method;

[0198] The final weight determination component is used to determine the final weight corresponding to each evaluation indicator based on the objective weight and the subjective weight by adopting a game theory mechanism.

[0199] Furthermore, the computing power demand type also includes: hot data demand class; the computing power resource scheduling module is also used for: if the computing power resource demand type area makes a request for the hot data demand class, then adopt a computing power scheduling solution that uses local or nearby resources and meets preset latency and cost requirements.

[0200] Figure 6 is a hardware block diagram of an electronic device 600 according to an embodiment of the present disclosure. The electronic device according to an embodiment of the present disclosure includes at least a processor; and a memory for storing computer-readable instructions. When the computer-readable instructions are loaded and executed by the processor, the processor executes the computing power resource scheduling method based on regional computing power resource efficiency evaluation as described above.

[0201] Figure 6 The electronic device 600 shown specifically includes: a central processing unit (CPU) 601, a graphics processing unit (GPU) 602 and a memory 603. These units are interconnected via a bus 604. The central processing unit (CPU) 601 and / or the graphics processing unit (GPU) 602 can be used as the above-mentioned processor, and the memory 603 can be used as the above-mentioned memory for storing computer-readable instructions. In addition, the electronic device 600 may also include a communication unit 605, a storage unit 606, an output unit 607, an input unit 608 and an external device 609, which are also connected to the bus 604.

[0202] Figure 7 is a schematic diagram illustrating a computer-readable storage medium according to an embodiment of the present disclosure. Figure 7As shown, a computer-readable storage medium 700 according to an embodiment of the present disclosure has computer-readable instructions 701 stored thereon. When the computer-readable instructions 701 are executed by a processor, the computing power resource scheduling method based on regional computing power resource efficiency evaluation according to an embodiment of the present disclosure described with reference to the above figures is executed. The computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may, for example, include a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may, for example, include a read-only memory (ROM), a hard disk, a flash memory, an optical disk, a magnetic disk, etc.

[0203] Finally, it can be added that according to a computer program product (which may include the above-mentioned device) of an embodiment of the present disclosure, when the computer program product is running on a terminal device, the terminal device executes the computing resource scheduling method based on regional computing resource efficiency evaluation of the aforementioned embodiment or an equivalent implementation. Through the description of the above implementation methods, it can be seen that those skilled in the art can clearly understand that all or part of the steps in the above implementation method can be implemented by means of software plus the necessary general hardware platform. Based on this understanding, the above-mentioned computer program product may include but is not limited to APP.

[0204] The above describes the computing resource scheduling method, device, electronic device, computer-readable storage medium and computer program product based on regional computing resource efficiency evaluation according to the embodiments of the present disclosure with reference to the accompanying drawings. The main idea is to solve the following problems:

[0205] First, from a global perspective, a regional computing resource efficiency evaluation system is constructed, which mainly evaluates the current basic computing resource capabilities and computing network application capabilities of each region, improves entropy weights through game theory, combines subjective and objective factors to determine the optimal weights, and outputs regional resource capacity values ​​to provide effective input for computing scheduling methods;

[0206] The second is to divide the types of computing resources within the administrative region based on collaborative analysis, and dynamically change the "computing resource supply type, demand type and balance type" according to the capacity value;

[0207] The third is to implement the optimal scheduling plan for computing power resources within the region based on the KM algorithm (Kuhn-Munkres), respond to the needs of multiple demand-oriented areas in a timely manner, consider factors such as latency, distance, resource capacity, and formulate a one-to-one "demand-supply" scientific scheduling plan to fully activate the computing power resources within the region, improve utilization and solve the situation of highly tight computing power resources in some areas.

[0208] Compared with the existing technology, this disclosure takes a global perspective and constructs a regional computing power performance evaluation scheme for the needs of the digital industry. It effectively evaluates the development level of computing power basic resource construction and computing power resource application effects in each region, and uses a subjective and objective combination weighting method to clarify the indicator weights and obtain the computing power development capacity value of the region. On this basis, the corresponding evaluation results of regional computing power basic resources and resource application effects are obtained, and a linear regression model is constructed using two-dimensional synergy analysis to quantitatively analyze the average application effect of unit computing power basic resources in the current region, further identify the type of regional computing power development, and provide effective input for subsequent cross-regional computing power resource scheduling. Through the above evaluation and classification, the KM bipartite graph algorithm is used to achieve optimal scheduling of computing resources for areas with computing power demand: specifically, by dynamically identifying computing power resource demand areas and supply areas within the region, for multiple demand areas at the same time, a demand-supply matrix is ​​constructed for warm and cold computing power demands respectively. From the perspective of latency requirements and cost control, the edge weights are calculated by the distance between regions and the computing power capacity value, and the augmentation path is found to obtain a solution that fully matches demand and supply, maximize the total weight, and complete the optimal scheduling, which solves the current problems of unbalanced computing power resources, concentrated supply without demand, and concentrated demand without response to the greatest extent.

[0209] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this disclosure.

[0210] The basic principles of the present disclosure are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present disclosure are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present disclosure. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, and are not limitations. The above details do not limit the present disclosure to the necessity of adopting the above specific details to be implemented.

[0211] The block diagrams of the devices, apparatuses, equipment, and systems involved in this disclosure are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including," "comprising," "having," and the like are open words, referring to "including but not limited to," and can be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or," and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and can be used interchangeably therewith.

[0212] Additionally, as used herein, "or" used in a list of items beginning with "at least one" indicates a separate list, so that, for example, a list of "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Furthermore, the word "exemplary" does not mean that the example described is preferred or better than other examples.

[0213] It should also be noted that in the system and method of the present disclosure, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present disclosure.

[0214] Various changes, substitutions, and modifications of the techniques described herein may be made without departing from the teachings defined by the appended claims. Furthermore, the scope of the claims of the present disclosure is not limited to the specific aspects of the processes, machines, manufactures, compositions of events, means, methods, and actions described above. Currently existing or later to be developed processes, machines, manufactures, compositions of events, means, methods, or actions that perform substantially the same functions or achieve substantially the same results as the corresponding aspects described herein may be utilized. Thus, the appended claims include such processes, machines, manufactures, compositions of events, means, methods, or actions within their scope.

[0215] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0216] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present disclosure to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A computing resource scheduling method based on regional computing resource efficiency evaluation, characterized in that: The scheduling method comprises: Using a multi-dimensional evaluation strategy, determine the computing power resource type of each region; wherein the computing power resource type includes at least: computing power resource demand type and computing power resource supply type; The computing power requirements of each region are classified according to the frequency of data retrieval and the real-time retrieval requirements. The computing power requirement types include at least the preset warm data requirement category and cold data requirement category. When a computing power resource demand area makes a computing power demand request and the request belongs to the warm data demand type or the cold data demand type, its computing power demand is matched with several computing power resource supply type areas within a preset range to obtain a computing power scheduling plan.

2. The computing resource scheduling method based on regional computing resource efficiency evaluation according to claim 1, characterized in that: The computing power scheduling scheme includes: If at least two regions initiate computing resource requests at the same time, the computing resource supply range is determined based on the computing resource type; Select target areas that provide computing resources within the computing resource supply range; Combine the request area and the target area with the bipartite graph matching algorithm to determine the computing power scheduling plan.

3. The computing resource scheduling method based on regional computing resource efficiency evaluation according to claim 2 is characterized in that: The method of determining a computing power scheduling plan includes: based on a bipartite graph constructed by matching computing power resource supply-type areas with computing power resource demand-type areas, edge weights are calculated by using the distance between areas and computing power capacity values ​​according to established latency and cost requirements, and augmented path queries are performed within the computing power demand and supply range to determine the optimal scheduling plan.

4. The computing resource scheduling method based on regional computing resource efficiency evaluation according to claim 1, characterized in that: Determining the computing resource type of each region includes: Determine multi-dimensional evaluation indicators and combine and weight them, where the multi-dimensional indicators include at least computing power infrastructure resources and computing power network application capabilities; Use the weighted weights to score the health of regional computing resources; Conduct multi-dimensional collaborative correlation analysis based on the scoring results and build a computing resource efficiency evaluation model; The computing power resource efficiency evaluation model is used to classify the computing power resources in each region.

5. The computing resource scheduling method based on regional computing resource efficiency evaluation according to claim 4 is characterized in that: The combined weighting method includes: Determine the objective weight of each evaluation indicator based on the entropy weight method; Determine the subjective weight of each evaluation indicator based on the analytic hierarchy process; A game theory mechanism is adopted to determine the final weight corresponding to each evaluation indicator based on the objective weight and the subjective weight.

6. The computing resource scheduling method based on regional computing resource efficiency evaluation according to any one of claims 1 to 5, characterized in that: The computing power demand types also include: hot data demand type; If a computing resource demanding area makes a request of the aforementioned hot data demand type, a computing resource scheduling solution that uses local or nearby resources and meets preset latency and cost requirements is adopted.

7. A computing resource scheduling device based on regional computing resource efficiency evaluation, characterized in that: The device comprises: A computing power resource evaluation module is used to determine the computing power resource type of each region using a multi-dimensional evaluation strategy; wherein the computing power resource type includes at least: computing power resource demand type and computing power resource supply type; The computing power demand classification module is used to classify the computing power demand of each region according to the frequency of data retrieval and the real-time retrieval demand. The computing power demand types include at least the preset warm data demand class and cold data demand class. The computing power resource scheduling module is used to match the computing power demand with several computing power resource supply areas within a preset range to obtain a computing power scheduling plan when a computing power resource demand area makes a computing power demand request and the request belongs to the warm data demand type or the cold data demand type.

8. An electronic device, characterized in that: include: a memory for storing computer readable instructions; as well as A processor is used to run the computer-readable instructions so that the electronic device executes the computing power resource scheduling method based on regional computing power resource efficiency evaluation as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium for storing computer-readable instructions, characterized in that: When the computer-readable instructions are executed by a processor, the processor executes the computing power resource scheduling method based on regional computing power resource efficiency evaluation as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, it implements the computing power resource scheduling method based on regional computing power resource efficiency evaluation as described in any one of claims 1 to 6.

Images