Intelligent computing power scheduling method and device, electronic equipment and storage medium
By combining genetic algorithms and multiple swarm scheduling methods, the problem of low network security in computing power networks was solved, achieving comprehensive optimization scheduling that balances security, cost, and load balancing, thereby improving the security and scheduling accuracy of computing power networks.
Patent Information
- Application Number
- CN202410956530.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-17
- Publication Date
- 2024-11-05
AI Technical Summary
Existing computing networks have low security and cannot accurately schedule resources when considering multiple dimensions such as security, cost, and load balancing, resulting in unreasonable resource scheduling.
A hybrid genetic algorithm is used for multi-objective optimization scheduling. By acquiring risky computing nodes in the computing power network, multiple populations are established, initialization processing and non-dominated sorting are performed, and genetic evolution processing is combined to determine the optimal solution set of the objectives. Safe computing nodes are determined according to preset priorities.
It improves the security and scheduling accuracy of the computing power network, prevents the slow convergence speed and local convergence problems of the genetic algorithm, and achieves optimal matching of secure computing power nodes.
Smart Images

Figure CN118897723A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computing power network technology, and in particular to a computing power intelligent scheduling method, device, electronic device and storage medium. Background Art
[0002] With the rapid development of computer networks, in order to improve the security and defense capabilities of computing power networks and ensure the safe operation of computing power networks, it is necessary to reasonably schedule computing power nodes. At present, the computing power resource scheduling scheme is mainly based on the user resource demand dimension. However, when other dimensional factors such as computing power cost and security capabilities need to be considered, its essence is a multi-objective optimization problem.
[0003] In related technologies, multi-objective optimization problems are usually converted into single-objective optimization problems by artificially presetting strategy factors to solve computing power scheduling solutions. However, since the weights cannot be accurately determined, the quality of the solution is difficult to guarantee, and security and cost are not comprehensively considered. Computing power nodes cannot be accurately scheduled, resulting in low security of the computing power network. Summary of the invention
[0004] The purpose of the embodiments of the present application is to provide a computing power intelligent scheduling method, device, electronic device and storage medium to solve the problem of low security of the computing power network.
[0005] To solve the above technical problems, the embodiments of the present application are implemented as follows:
[0006] In a first aspect, an embodiment of the present application provides a method for intelligent scheduling of computing power, comprising: obtaining risk computing power nodes in a computing power network, performing optimization scheduling modeling on the risk computing power nodes, and obtaining multi-dimensional scheduling indicators, wherein the scheduling indicators include: one or more of the security indicators, load balancing indicators, network performance indicators, and cost indicators corresponding to the risk computing power nodes, and the risk computing power nodes include one or more; based on a hybrid genetic algorithm, establishing multiple populations of the risk computing power nodes, initializing the multiple populations, and obtaining parallel parent populations with different control parameters; according to the scheduling indicators, performing non-dominated sorting and genetic evolution processing on the individuals in each of the parent populations to obtain an optimal solution set of the parent population; calculating the similarity of the optimal solution sets of each of the parent populations, determining a target optimal solution set based on the similarity, and according to a preset priority of the scheduling indicator, determining the safe computing power nodes in the target optimal solution set that meet the preset priority as the target computing power nodes.
[0007] In a second aspect, an embodiment of the present application provides an intelligent scheduling device for computing power, comprising: an acquisition module, used to acquire risk computing power nodes in a computing power network, perform optimization scheduling modeling on the risk computing power nodes, and obtain multi-dimensional scheduling indicators, wherein the scheduling indicators include: one or more of the security indicators, load balancing indicators, network performance indicators and cost indicators corresponding to the risk computing power nodes, and the risk computing power nodes include one or more; an initialization module, used to establish multiple populations of the risk computing power nodes based on a hybrid genetic algorithm, initialize the multiple populations, and obtain parallel parent populations with different control parameters; a non-dominated sorting module, used to perform non-dominated sorting and genetic evolution processing on the individuals in each of the parent populations according to the scheduling indicators, and obtain the optimal solution set of the parent population; a determination module, used to calculate the similarity of the optimal solution sets of each of the parent populations, determine the target optimal solution set according to the similarity, and determine the safe computing power nodes in the target optimal solution set that meet the preset priority according to the preset priority of the scheduling indicator as the target computing power nodes.
[0008] In a third aspect, an embodiment of the present application provides an electronic device, comprising a processor and a memory electrically connected to the processor, the memory storing a computer program, and the processor being used to call and execute the computer program from the memory to implement the above-mentioned intelligent computing power scheduling method.
[0009] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium for storing a computer program, wherein the computer program can be executed by a processor to implement the above-mentioned method for intelligent scheduling of computing power.
[0010] In a fifth aspect, an embodiment of the present application provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the above-mentioned intelligent computing power scheduling method.
[0011] In a sixth aspect, an embodiment of the present application provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the above-mentioned method for intelligent scheduling of computing power.
[0012] By adopting the technical solution of the embodiment of the present application, the risk computing power nodes in the computing power network are obtained, and the risk computing power nodes are optimized and scheduled modeled to obtain multi-dimensional scheduling indicators, wherein the scheduling indicators include: one or more of the security indicators, load balancing indicators, network performance indicators and cost indicators corresponding to the risk computing power nodes, and the risk computing power nodes include one or more; based on the hybrid genetic algorithm, multiple populations of risk computing power nodes are established, and the multiple populations are initialized to obtain parallel parent populations with different control parameters; according to the scheduling indicators, the individuals in each parent population are non-dominated sorting and genetic evolution processed to obtain the optimal solution set of the parent population; the similarity of the optimal solution sets of each parent population is calculated, and the target optimal solution set is determined according to the similarity, and according to the preset priority of the scheduling indicator, the security computing power nodes that meet the preset priority in the target optimal solution set are determined as the target computing power nodes. Among them, incorporating multi-dimensional scheduling indicators into the decision factors for scheduling computing nodes and conducting multi-objective resource optimization scheduling modeling can improve the security and accuracy of determining the target computing nodes; and by combining the genetic algorithm with multiple group queue competition and non-dominated sorting methods, it can prevent the problems of slow convergence and local convergence of the genetic algorithm. By combining multi-objective resource optimization scheduling modeling and hybrid genetic algorithms to determine the target computing nodes, comprehensively consider the safety performance of computing nodes, and achieve the optimal matching of secure computing nodes, the problem of low security of computing networks can be solved. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] In order to more clearly illustrate one or more embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in one or more embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0014] Figure 1 is a schematic block diagram of an improved hybrid genetic algorithm according to an embodiment of the present application;
[0015] Figure 2 is a schematic flow chart of a computing power intelligent scheduling method according to an embodiment of the present application;
[0016] Figure 3 is a schematic block diagram of the internal evolution of various populations according to an embodiment of the present application;
[0017] Figure 4 is a schematic flow chart of a computing power intelligent scheduling method according to another embodiment of the present application;
[0018] Figure 5is a schematic block diagram of a computing power intelligent scheduling device according to an embodiment of the present application;
[0019] Figure 6 It is a hardware structure diagram of a computing power intelligent scheduling device according to an embodiment of the present application. DETAILED DESCRIPTION
[0020] The embodiments of the present application provide a computing power intelligent scheduling method, device, electronic device and storage medium to solve the problem of low security of the computing power network.
[0021] In order to enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.
[0022] The computing power intelligent scheduling method provided in the embodiment of the present application can be executed by an electronic device, or by software installed in the electronic device. Specifically, the electronic device can be a terminal device or a server device. Among them, the terminal device can include a smart phone, a laptop computer, a smart wearable device, a vehicle terminal, etc., and the server device can include an independent physical server, a server cluster composed of multiple servers, or a cloud server capable of cloud computing.
[0023] Figure 1It is a schematic block diagram of an improved hybrid genetic algorithm according to an embodiment of the present application. It mainly includes initialization, evolution process and communication between populations through immigration operators. Among them, initialization includes initializing algorithm parameters, randomly initializing populations and queuing evolution of multiple populations to obtain multiple populations after initialization processing, such as population 1, population 2...population n; the evolution process includes genetic evolution processing inside each multiple population after initialization processing, specifically including selection operation, crossover operation and mutation operation on individuals in each population, and finally obtaining Pareto optimal solution set, each population corresponds to a Pareto optimal solution set, such as population 1 corresponds to Pareto optimal solution set 1, population 2 corresponds to Pareto optimal solution set 2, population n corresponds to Pareto optimal solution set n, and judge whether the similarity of the Pareto optimal solution set corresponding to each population is greater than the preset probability. If so, the Pareto target optimal solution set is output, if not, the iterative evolution is continued. The use of a multi-population parallel search mechanism for collaborative search and the realization of information exchange between multiple populations through immigration operators can improve the global search and local search capabilities of the hybrid genetic algorithm, avoid the premature convergence phenomenon that occurs during the evolution of a single population, and accurately obtain the target optimal solution set. According to the target optimal solution set, the target computing power node is determined and intelligent scheduling is performed.
[0024] In the following, in conjunction with the accompanying drawings, a computing power intelligent scheduling method provided by an embodiment of the present application is described in detail through specific embodiments and their application scenarios.
[0025] Figure 2 A schematic flow chart of a computing power intelligent scheduling method provided by an embodiment of the present invention is shown, and the method includes the following steps:
[0026] S202, obtaining risk computing power nodes in the computing power network, optimizing the scheduling modeling of the risk computing power nodes, and obtaining multi-dimensional scheduling indicators.
[0027] Among them, the scheduling indicators include: one or more of the security indicators, load balancing indicators, network performance indicators and cost indicators corresponding to the risk computing power nodes, and the risk computing power nodes include one or more.
[0028] In the computing power network, the computing power network is like an electric grid, and computing power is like electricity. In the world of artificial intelligence, autonomous driving, facial recognition, the games we play, the videos we watch, etc. can only be used normally with computing power.
[0029] Computing nodes refer to high-performance computing devices distributed across the country. They can be physical computers, virtual machines, or cloud services. These computing nodes have powerful processing and storage capabilities and can carry large-scale computing tasks and data processing needs.
[0030] Risky computing nodes include computing nodes with low security after security testing of the computing nodes. There are multiple computing nodes in a computing network, and there may be one or more risky computing nodes among the multiple computing nodes.
[0031] Optimize the scheduling model of risk computing nodes, including, in the scenario of network security protection, multi-dimensional calculation of risk computing node indicators, and determine them as multi-dimensional scheduling indicators of risk computing nodes. Scheduling indicators include: one or more of the security indicators, load balancing indicators, network performance indicators and cost indicators corresponding to risk computing nodes. In general, multiple indicators are considered together for multi-objective resource optimization scheduling.
[0032] Specifically, the optimization scheduling modeling of risk computing nodes refers to multi-objective computing resource scheduling. A risk computing node can be a computing instance, and each risk computing node corresponds to a virtual machine. Multi-objective computing resource scheduling is to efficiently allocate N computing instances on M computing nodes, that is, to determine the allocation of risk computing nodes in the computing nodes of the computing network. Among them, the resource demand of each computing instance i (i = 1, 2, ..., N) is expressed as R i , that is, the resource demand of the virtual machine in each risk computing node; the security score of each computing node j (j = 1, 2, ..., M) is expressed as S j The security score includes the score corresponding to the security performance of the computing power node when detecting the computing power node, and the available resource is A j , the available resources include the available resources of the virtual machines in each computing node, and the cost is C j The specific calculation and description are as follows:
[0033] (1) Security securitv The indicator calculation formula is expressed as:
[0034]
[0035] Among them, S i→j Represents the security score of computing node j assigned to computing instance i. After taking the reciprocal of the mean value of the security score, the smaller the value, the higher the overall security of the system.
[0036] (2) The calculation formula of load balancing LB index is expressed as:
[0037]
[0038] Among them, R i A represents the resource requirement of computing instance i. i→j Indicates the available resources of computing node j assigned to computing instance i. The standard deviation LB is used to measure the load of each computing node.ave Represents the average standard deviation. The smaller the coefficient of the standard deviation LB, the more balanced the load allocated to the risk computing power node.
[0039] (3) Network performance P network The indicator calculation formula is expressed as:
[0040]
[0041] Among them, D i→j Indicates the network delay from computing node j assigned to computing instance i to the user side. B i→j S represents the network bandwidth from computing node j allocated by computing instance i to the user side, i→j represents the network jitter from computing node j assigned to computing instance i to the user side, L i→j Indicates the packet loss rate from computing node j assigned to computing instance i to the user side. network A smaller value means better network performance.
[0042] (4) Cost C total The indicator calculation formula is expressed as:
[0043]
[0044] in, represents the computing performance cost of computing instance i assigned to computing node j, represents the network bandwidth cost of computing instance i allocated to computing node j, represents the storage resource cost of computing instance i allocated to computing node j. total The smaller the value, the more reasonable the cost.
[0045] In summary, the multi-objective resource optimization scheduling problem can be described as follows:
[0046] Minx
[0047]
[0048] Among them, x represents i→j, that is, the allocation result of computing instance i to computing node j, st represents the constraint condition, and R i A represents the resource requirement of computing instance i. j Represents the available resources of computing node j. When the constraint condition is met, the R i Less than A j , and at the same time meet the minimum of multiple indicator data, the allocation of computing instance i to computing node j is the optimal allocation solution.
[0049] By obtaining risk computing power nodes and calculating the multi-dimensional scheduling indicators of risk computing power nodes, and on the premise of meeting the resource requirements of computing instances, the target computing power nodes are determined from the computing power nodes of the computing power network through intelligent decision-making, which can achieve collaborative optimization of computing instances in terms of security, cost, network performance and load balancing.
[0050] S204, based on the hybrid genetic algorithm, multiple populations of risk computing power nodes are established, and the multiple populations are initialized to obtain parallel parent populations with different control parameters.
[0051] The hybrid genetic algorithm, including the genetic algorithm and the multi-objective optimization algorithm, constructs chromosomes that meet the constraints, randomly generates the initial population, calculates the fitness of each chromosome, generates sub-populations through replication, crossover and mutation operators, and outputs the search results if the termination conditions are met.
[0052] Based on the hybrid genetic algorithm, multiple populations related to risk computing nodes are randomly generated, and the multiple populations are initialized to obtain parent populations with different control parameters. Through the improved hybrid genetic algorithm, multiple populations are used to replace the original single population, and different control parameters are assigned to achieve different search purposes. Each population in the multiple populations can be initialized from different dimensions to obtain parent populations initialized in different dimensions. For example, parent population 1 is initialized for the security dimension, parent population 2 is initialized for the cost dimension, etc.
[0053] S206, performing non-dominated sorting and genetic evolution processing on the individuals in each parent population according to the scheduling index to obtain the optimal solution set of the parent population.
[0054] The individuals in the parent population include the mapping relationship between each computing instance and the allocated computing power node.
[0055] Non-dominated sorting processing includes dividing individuals into different levels according to their superiority and inferiority relationships in order to carry out further selection and evolution operations.
[0056] Genetic evolution, including optimization algorithms based on biological evolution theory and genetic principles, continuously produces new excellent individuals by simulating processes such as natural selection, crossover, and mutation, and gradually screens and improves them to obtain the final population.
[0057] Based on the scheduling indicators corresponding to the individuals in the parent populations with different control parameters mentioned above, non-dominated sorting is performed on the individuals in each parent population, that is, according to the multi-dimensional indicators of the individuals in each parent population, the individuals in the parent population are hierarchically sorted, and the individuals in the parent population are genetically evolved to obtain the optimal solution set of the parent population after the queue evolution process. Among them, the optimal solution set is the Pareto optimal solution set corresponding to the non-dominated sorting. The solutions in the Pareto optimal solution set are mutually non-dominated, and there is no domination relationship between them. As an example, among the individuals in the parent population, there is a risk computing power node, namely, risk computing power node 1. After non-dominated sorting and genetic evolution processing, the solution set is obtained as follows: the security index and load balancing index of the safety computing power node A corresponding to the risk computing power node 1 are optimal, and the network performance index and cost index are suboptimal. The network performance index and cost index of the safety computing power node B corresponding to the risk computing power node 1 are optimal, and the security index and load balancing index are suboptimal. The above two solutions do not dominate each other, so the solution set of the safety computing power node A corresponding to the risk computing power node 1 and the computing power node B corresponding to the risk computing power node 1 are determined as the optimal solution set of the parent population. If there are two risk computing nodes among the individuals in the parent population, namely, risk computing node 1 and risk computing node 2, after non-dominated sorting and genetic evolution processing, the solution set is: the security index and load balancing index of risk computing node 1 corresponding to security computing node A and risk computing node 2 corresponding to security computing node B are optimal, and the network performance index and cost index are suboptimal; the network performance index and cost index of risk computing node 1 corresponding to security computing node C and risk computing node 2 corresponding to security computing node D are optimal, and the security index and load balancing index are suboptimal. The above two solutions do not dominate each other. Then, risk computing node 1 corresponding to security computing node A and risk computing node 2 corresponding to security computing node B, risk computing node 1 corresponding to security computing node C and risk computing node 2 corresponding to security computing node D are determined as the optimal solution set of the parent population.
[0058] S208, calculate the similarity of the optimal solution sets of each parent population, determine the target optimal solution set based on the similarity, and according to the preset priority of the scheduling indicator, determine the security computing power node in the target optimal solution set that meets the preset priority as the target computing power node.
[0059] According to the optimal solution set of the parent population obtained in step 206, the similarity of the optimal solution sets of each parent population is calculated, and the target optimal solution set is determined according to the similarity. Specifically, if the similarity of the optimal solution sets of each parent population meets the preset probability, the iterative process of the genetic evolution of the parent population is terminated. Otherwise, return to step 206 for the next iteration. For example, when the similarity of the optimal solution sets of each parent population is greater than 90%, the optimal solution sets of each parent population are merged to obtain the target optimal solution set. If the similarity of the optimal solution sets of each parent population is less than or equal to 90%, the individuals in each parent population continue to perform optimization iterative calculations until the similarity of the optimal solution sets of each parent population meets the preset probability.
[0060] The preset priority of the scheduling index includes the method of using artificial preset scheme decision strategy, flexibly weighing the priority of the scheduling index according to the actual business needs, and selecting the target computing power node from the target optimal solution set generated by the multi-objective hybrid genetic algorithm, so as to achieve safe and efficient scheduling of computing power resources in different scenarios and different businesses. For example, in the scheduling index, it is considered that the preset priority is the most important for security, so the computing power node with the highest security index in the multi-dimensional is determined as the target computing power node, which can intelligently schedule the computing power nodes and improve the security of the computing power network. Following the example of step 206 above, the security index and load balancing index of the risk computing power node 1 corresponding to the security computing power node A are the best. Since the preset priority of the scheduling index is the most important for security, the security computing power node A with the best security index and load balancing index is determined as the target computing power node.
[0061] By adopting the technical solution of the embodiment of the present application, the risk computing power nodes in the computing power network are obtained, and the risk computing power nodes are optimized and scheduled modeled to obtain multi-dimensional scheduling indicators, wherein the scheduling indicators include: one or more of the security indicators, load balancing indicators, network performance indicators and cost indicators corresponding to the risk computing power nodes, and the risk computing power nodes include one or more; based on the hybrid genetic algorithm, multiple populations of risk computing power nodes are established, and the multiple populations are initialized to obtain parallel parent populations with different control parameters; according to the scheduling indicators, the individuals in each parent population are non-dominated sorting and genetic evolution processed to obtain the optimal solution set of the parent population; the similarity of the optimal solution sets of each parent population is calculated, and the target optimal solution set is determined according to the similarity, and according to the preset priority of the scheduling indicator, the security computing power nodes that meet the preset priority in the target optimal solution set are determined as the target computing power nodes. Among them, incorporating multi-dimensional scheduling indicators into the decision factors for scheduling computing nodes and conducting multi-objective resource optimization scheduling modeling can improve the security and accuracy of determining the target computing nodes; and by combining the genetic algorithm with multiple group queue competition and non-dominated sorting methods, it can prevent the problems of slow convergence and local convergence of the genetic algorithm. By combining multi-objective resource optimization scheduling modeling and hybrid genetic algorithms to determine the target computing nodes, comprehensively consider the safety performance of computing nodes, and achieve the optimal matching of secure computing nodes, the problem of low security of computing networks can be solved.
[0062] In one embodiment, to obtain the risk computing power node in the computing power network (i.e., S202), the following steps A1-A5 may be performed:
[0063] Step A1: Monitor and collect risk assessment data of each computing power node in the computing power network.
[0064] With the ubiquitous distribution of cloud-edge computing power, the security risks faced by network connections have become more complex. For grid-connected computing nodes, it is necessary to conduct equipment security risk assessments regularly to promptly discover and resolve potential risks and vulnerabilities. Therefore, it is necessary to timely or regularly monitor and record the risk assessment data of each computing node, and continuously monitor the safe operation status of each computing node.
[0065] Risk assessment data includes the collection of security protection construction status of each computing power node, such as access control, network security, data protection, monitoring and logs, and other comprehensive protection measures for computing power nodes, as well as the construction of security audits, emergency response and recovery mechanisms for computing power nodes; risk assessment status of each computing power node, such as regular risk assessment of grid-connected computing power node equipment from the dimensions of equipment security vulnerabilities, password security, platform website vulnerabilities, baseline configuration and security policy audits, and collection of assessment results; and security assessment status of each computing power node, such as security assessment based on the alarm and monitoring data of security equipment. The security protection construction status data, risk assessment status data and security assessment status data are determined as risk assessment data, and node security risks are dynamically monitored from these three dimensions.
[0066] Step A2: Determine the subjective weight and objective weight of the risk assessment data indicators based on the risk assessment data.
[0067] Subjective weights include using the analytic hierarchy process (AHP) to determine the importance of each security risk assessment data indicator based on expert experience, and constructing a judgment matrix, and determining the subjective weight through a series of steps such as the weight value, characteristic root and consistency test of each indicator. Specifically, the corresponding risk assessment data indicator is determined based on the risk assessment data obtained in step A1, wherein the risk assessment data indicator is determined by expert experience, the risk assessment data is scored based on the risk assessment data indicator, and the weight of the risk assessment data in the risk assessment data indicator is determined. The importance of each data in the computing power node is determined through the subjective weights of these risk assessment data.
[0068] Due to the complexity and professionalism of the security risk assessment of computing network nodes, the experience of experts in this field is particularly important. However, given the differences in the level of experts, which may be affected by subjective factors, an objective analysis is needed to make up for the differences in subjective factors between experts. Therefore, it is also necessary to use the objective weighting method CRITIC to calculate the objective weight. The specific calculation method of the objective weight includes: by calculating the comparison strength of the security risk assessment data indicators and the conflict between the security risk assessment data indicators, the objective weight of the security risk assessment data indicators is comprehensively measured. Among them, the comparison strength of the security risk assessment data indicators is reflected by calculating the standard deviation to reflect the difference in the values of the same security risk assessment data indicator between the evaluation schemes. The larger the standard deviation, the greater the difference in the values between the schemes, the more information is provided, and the greater the weight should be; while the conflict of the security risk assessment data indicators is reflected by the correlation coefficient between the security risk assessment data indicators. If the two security risk assessment data indicators are strongly positively correlated, it means that the conflict between the security risk assessment data indicators is weak, the amount of information provided is small, and the weight is also small.
[0069] Step A3: Weighted fusion of subjective weight and objective weight is performed to obtain the risk indicator weight corresponding to each computing power node.
[0070] The risk indicator weight includes the weighted fusion of subjective weight and objective weight by using the combined weighting method to obtain a reasonable and reliable risk indicator weight. The calculation formula for weighted fusion is as follows:
[0071]
[0072] Among them, W i is the risk indicator weight; α i is the subjective weight calculated by AHP, β i The objective weights calculated by the CRITIC weighting method.
[0073] Step A4: Determine the total risk assessment score of each computing power node based on the risk indicator weight and the score of the risk assessment data indicator.
[0074] According to the risk indicator weights and corresponding scores in the risk assessment data indicators corresponding to the security score data of each computing power node, a weighted sum is performed to determine the total risk assessment score of each computing power node. The summation formula is as follows:
[0075]
[0076] Among them, S is the total risk assessment score of the computing power node, W i is the risk indicator weight corresponding to the computing power node; X i is the score corresponding to indicator item i, and i is the risk assessment data indicator corresponding to the computing power node. That is, the total risk assessment score of the computing power node is determined based on the risk indicator weight of each computing power node and the score corresponding to the risk assessment data indicator of the computing power node.
[0077] Step A5: When the total risk assessment score is less than a preset threshold, the computing power nodes with a score less than the preset threshold are determined as risky computing power nodes.
[0078] According to the total risk assessment score of the computing power node determined in step A4, it is determined whether the total risk assessment score is less than a preset threshold. The preset threshold is the security risk threshold that is considered to be set. If the total risk assessment score is less than the preset threshold, an abnormal risk alarm is issued, and the computing power node is determined to be a risky computing power node. It is necessary to schedule computing power tasks to avoid risky computing power nodes.
[0079] In this embodiment, the risk assessment data of the computing power nodes in the computing power network are detected, and the risk indicator weights of the risk assessment data indicators corresponding to the risk assessment data are calculated, wherein the risk indicator weights are determined by the AHP-CRITIC combined weighting method, which can combine the subjective weights and the objective weights to obtain reasonable and reliable risk indicator weights, so that the risk assessment scores of the computing power nodes are more accurate and objective, thereby improving the accuracy of determining risk computing power nodes.
[0080] In one embodiment, based on a hybrid genetic algorithm, multiple populations of risk computing nodes are established, and the multiple populations are initialized to obtain parallel parent populations with different control parameters (i.e., S204), and the following steps B1-B4 may be performed:
[0081] Step B1: Based on a hybrid genetic algorithm, randomly generate multiple parallel populations of risk computing nodes.
[0082] Among them, the hybrid genetic algorithm can include a genetic algorithm and a multi-objective optimization algorithm; the parallel multiple populations are populations randomly generated about risk computing power nodes for determining the optimal solution set.
[0083] Step B2: Initialize each population in the multiple populations, assign different control parameters, and sort the individuals in each population according to the scheduling index to obtain an individual queue.
[0084] Each population in the multi-population randomly generated in step B1 is initialized and different control parameters are assigned, including: initializing the algorithm parameters of each population in the multi-population and randomly initializing the population. For example, n initial populations are randomly generated, that is, multi-populations. Then, the number of each population in the multi-population is n, each population randomly generates m individuals, the maximum evolutionary generation of the queue is T, and the crossover probability of each population in the multi-population is P c , the mutation probability is P m , the crossover and mutation probabilities are randomly generated within a specified range and can be different for each population.
[0085] The target rotation method is used to assign a target to each of the multiple populations. For example, population 1 can be assigned safety as a scheduling indicator for evolution, and population 2 can be assigned cost as a scheduling indicator for evolution, and so on. Then, according to the calculation formulas corresponding to the scheduling indicators of different dimensions, the objective function values corresponding to the scheduling indicators of different dimensions of individuals in the population are determined.
[0086] By obtaining the objective function value corresponding to the individual scheduling index in each population, the individuals in each population are sorted according to the size of the objective function value corresponding to the scheduling index. That is, based on the corresponding different dimensions in each population, the individuals in the population are sorted in ascending order according to the size of the corresponding objective function value to obtain an ordered individual queue.
[0087] Step B3: By performing evolutionary processing on the individuals in the individual queue, the individual objective function values in the individual queue are obtained, and based on the individual objective function values in the individual queue, the population after the evolutionary processing of the individual queue is determined.
[0088] The individuals in the individual queue are subjected to evolutionary processing, specifically including: according to the individual queue determined in step 3, the number of mutations of the individuals in the ordered individual queue is determined according to their positions in the queue, for example, the first individual mutates once, the second individual mutates twice, and so on, the mth individual mutates m times, and finally the objective function value after the m individuals mutate is calculated.
[0089] The offspring of each individual in the individual queue after mutation competes for survival with the parent before mutation, that is, compares the size of their objective function values, and retains the best ones as offspring to participate in the next queue status competition. Repeat the steps of sorting and queue evolution until the predetermined maximum queue evolution generation is reached, and the population after the individual queue evolution processing corresponding to the population is determined.
[0090] Step B4: The population after each individual queue evolution process is determined as the parent population with different control parameters.
[0091] In this embodiment, a multi-population evolution strategy is utilized to randomly generate multiple populations, and each of the multiple populations is initialized and queued for evolution to obtain a parent population after the queued evolution process. This can reduce the randomness of the algorithm and enable each of the multiple populations to evolve in a specific direction, thereby expanding the search space. This can prevent the algorithm from falling into a local optimum and accelerate the convergence of the algorithm.
[0092] In one embodiment, according to the scheduling index, the individuals in each parent population are subjected to non-dominated sorting processing (i.e., S206), and the following steps C1-C2 may be performed:
[0093] Step C1: Based on the individuals in each parent population and the scheduling indicators corresponding to the individuals in the parent population, the individuals in the parent population are subjected to non-dominated hierarchical sorting.
[0094] Specifically, the individuals in the parent population include the individuals in the parent population that have been encoded in the process of initializing multiple populations. When there are n risk computing nodes, it is equivalent to having n computing tasks and n virtual machines, and correspondingly there are m resource pools available. The individuals in the parent population represent a mapping relationship, that is, a mapping relationship from computing tasks to resource pools. For example, the individuals in the parent population contain i risk computing nodes, which are equivalent to i computing instances, and resource pool m corresponds to j computing nodes. Then, allocating risk computing node i to computing node j represents a mapping relationship.
[0095] It should be understood that the individuals in the parent population include one or more, and the mapping relationship can be determined according to the encoding, and the scheduling index of the computing instance allocated to the computing power node is calculated according to the mapping relationship. The calculated scheduling index includes security index, load balancing index, network performance index and cost index. Each of the indicators corresponds to an indicator function value.
[0096] Since the indicator function value determined by the indicator calculation formula for each dimension in the scheduling indicator is different, the individuals in the parent population can be non-dominated hierarchical sorting according to the indicator function value. Specifically, by calculating the indicator function value of each dimension of the individuals in the parent population, it is determined that among all dimensions, the individuals in the parent population with the best indicator function value are arranged in the first layer, and the individuals in this layer are not dominated, and they can dominate the individuals in the remaining layers; the individuals in the parent population with the second best indicator function value are arranged in the second layer, and the individuals in the second layer can be dominated by the individuals in the first layer, and can dominate the individuals after the second layer. The individuals in the parent population are all stratified according to the non-dominated hierarchical sorting method.
[0097] Step C2: For individuals at the same level in the parent population, sort the individuals at the same level in the parent population according to their crowding degree.
[0098] According to step C1, when the function values corresponding to the scheduling indicators calculated by individuals at the same level in the parent population are the same, then the same level also includes individuals from multiple parent populations. In this case, the individuals at the same level in the parent population are sorted according to the crowding degree of the individuals at the same level in the parent population. The greater the crowding degree, the sparser the individuals around it. In order to maintain the diversity of individuals in the population, individuals with greater crowding degree at the same level are preferentially selected for subsequent genetic evolution processing.
[0099] Among them, the crowding degree is used to measure the density of each individual in its non-dominated layer. The specific calculation method includes: traversing N individuals in the same level, calculating the value of the objective function in the M scheduling indicators of the N individuals, and obtaining an N*M matrix; traversing the M objective functions, that is, sorting the corresponding columns in N*M according to the values, and then traversing the N individuals in the sorting to obtain the crowding degree of each individual in this objective function. The crowding degree is the difference between the left and right neighbor values in the sorting. When the left and right neighbors are empty, the crowding degree is considered to be infinite. By traversing the individuals and accumulating multiple crowding degrees of the individuals, the accumulated crowding degree of the individuals is obtained, which is the crowding degree of the individuals.
[0100] In this embodiment, compared with the standard genetic algorithm, by using a fast non-dominated sorting algorithm within each parent population to perform non-dominated hierarchical sorting processing on the individuals in the parent population, the complexity of the calculation can be effectively reduced.
[0101] In one embodiment, genetic evolution is performed on individuals in each parent population to obtain the optimal solution set of the parent population (i.e., S206), and the following steps D1-D3 may be performed:
[0102] Step D1: Perform non-dominated sorting on the individuals in the parent population to obtain individuals of different levels in the parent population after non-dominated sorting.
[0103] Step D2: Randomly select individuals from the parent population, and perform selection, crossover and mutation operations on the individuals from the parent population according to their different levels to obtain child populations.
[0104] According to the tournament selection method, a certain number of parent population individuals are randomly selected, and the individuals with the highest fitness are selected from the parent population individuals for genetic operations to obtain the child population. Among them, the individuals with the highest fitness are the individuals with lower levels after the non-dominated sorting process. When the levels are the same, the individuals with the higher crowding degree are given priority. This process can greatly maintain the diversity of the population.
[0105] Genetic operations include selection, crossover, and mutation of parent population individuals, which can modify the mapping relationship of parent population individuals. When the encoding length of the parent population individual that needs genetic operation is n, n computing instances are migrated to m resource pools, and the latest population after migration is determined as the sub-population. For example, 3 computing instances are migrated to 5 resource pools, and the genotype is 152. 1 represents that the first computing instance can be moved to the resource pool numbered 1; 5 represents that the second computing instance can be moved to the resource pool numbered 5; 2 represents that the third computing instance can be moved to the resource pool numbered 2. In this correspondence, one gene code is equivalent to one individual. Through genetic crossover mutation, the original 152 is changed to 151 or 141, and finally a new gene code is generated as a sub-population.
[0106] Among them, the selection, crossover and mutation operations are all based on the probability when the population is initialized. By changing the gene, the mapping relationship of the parent population individuals is changed to obtain a new sub-population, which can make the value of the scheduling index corresponding to the new sub-population larger. It should be understood that the genotype of the parent population individual generated after the queue evolution process in step B4 above needs to be used as the parent genotype of each parent population for subsequent genetic evolution.
[0107] Step D3: Merge the parent population and the child population to obtain a merged population, and perform non-dominated sorting and crowding calculation on the merged population to screen out a new population with the same size as the parent population and the child population. Based on the new population, determine the optimal solution set corresponding to the parent population.
[0108] According to the elite strategy, the parent population and the child population are merged, and the parent population and the child population obtained in step D2 are merged to obtain a merged population. The number of individuals in the merged population is the sum of the parent population and the child population. By performing non-dominated sorting and crowding calculation on the individuals in the merged population, a new population is screened out so that the number of individuals in the new population is consistent with that of the parent population and the child population. Among them, by performing non-dominated sorting and crowding calculation on the merged population, the new population obtained can retain the best parent population individuals in the parent population and the excellent individuals in the child population after genetic evolution.
[0109] Specifically, the merged population includes all individuals in the parent population and the child population. For example, if the size of the parent population is N, the size of the child population after genetic evolution is also N, and the size of the merged population is 2N. The individuals in the merged population are sorted non-dominatedly. By sorting the individuals in the merged population in layers, the Pareto non-dominated layers F1, F2, ..., F n , and calculate the crowding degree of individuals in the merged population, and perform rank-first selection. If the ranks are the same, individuals with large crowding degree are given priority until the size of the merged population is equal to N, and the new population after non-dominated sorting is output as the latest generation of parent population; when there is no need for iterative optimization again, the new population after sorting can be used as the optimal solution set of the parent population corresponding to the parent population.
[0110] It should be noted that if Figure 3 The figure shows a schematic block diagram of the internal evolution of various populations. Within each population, one of the initialized parent populations is obtained, and the parent population is subjected to non-dominated sorting and crowding calculation processing to obtain the non-dominated sorted parent population. Parent population individuals in the non-dominated sorted parent population are randomly obtained for genetic evolution processing. The genetic evolution processing includes: obtaining a child population through selection, crossover, and mutation processing operations, merging the parent population and the child population, and performing non-dominated sorting and crowding calculation processing on the merged population to generate a new population. The new population is the optimal solution set of the parent population, that is, the Pareto optimal solution set of the parent population. The optimal solution set of the parent population is compared with the optimal solution sets of other parallel parent populations with different control parameters to determine whether the Pareto optimal solution set of the parent population is consistent with the Pareto optimal solution sets of other parent populations. If they are consistent, the internal evolution of the parent population is terminated. If not, the optimization iteration is continued from the genetic evolution processing of the parent population.
[0111] In this embodiment, by performing heritage evolution processing on the individuals of the parent population at different levels after non-dominated sorting, such as performing selection, crossover and mutation processing operations on the individuals of the parent population in the parent population, a child population is obtained, and the parent population and the child population are merged using an elite strategy to obtain a merged population. Based on the merged population, the individuals in the merged population are non-dominated sorted and crowdedness is calculated to screen out a new population with the same number of individuals as the parent population and the child population, and the new population is used as the optimal solution set of the parent population corresponding to the parent population. Multi-objective non-dominated sorting processing and genetic evolution processing are performed simultaneously within each parent population, and the non-dominated sorting processing can screen out excellent individuals in the parent population and the child population, which can improve the accuracy and efficiency of obtaining the optimal solution set of the parent population.
[0112] In one embodiment, the method may further perform the following step E:
[0113] Step E: According to the immigration operator, the best individual in the parent population including the best individual is introduced into the parent population not including the best individual for information exchange between the parent populations, wherein the best individual is the individual with the highest any one of the scheduling indicators.
[0114] The immigration operator includes introducing the best individuals that appear in the evolution process of each population into other populations regularly (at certain evolutionary generations). Specifically, the individuals in the population are sorted according to the non-dominated sorting stratification results and individual crowding, and the worst individuals in other populations are replaced by the best individuals in the original population.
[0115] In the multi-population genetic algorithm, there are multiple parent populations with different control parameters. Since the parent populations with different control parameters are relatively independent, they can be connected to each other through the immigration operator. The best individual corresponding to the parent population is migrated to the target parent population to replace the worst individual under this dimension, where the target parent population is the parent population that does not include the best individual, and the best individual is the individual with the highest any indicator in the scheduling indicator.
[0116] As an example, each evolution of the parent population with different control parameters will generate an optimal solution set. Every one or several generations of evolution, the best individual in parent population 1 can be replaced by the worst individual in parent population 2. The best individual between populations replaces the worst individual. Because non-dominated sorting processing will be performed, the worst individual in the obtained hierarchy will be replaced by the best individual in other populations including the best individual.
[0117] In this embodiment, information exchange between populations is achieved through the immigration operator, thereby achieving the effect of improving the evolution accuracy and efficiency between parent populations with different control parameters.
[0118] In one embodiment, the similarity of the optimal solution sets of each parent population is calculated, and the target optimal solution set is determined according to the similarity (i.e., 208), and the following steps F1-F2 may be performed:
[0119] Step F1: Determine the similarity of the optimal solution sets of each parent population according to the optimal solution sets of each parent population after the multi-population initialization process and the genetic evolution process;
[0120] According to the optimal solution sets of each parent population obtained after the initialization and genetic evolution of multiple populations, the similarity of the optimal solution sets of all parent populations is calculated, wherein the similarity of the optimal solution sets of all parent populations can be set by the user, and the higher the similarity, the more accurate the optimal solution set obtained. The similarity can be calculated according to the cosine similarity, which is not specifically limited here.
[0121] Step F2: When the similarity satisfies the preset probability, the optimal solution set of the parent population whose similarity satisfies the preset probability is determined as the target optimal solution set.
[0122] When the calculated similarity satisfies a preset probability, such as when the similarity of the optimal solution sets of all parent populations satisfies 90%, the optimal solution set of each parent population is determined as the target optimal solution set.
[0123] When the calculated similarity does not meet the preset probability, the latest generation of parent population corresponding to the optimal solution set of the parent population is iteratively optimized until the optimal solution set of the latest generation of parent population meets the preset probability.
[0124] In this embodiment, by obtaining the optimal solution sets of each parent population after the initialization processing of multiple populations, the similarity of the optimal solution sets of each parent population is determined, and when the similarity meets the preset probability, the optimal solution set of the parent population is determined as the target optimal solution set. The process of determining the target optimal solution set through multiple populations can control the number of iterations of the parent population, so that the target optimal solution set is obtained according to the optimal solution set of each parent population, and the computing power nodes are intelligently scheduled based on the target optimal solution set.
[0125] Figure 4 is a schematic flow chart of a computing power intelligent scheduling method according to another embodiment of the present application, such as Figure 4 As shown, the method comprises the following steps:
[0126] S401, monitor and collect risk assessment data of each computing power node in the computing power network, and calculate the subjective weight and objective weight of the risk assessment data indicators.
[0127] S402, weighting and summing the subjective weight and the objective weight to obtain the risk indicator weight corresponding to each computing power node.
[0128] S403, based on the risk indicator weight and the score of the risk assessment data indicator, determine the total risk assessment score of each computing power node, and determine the computing power node with a total risk assessment score less than a preset threshold as a risky computing power node.
[0129] S404, calculate the security index, load balancing index, network performance index and cost index corresponding to the risk computing power node to obtain a multi-dimensional scheduling index.
[0130] S405, based on the hybrid genetic algorithm, multiple populations of risk computing nodes are established, and algorithm parameters of the multiple populations and randomly initialized populations are initialized to obtain multiple parallel parent populations with different control parameters.
[0131] S406, performing non-dominated sorting processing on the individuals in each parent population according to the scheduling index.
[0132] Among them, the individuals at the same level in the parent population are sorted according to the crowding degree of the individuals at the same level.
[0133] S407, performing non-dominated sorting processing on the individuals in the parent population to obtain individuals of different levels of the parent population after the non-dominated sorting processing.
[0134] S408, randomly selecting individuals from the parent population, and performing selection, crossover and mutation processing operations on the individuals from the parent population according to different levels of the individuals from the parent population to obtain a child population.
[0135] S409, according to the elite strategy, the parent population and the child population are merged to obtain a merged population, and the merged population is subjected to non-dominated sorting and crowding calculation to obtain a new population with the same size as the parent population and the child population. Based on the new population, the optimal solution set corresponding to the parent population is determined.
[0136] S410, determining the similarity of the optimal solution sets of each parent population according to the optimal solution sets of each parent population after the multi-population initialization processing.
[0137] S411, when the similarity satisfies the preset probability, determining the optimal solution set of the parent population whose similarity satisfies the preset probability as the target optimal solution set.
[0138] S412, when the similarity does not meet the preset probability, jump to S309, perform the next generation iteration process until the similarity meets the preset probability, and take the optimal solution set of the parent population whose similarity meets the preset probability as the target optimal solution set.
[0139] Among them, the immigration operator can introduce the best individual in the parent population into the parent population that does not include the best individual, and is used for information exchange between parent populations with different control parameters.
[0140] S413, according to the preset priorities of the scheduling indicators, the security computing power nodes that meet the preset priorities in the target optimal solution are determined as the target computing power nodes.
[0141] The specific process from S401 to S413 has been described in detail in the above embodiment and will not be repeated here.
[0142] By adopting the technical solution of the embodiment of the present application, the risk computing power nodes in the computing power network are obtained, and the risk computing power nodes are optimized and scheduled modeled to obtain multi-dimensional scheduling indicators, wherein the scheduling indicators include: one or more of the security indicators, load balancing indicators, network performance indicators and cost indicators corresponding to the risk computing power nodes, and the risk computing power nodes include one or more; based on the hybrid genetic algorithm, multiple populations of risk computing power nodes are established, and the multiple populations are initialized to obtain parallel parent populations with different control parameters; according to the scheduling indicators, the individuals in each parent population are non-dominated sorting and genetic evolution processed to obtain the optimal solution set of the parent population; the similarity of the optimal solution sets of each parent population is calculated, and the target optimal solution set is determined according to the similarity, and according to the preset priority of the scheduling indicator, the security computing power nodes that meet the preset priority in the target optimal solution set are determined as the target computing power nodes. Among them, incorporating multi-dimensional scheduling indicators into the decision factors for scheduling computing nodes and conducting multi-objective resource optimization scheduling modeling can improve the security and accuracy of determining the target computing nodes; and by combining the genetic algorithm with multiple group queue competition and non-dominated sorting methods, it can prevent the problems of slow convergence and local convergence of the genetic algorithm. By combining multi-objective resource optimization scheduling modeling and hybrid genetic algorithms to determine the target computing nodes, comprehensively consider the safety performance of computing nodes, and achieve the optimal matching of secure computing nodes, the problem of low security of computing networks can be solved.
[0143] In summary, specific embodiments of the present subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recorded in the claims can be performed in a different order and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing can be advantageous.
[0144] The above is a computing power intelligent scheduling method provided in an embodiment of the present application. Based on the same idea, an embodiment of the present application also provides a computing power intelligent scheduling device.
[0145] Figure 5 Schematic diagram of a computing power intelligent scheduling device according to an embodiment of the present invention. Figure 5As shown, the computing power intelligent scheduling device includes: an acquisition module 51, an initialization module 52, a non-dominated sorting module 53, and a determination module 54:
[0146] The acquisition module 51 is used to acquire risk computing nodes in the computing network, optimize the scheduling modeling of the risk computing nodes, and obtain multi-dimensional scheduling indicators, wherein the scheduling indicators include: one or more of the security indicators, load balancing indicators, network performance indicators and cost indicators corresponding to the risk computing nodes, and the risk computing nodes include one or more;
[0147] Initialization module 52, used to establish multiple populations of risk computing power nodes based on a hybrid genetic algorithm, initialize the multiple populations, and obtain parallel parent populations with different control parameters;
[0148] The non-dominated sorting module 53 is used to perform non-dominated sorting and genetic evolution processing on the individuals in each parent population according to the scheduling index to obtain the optimal solution set of the parent population;
[0149] The determination module 54 is used to calculate the similarity of the optimal solution sets of each parent population, determine the target optimal solution set according to the similarity, and determine the security computing power node in the target optimal solution set that meets the preset priority as the target computing power node according to the preset priority of the scheduling indicator.
[0150] In one embodiment, the acquisition module 51 specifically includes:
[0151] A detection unit, used to monitor and collect risk assessment data of each computing power node in the computing power network;
[0152] A first determination unit, used to determine the subjective weight and objective weight of the risk assessment data indicator according to the risk assessment data;
[0153] The fusion unit is used to perform weighted fusion of subjective weights and objective weights to obtain the risk indicator weights corresponding to each computing power node;
[0154] The second determination unit is used to determine the total risk assessment score of each computing power node according to the risk indicator weight and the score of the risk assessment data indicator;
[0155] The third determination unit is used to determine the computing power nodes with a score less than the preset threshold as risk computing power nodes when the total risk assessment score is less than the preset threshold.
[0156] In one embodiment, the initialization module 52 specifically includes:
[0157] A random unit, used to randomly generate multiple parallel populations of risk computing nodes based on a hybrid genetic algorithm;
[0158] An initialization processing unit is used to initialize each population in the multiple populations, allocate different control parameters, and sort the individuals in each population according to the scheduling index to obtain an individual queue;
[0159] An evolution unit, used for obtaining individual objective function values of the individual queues by performing evolutionary processing on the individuals in the individual queues, and determining a population of the individual queues after the evolutionary processing based on the individual objective function values;
[0160] The fourth determination unit is used to determine the population after each individual queue evolution process as a parent population with different control parameters.
[0161] In one embodiment, the non-dominated sorting module 53 is specifically used to perform non-dominated hierarchical sorting processing on the individuals in the parent population based on the scheduling index corresponding to the individuals in each parent population;
[0162] For individuals at the same level in the parent population, they are sorted according to their crowding degree.
[0163] In one embodiment, the non-dominated sorting module 53 is further configured to perform non-dominated sorting on the individuals in the parent population to obtain individuals in the parent population at different levels after the non-dominated sorting.
[0164] Randomly select individuals from the parent population, and perform selection, crossover and mutation operations on the individuals from the parent population according to their different levels to obtain sub-populations;
[0165] The parent population and the child population are merged to obtain a merged population, and the merged population is subjected to non-dominated sorting and crowding calculation to obtain a new population with the same size as the parent population and the child population. Based on the new population, the optimal solution set corresponding to the parent population is determined.
[0166] In one embodiment, the device is also used to introduce the best individual from the parent population including the best individual into the parent population not including the best individual according to the immigration operator, for information exchange between the parent populations, wherein the best individual is the individual with the highest any one of the scheduling indicators.
[0167] By adopting the technical solution of the embodiment of the present application, the risk computing power nodes in the computing power network are obtained, and the risk computing power nodes are optimized and scheduled modeled to obtain multi-dimensional scheduling indicators, wherein the scheduling indicators include: one or more of the security indicators, load balancing indicators, network performance indicators and cost indicators corresponding to the risk computing power nodes, and the risk computing power nodes include one or more; based on the hybrid genetic algorithm, multiple populations of risk computing power nodes are established, and the multiple populations are initialized to obtain parallel parent populations with different control parameters; according to the scheduling indicators, the individuals in each parent population are non-dominated sorting and genetic evolution processed to obtain the optimal solution set of the parent population; the similarity of the optimal solution sets of each parent population is calculated, and the target optimal solution set is determined according to the similarity, and according to the preset priority of the scheduling indicator, the security computing power nodes that meet the preset priority in the target optimal solution set are determined as the target computing power nodes. Among them, incorporating multi-dimensional scheduling indicators into the decision factors for scheduling computing nodes and conducting multi-objective resource optimization scheduling modeling can improve the security and accuracy of determining the target computing nodes; and by combining the genetic algorithm with multiple group queue competition and non-dominated sorting methods, it can prevent the problems of slow convergence and local convergence of the genetic algorithm. By combining multi-objective resource optimization scheduling modeling and hybrid genetic algorithms to determine the target computing nodes, comprehensively consider the safety performance of computing nodes, and achieve the optimal matching of secure computing nodes, the problem of low security of computing networks can be solved.
[0168] Those skilled in the art should understand that Figure 5 The computing power intelligent scheduling device in can be used to implement the computing power intelligent scheduling method described above, and the detailed description thereof should be similar to the description of the method part above. To avoid tediousness, it will not be repeated here.
[0169] Based on the same technical concept, an embodiment of the present application further provides an electronic device, which is used to execute the above-mentioned computing power intelligent scheduling method. Figure 6 A schematic diagram of the structure of an electronic device for implementing various embodiments of the present application. The electronic device may have relatively large differences due to different configurations or performances, and may include a processor (processor) 610, a communication interface (Communications Interface) 620, a memory (memory) 930 and a communication bus 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call a computer program stored in the memory 630 and executable on the processor 610 to perform the following steps:
[0170] Obtain risk computing nodes in the computing power network, optimize the scheduling modeling of the risk computing nodes, and obtain multi-dimensional scheduling indicators, wherein the scheduling indicators include: one or more of the security indicators, load balancing indicators, network performance indicators and cost indicators corresponding to the risk computing nodes, and the risk computing nodes include one or more;
[0171] Based on the hybrid genetic algorithm, multiple populations of risk computing nodes are established, and the multiple populations are initialized to obtain parallel parent populations with different control parameters.
[0172] According to the scheduling index, the individuals in each parent population are processed by non-dominated sorting and genetic evolution to obtain the optimal solution set of the parent population;
[0173] Calculate the similarity of the optimal solution sets of each parent population, determine the target optimal solution set based on the similarity, and according to the preset priority of the scheduling indicator, determine the security computing power nodes in the target optimal solution set that meet the preset priority as the target computing power nodes.
[0174] By adopting the technical solution of the embodiment of the present application, the risk computing power nodes in the computing power network are obtained, and the risk computing power nodes are optimized and scheduled modeled to obtain multi-dimensional scheduling indicators, wherein the scheduling indicators include: one or more of the security indicators, load balancing indicators, network performance indicators and cost indicators corresponding to the risk computing power nodes, and the risk computing power nodes include one or more; based on the hybrid genetic algorithm, multiple populations of risk computing power nodes are established, and the multiple populations are initialized to obtain parallel parent populations with different control parameters; according to the scheduling indicators, the individuals in each parent population are non-dominated sorting and genetic evolution processed to obtain the optimal solution set of the parent population; the similarity of the optimal solution sets of each parent population is calculated, and the target optimal solution set is determined according to the similarity, and according to the preset priority of the scheduling indicator, the security computing power nodes that meet the preset priority in the target optimal solution set are determined as the target computing power nodes. Among them, incorporating multi-dimensional scheduling indicators into the decision factors for scheduling computing nodes and conducting multi-objective resource optimization scheduling modeling can improve the security and accuracy of determining the target computing nodes; and by combining the genetic algorithm with multiple group queue competition and non-dominated sorting methods, it can prevent the problems of slow convergence and local convergence of the genetic algorithm. By combining multi-objective resource optimization scheduling modeling and hybrid genetic algorithms to determine the target computing nodes, comprehensively consider the safety performance of computing nodes, and achieve the optimal matching of secure computing nodes, the problem of low security of computing networks can be solved.
[0175] The specific execution steps can refer to the various steps of the above-mentioned computing power intelligent scheduling method embodiment, and can achieve the same technical effect. To avoid repetition, they will not be repeated here.
[0176] It should be noted that the electronic devices in the embodiments of the present application include: servers, terminals, or other devices except terminals.
[0177] The above electronic device structure does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. For example, the input unit may include a graphics processing unit (GPU) and a microphone, and the display unit may be configured with a display panel in the form of a liquid crystal display, an organic light-emitting diode, etc. The user input unit includes a touch panel and at least one of other input devices. The touch panel is also called a touch screen. Other input devices may include, but are not limited to, a physical keyboard, function keys (such as volume control buttons, switch buttons, etc.), a trackball, a mouse, and a joystick, which will not be repeated here.
[0178] The memory can be used to store software programs and various data. The memory may mainly include a first storage area for storing programs or instructions and a second storage area for storing data, wherein the first storage area may store an operating system, an application program or instructions required for at least one function (such as a sound playback function, an image playback function, etc.), etc. In addition, the memory may include a volatile memory or a non-volatile memory, or the memory may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct memory bus random access memory (DRRAM).
[0179] The processor may include one or more processing units; optionally, the processor integrates an application processor and a modem processor, wherein the application processor mainly processes operations related to the operating system, user interface, and application programs, and the modem processor mainly processes wireless communication signals, such as a baseband processor. It is understandable that the modem processor may not be integrated into the processor.
[0180] An embodiment of the present application also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, each process of the above-mentioned embodiment of the method for intelligent scheduling of computing power is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.
[0181] The processor is a processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.
[0182] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned embodiment of the computing power intelligent scheduling method, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.
[0183] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.
[0184] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the processor is used to run the program or instructions to implement the various processes of the above-mentioned product recommendation method embodiment and can achieve the same technical effect. To avoid repetition, it will not be repeated here.
[0185] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.
[0186] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.
[0187] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.
Claims
1. A computing power intelligent scheduling method, characterized in that: The method comprises: Acquire risk computing nodes in the computing power network, perform optimization scheduling modeling on the risk computing nodes, and obtain multi-dimensional scheduling indicators, wherein the scheduling indicators include: one or more of a security indicator, a load balancing indicator, a network performance indicator, and a cost indicator corresponding to the risk computing node, and the risk computing node includes one or more; Based on a hybrid genetic algorithm, multiple populations of risk computing nodes are established, and the multiple populations are initialized to obtain parallel parent populations with different control parameters; According to the scheduling index, the individuals in each parent population are subjected to non-dominated sorting processing and genetic evolution processing to obtain an optimal solution set of the parent population; Calculate the similarity of the optimal solution sets of each parent population, determine the target optimal solution set according to the similarity, and according to the preset priority of the scheduling indicator, determine the security computing power node in the target optimal solution set that meets the preset priority as the target computing power node.
2. The method according to claim 1, characterized in that The step of obtaining risky computing power nodes in the computing power network includes: Monitor and collect risk assessment data of each computing power node in the computing power network; Determining subjective weights and objective weights of risk assessment data indicators based on the risk assessment data; The subjective weight and the objective weight are weighted and integrated to obtain the risk indicator weight corresponding to each computing power node; Determine the total risk assessment score of each computing power node according to the risk indicator weight and the score of the risk assessment data indicator; When the total risk assessment score is less than a preset threshold, the computing power node with a score less than the preset threshold is determined as the risky computing power node.
3. The method according to claim 1, characterized in that: The method of establishing multiple populations of risk computing nodes based on a hybrid genetic algorithm and initializing the multiple populations to obtain parallel parent populations with different control parameters includes: Based on the hybrid genetic algorithm, randomly generate the multiple populations in parallel about the risk computing power node; Initializing each of the multiple populations, assigning the different control parameters, and sorting the individuals in each population according to the scheduling index to obtain an individual queue; By performing evolutionary processing on the individuals in the individual queue, an objective function value of the individuals in the individual queue is obtained, and based on the individual objective function value, a population of the individual queue after the evolutionary processing is determined; The population after each of the individuals is queued for evolution is determined as the parent population of the different control parameters.
4. The method according to claim 1, characterized in that: The performing non-dominated sorting processing on the individuals in each of the parent populations according to the scheduling index includes: Based on the scheduling index corresponding to each individual in the parent population, the individuals in the parent population are subjected to non-dominated hierarchical sorting processing; For individuals at the same level in the parent population, the individuals at the same level in the parent population are sorted according to the crowding degree of the individuals at the same level.
5. The method according to claim 4, characterized in that The performing genetic evolution processing on the individuals in each of the parent populations to obtain the optimal solution set of the parent populations includes: By performing non-dominated sorting processing on the individuals in the parent population, individuals of the parent population at different levels after the non-dominated sorting processing are obtained; Randomly selecting individuals of the parent population, and performing selection, crossover and mutation processing operations on the individuals of the parent population according to different levels of the individuals of the parent population to obtain a child population; The parent population and the child population are merged to obtain a merged population, and non-dominated sorting and crowding calculation are performed on the merged population to screen out a new population with the same size as the parent population and the child population, and based on the new population, an optimal solution set corresponding to the parent population is determined.
6. The method according to claim 1, characterized in that The method further comprises: According to the immigration operator, the best individual in the parent population including the best individual is introduced into the parent population not including the best individual for information exchange between the parent populations, wherein the best individual is the individual with the highest any one of the scheduling indicators.
7. A computing power intelligent scheduling device, characterized in that: The device comprises: An acquisition module is used to acquire risk computing nodes in the computing network, optimize the scheduling modeling of the risk computing nodes, and obtain multi-dimensional scheduling indicators, wherein the scheduling indicators include: one or more of the security indicators, load balancing indicators, network performance indicators and cost indicators corresponding to the risk computing nodes, and the risk computing nodes include one or more; An initialization module, used to establish multiple populations of risk computing power nodes based on a hybrid genetic algorithm, initialize the multiple populations, and obtain parallel parent populations with different control parameters; A non-dominated sorting module is used to perform non-dominated sorting and genetic evolution processing on the individuals in each parent population according to the scheduling index to obtain an optimal solution set of the parent population; A determination module is used to calculate the similarity of the optimal solution sets of each parent population, determine the target optimal solution set according to the similarity, and according to the preset priority of the scheduling indicator, determine the security computing power node in the target optimal solution set that meets the preset priority as the target computing power node.
8. An electronic device, characterized in that: It includes a processor and a memory electrically connected to the processor, the memory stores a computer program, and the processor is used to call and execute the computer program from the memory to implement a computing power intelligent scheduling method as described in any one of claims 1-6.
9. A computer-readable storage medium, characterized in that: The storage medium is used to store a computer program, and the computer program can be executed by a processor to implement a computing power intelligent scheduling method as described in any one of claims 1-6.
10. A computer program product, characterized in that It includes a computer program, which, when executed by a processor, implements a computing power intelligent scheduling method as described in any one of claims 1 to 6.
Citation Information
Cited By
Distributed data asset storage method and system
CN119760710A
Container instance migration method and device, electronic equipment and storage medium
CN120803619A
