A one-click data center intelligent layout method and system based on particle swarm algorithm

Abstract

The invention relates to the technical field of database detection, and discloses a one-key data center intelligent layout method based on a particle swarm algorithm. The method comprises: constructing a symmetric matrix of a data link set according to the connectivity relationship and weight of each device node in the data center, and representing it through an AOE directed acyclic graph network; dividing the data link set into a plurality of cluster center data sets through a clustering algorithm; respectively correcting the plurality of cluster center data sets in combination with the AOE directed acyclic graph network to obtain corresponding cluster family sub-data link sets; constructing a corresponding membership function according to the balance index of each cluster family sub-data link set; obtaining the common energy consumption influencing parameters of each cluster family sub-data link set, and establishing a multi-objective optimized sub-virtual machine scheduling model in combination with constraint conditions; solving the multi-objective optimization problem based on an improved particle swarm algorithm to obtain an optimal sub-solution, which is the corresponding virtual machine scheduling scheme.

Description

Technical Field

[0001] The present invention relates to the technical field of cloud computing and machine learning, and in particular to a one-click data center intelligent layout method based on a particle swarm algorithm. Background Art

[0002] In current data center management, intelligent layout and resource optimization scheduling are the key to improving efficiency and reducing costs. With the rapid development of cloud computing, big data and artificial intelligence technologies, data centers are facing the challenge of a sharp increase in processing power requirements. The application of these technologies forces data centers to optimize their resource allocation to ensure efficient data processing and minimize energy consumption. In order to achieve this goal, advanced algorithms and strategies are needed for data center layout and resource scheduling.

[0003] Traditional data center management mainly relies on experience and simple rules to allocate resources. This method is often inadequate when dealing with complex network structures and dynamically changing load requirements. For example, virtual machine allocation and migration decisions are often based on preset rules and cannot reflect the current status and future trends of the data center in real time. In addition, due to the lack of an effective global management perspective, traditional methods also make it difficult to optimize the configuration of device nodes within the data center, resulting in low resource utilization and serious energy waste.

[0004] In addition, the existing virtual machine scheduling methods for cloud data centers have problems such as high energy consumption, load imbalance, and low resource utilization. Although there are many algorithms that can reduce the energy consumption of cloud centers to a certain extent, there are still some shortcomings. On the one hand, with the continuous updating of equipment and the expansion of scale, the existing virtual machine scheduling methods for mobile cloud computing cannot meet the current needs and cannot meet the large-scale virtual machine scheduling needs; on the other hand, most of the existing virtual machine scheduling strategies are not comprehensive, or only consider different angles from one aspect, such as only considering the CPU energy consumption without considering the energy consumption of other components. Summary of the invention

[0005] In view of this, the present invention provides a one-click data center intelligent layout method based on a particle swarm algorithm, the purpose of which is to: 1) propose a one-click data center intelligent layout method based on a particle swarm algorithm, which can effectively improve the automation and intelligence level of data center scheduling construction through the construction of a data chain; 2) adopt a local anomaly recognition index method to capture anomalies in a data set, which can more effectively identify local anomalies and improve the accuracy of anomaly detection; 3) solve the multi-objective optimization problem by improving the particle swarm algorithm to obtain the optimal sub-solution, which is the corresponding virtual machine scheduling solution, fuse the sub-models of different data sets, and use the information fusion algorithm to synthesize the optimal sub-solutions of different data sets to construct a detection chain for the sub-detection models of different data sets, so as to achieve the optimal detection method of the global final virtual machine scheduling solution, realize real-time updating of detection parameters, so as to adapt to the dynamic changes of structural data in the data set, maintain the effectiveness of the detection model, and dynamically adjust it as the data changes.

[0006] To achieve the above object, the present invention provides a one-click data center intelligent layout method based on particle swarm algorithm, comprising the following steps:

[0007] S1: According to the connectivity and weight of each device node in the data center, a symmetric matrix of the data link set is constructed and represented by an AOE directed acyclic graph network;

[0008] S2: Divide the data link set into multiple cluster center data sets through clustering algorithm;

[0009] S3: Combine the AOE directed acyclic graph network to modify multiple cluster center data sets respectively to obtain the corresponding cluster family sub-data link sets;

[0010] S4: construct the corresponding membership function according to the balance index of each cluster family sub-data link set;

[0011] S5: Obtain the common energy consumption influencing parameters of each cluster group sub-data link set, combined with the constraint conditions,

[0012] Establish a sub-virtual machine scheduling model with multi-objective optimization;

[0013] S6: Solve the multi-objective optimization problem based on the improved particle swarm algorithm and obtain the optimal sub-solution.

[0014] The optimal sub-solution is the corresponding virtual machine scheduling solution;

[0015] S7: Fuse the sub-models of different data sets, use the information fusion algorithm to integrate the optimal sub-solutions of different data sets to construct a detection chain for the sub-detection models of different data sets, and achieve the optimal detection of the final global virtual machine scheduling solution.

[0016] As a further improvement method of the present invention:

[0017] Optionally, in step S1, according to the connectivity relationship and weights of each device node in the data center, a symmetric matrix of the data link set is constructed and represented by an AOE directed acyclic graph network, including:

[0018] S11: For each data node I in the data center i , through the data transmission program, determine whether there is a connected path from the sending point to the data receiving point, where i is the node serial number, 0 < i ≤ N, and N is the total number of nodes;

[0019] S12: Use 1 to represent the nodes with a connected path, and use 0 to represent the nodes without a connected path, and construct a symmetric matrix B of order N;

[0020] S13: Screen the transmission routes passed by the connected paths according to the symmetric matrix B, and construct an AOE directed acyclic graph network in combination with the weights.

[0021] As a further improvement method of the present invention:

[0022] Optionally, in step S3, the AOE directed acyclic graph network is combined to correct multiple clustering centers respectively to obtain corresponding clustering family sub-data link sets, including:

[0023] S31: Obtain the neighborhood set P(pi) = [p1... pk] of each clustering center data set, where k is the number of nodes in the neighborhood set and o is the clustering center;

[0024] S32: Calculate the connected distance Lp-o between the data point pi in the neighborhood set and the clustering center o point, and calculate the connected distance Lp-p' between the data point pi in the neighborhood set and the adjacent point pi';

[0025] S33: Calculate the local outlier recognition index LOF p-p'(pi) of the data point pi, and the calculation formula is:

[0026]

[0027] Where:

[0028] P(pi) represents the neighborhood set of the data point pi;

[0029] Lp-p' represents the distance from the data point pi to its nearest neighbor point;

[0030] Lp-o represents the distance from the data point pi to the clustering center o;

[0031] LOF p-p′(pi) represents the local outlier recognition index of the data point pi;

[0032] S34: If the local anomaly recognition index exceeds the specified threshold, the data point pi is judged as an abnormal data point and is not added to the corresponding cluster family sub-data link set; if the local anomaly recognition index does not exceed the specified threshold, the data point pi is judged as a non-abnormal data point and is added to the corresponding cluster family sub-data link set.

[0033] As a further improvement method of the present invention:

[0034] Optionally, in step S4, a corresponding membership function is constructed according to the balance index of each cluster family sub-data link set, including:

[0035] S41: If a connected path exists between every two vertices in the AOE directed acyclic graph network, determine the consumption cost p corresponding to the edge between every two vertices according to the weight g(i, j) between every two storage nodes. cov ;

[0036] Wherein, if a connected path between every two vertices in the AOE directed acyclic graph network exists, the corresponding occupancy coefficient p(i, j) is 1; if a connected path between every two vertices in the AOE directed acyclic graph network does not exist, the corresponding occupancy coefficient p(i, j) is 0;

[0037] S42: Calculate the cost p of the edge between each two vertices cov ,include:

[0038] Determine the corresponding sub-symmetric matrix Z according to the cluster family sub-data link set m×n , where the position of each site is s ij , the calculation formula is:

[0039] s ij =(x ij ,y ij );

[0040]

[0041] Among them, m and n are the sub-symmetric matrices Z m×n The number of rows and columns;

[0042] S43: According to the consumption cost p cov Determine the type of sub-model and construct the corresponding evaluation function, including:

[0043]

[0044]

[0045] Among them, A1 and A2 are the types of models;

[0046] S44: Determine the type of each sub-model according to the evaluation function, and then process the multi-objective function to optimize each sub-model, including: if the sub-model belongs to the A1 type, increase the importance of each indicator of the sub-model; if the sub-model belongs to the A2 type, reduce the importance of each indicator of the sub-model.

[0047] As a further improvement method of the present invention:

[0048] Optionally, in step S6, solving the multi-objective optimization problem based on an improved particle swarm algorithm to obtain an optimal sub-solution includes:

[0049] S61: Initialize the population and its parameters, including population size, particle dimension, number of iterations, indicator membership, indicator weight, determine the encoding method, use binary encoding, and represent the physical host where each virtual machine is placed as a binary string;

[0050] S62: Optimizing the system objective function using the evaluation function;

[0051] S63: Calculate the fitness function value of the particle corresponding to the path through the statistical information of the link, and calculate the aggregation degree of the current particle from the fitness value;

[0052] S64: Calculating the individual historical optimal solution of each particle according to the aggregation degree;

[0053] S65: Take the position in the particle swarm as a candidate solution, and select a part of the particles to generate the next generation of population and particle swarm according to the similarity between the candidate solution and the individual historical optimal solution;

[0054] S66: For multi-objective problems, the fitness value of each particle is calculated using a multi-objective fitness function;

[0055] S67: Compare the fitness value of the current particle with the optimal fitness value of the individual particle of the previous generation. If the fitness value of the current particle is less than the optimal fitness value of the individual particle of the previous generation, update the optimal solution of the individual particle, otherwise do not update.

[0056] S68: Determine whether the termination condition is met: Determine whether the algorithm has converged based on whether the updated particle individuals are stable; if it has converged.

[0057] Particle swarm optimization (PSO) is an evolutionary computing technology that finds the optimal solution by simulating information exchange and coordination between individuals in a group. The algorithm simulates birds in a flock by designing a massless particle. The particle has only two properties: speed and position. The speed represents the speed of movement, and the position represents the direction of movement. Each particle searches for the optimal solution in the search space alone, records it as the current individual extreme value, and shares the individual extreme value with other particles in the entire particle swarm. The best individual extreme value is found as the current global optimal solution of the entire particle swarm. All particles in the particle swarm adjust their speed and position according to the current individual extreme value they find and the current global optimal solution shared by the entire particle swarm.

[0058] The PSO algorithm regards the solution space as a multidimensional space, and each point is called a particle. The main steps of the algorithm include group initialization, fitness value evaluation, updating of historical optimal position and global optimal position, and updating of speed and position.

[0059] Among them, in the particle swarm optimization algorithm, particles update their speed and position through the following two formulas:

[0060]

[0061] x i =x i +v i ;

[0062] Where: i = 1, 2, ..., K, K is the total number of particles in the particle group; v i is the speed of the particle; rand() is a random number between (0,1); x i is the current position of the particle; c 1 With c 2 is the learning factor; v i The maximum value of v max , when v i >v max When,v i =v max .

[0063] As a further improvement method of the present invention:

[0064] Optionally, in step S6, the multi-objective optimization problem is solved based on the improved particle swarm algorithm to obtain an optimal sub-solution, which is the corresponding virtual machine scheduling solution, including:

[0065] S61: Obtaining the objective function of each sub-model;

[0066] S62: Perform crossover and mutation operations on each sub-model data set based on the improved particle swarm algorithm;

[0067] S63: sort each objective function, and then calculate a new speed and a new position respectively;

[0068] S64: Based on the fact that each particle position corresponds to a flow scheduling scheme, the particle position corresponds to the scheduling path of the network flow in the flow scheduling scheme, and the best path is obtained, that is, the optimal sub-solution and the corresponding virtual machine scheduling scheme are obtained.

[0069] As a further improvement method of the present invention:

[0070] Optionally, in step S7, the sub-models of different data sets are fused, and the optimal sub-solutions of different data sets are integrated with the information fusion algorithm to construct a detection chain for the sub-detection models of different data sets, so as to achieve optimal detection of the global final virtual machine scheduling solution, including:

[0071] S71: Fusing the sub-model data sets of different data sets to obtain a fused matrix data set again, and constructing a fused AOE directed acyclic graph network;

[0072] S72: according to the fused AOE directed acyclic graph network, taking the sum of the weights corresponding to all the edges in each feasible link as the consumption cost corresponding to each feasible link;

[0073] S73: Obtain the optimal link according to the consumption cost corresponding to each feasible link, the CPU utilization rate and the transmission load of each feasible link;

[0074] S74: Based on the node path on the optimal link, final global virtual machine scheduling is implemented.

[0075] As a further improvement method of the present invention:

[0076] Optionally, the present invention further provides a one-key data center intelligent layout method system based on a particle swarm algorithm, comprising:

[0077] Original data set module: According to the connectivity and weight of each device node in the data center, a symmetric matrix of the data link set is constructed and represented by an AOE directed acyclic graph network;

[0078] Sub-dataset module: divides the data link set into multiple cluster center data sets through clustering algorithm;

[0079] Screening module: Combined with the AOE directed acyclic graph network, multiple cluster center data sets are corrected to obtain the corresponding cluster family sub-data link set;

[0080] Evaluation module: construct the corresponding membership function according to the balance index of each cluster group sub-data link set;

[0081] Correction module: obtain the common energy consumption influencing parameters of each cluster family sub-data link set, and establish a multi-objective optimization sub-virtual machine scheduling model based on constraint conditions;

[0082] Optimization module: solves the multi-objective optimization problem based on the improved particle swarm algorithm to obtain the optimal sub-solution, which is the corresponding virtual machine scheduling solution;

[0083] Detection module: fuses the sub-models of different data sets, uses the information fusion algorithm to integrate the optimal sub-solutions of different data sets to build a detection chain for the sub-detection models of different data sets, and realizes the optimal detection of the final global virtual machine scheduling solution.

[0084] In order to solve the above problem, the present invention provides an electronic device, the electronic device comprising:

[0085] A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the one-key data center intelligent layout method based on particle swarm algorithm when executing the computer program.

[0086] In order to solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the steps of the one-click data center intelligent layout method based on a particle swarm algorithm.

[0087] First, this scheme proposes a one-click data center intelligent layout method based on particle swarm algorithm. Through the construction of data chain, it can effectively improve the automation and intelligence level of data center scheduling construction.

[0088] At the same time, this scheme adopts a local anomaly identification index method to capture anomalies in the data set, which can more effectively identify local anomalies and improve the accuracy of anomaly detection.

[0089] In addition, this solution solves the multi-objective optimization problem by improving the particle swarm algorithm to obtain the optimal sub-solution, which is the corresponding virtual machine scheduling solution. The sub-models of different data sets are fused, and the information fusion algorithm is used to integrate the optimal sub-solutions of different data sets to construct a detection chain for the sub-detection models of different data sets, so as to achieve the optimal detection of the final global virtual machine scheduling solution, and realize real-time update of detection parameters to adapt to the dynamic changes of structural data in the data set, so as to maintain the effectiveness of the detection model and make dynamic adjustments as the data changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] Figure 1 A schematic diagram of a process flow of a one-click data center intelligent layout method based on a particle swarm algorithm provided by an embodiment of the present invention;

[0091] Figure 2 A schematic diagram of the structure of an electronic device for implementing a method for building an adaptive database detection data link provided by an embodiment of the present invention.

[0092] In the figure: 1 electronic device, 10 processor, 11 memory, 12 program, 13 communication interface.

[0093] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0094] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0095] The embodiment of the present application provides a one-click data center intelligent layout method based on a particle swarm algorithm. The execution subject of the method includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the method can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0096] Embodiment 1:

[0097] S1: According to the connectivity relationship and weight of each device node in the data center, a symmetric matrix of the data link set is constructed and represented by an AOE directed acyclic graph network; a small data center has the following four device nodes: A, B, C, and D. The connectivity relationship and weights between these nodes are as follows: the weight from A to B is 2; the weight from A to C is 3; the weight from B to C is 1; the weight from C to D is 4; the weight from B to D is 5.

[0098] Constructing the data link set symmetric matrix,First, according to the connectivity relationship and weight of each device node in the data center, a data link set symmetric matrix is ​​constructed. In this matrix, the rows and columns represent the device nodes of the data center, and the elements in the matrix represent the weight from the row node to the column node. If there is no direct connectivity between two nodes, the corresponding matrix element is 0.

[0099] An AOE (Activity On Edge) network is a directed graph in which nodes represent events, edges represent activities, and each edge has a weight, where the weight represents the connectivity between two nodes and its cost or weight. For the above data center example, the following AOE network graph can be constructed: create nodes A, B, C, and D; according to matrix B, create an edge from node A to B with a weight of 2; create an edge from A to C with a weight of 3; create an edge from B to C with a weight of 1; create an edge from C to D with a weight of 4; create an edge from B to D with a weight of 5. This AOE network graph can help us more intuitively understand the connection mode and communication cost between nodes in the data center. By optimizing the paths in this graph, we can effectively manage the resources and data flow of the data center and improve the overall operating efficiency.

[0100] S2: The data link set is divided into multiple cluster center data sets by a clustering algorithm; in one embodiment, in the application scenario of the data center, a variety of clustering algorithms can be selected to process the division of the data link set. In some embodiments, the K-means clustering algorithm is used, which can effectively group data points to the nearest cluster center. Here, we choose the K-means algorithm for clustering. Our goal is to divide the four device nodes A, B, C, and D in the previous example into two cluster centers. Each cluster center will represent a data subset and can be considered independently when optimizing resource allocation and network traffic management. Randomly select two nodes as the initial cluster centers, A and C. For each node, calculate its distance to the two cluster centers and assign it to the nearest cluster center. The distance here can be measured by the weight value in the symmetric matrix constructed previously. For node B, the distance to A is 2 and the distance to C is 1, so B is assigned to cluster center C. For node D, the distance to A is 0 (no direct connection) and the distance to C is 4, so D is assigned to cluster center C. Recalculate the center of each cluster based on the assigned nodes. This may involve averaging the weights of all nodes in the cluster. Repeat steps 2 and 3 until the cluster center no longer changes or a predetermined number of iterations is reached. In the end, we get two cluster centers, cluster center 1 contains node A; cluster center 2 contains nodes B, C, and D. By grouping device nodes in the data center into different cluster centers, resources and communication efficiency in each cluster can be managed and optimized more accurately. For example, it may be found that the communication demand within some clusters is much higher than the communication demand with other clusters, so that higher-performance network devices can be deployed preferentially in these clusters or specific network policies can be implemented.

[0101] S3: Combined with the AOE directed acyclic graph network, multiple cluster center data sets are corrected respectively to obtain the corresponding cluster family sub-data link set. The clustered data set may not fully meet the optimization requirements of the data center network flow. In some embodiments, the purpose of correcting the cluster family sub-data link set is to adjust it according to the actual data flow and network structure to ensure the efficiency and load balancing of the data link. Use the AOE directed acyclic graph to analyze the connectivity and path efficiency between nodes in each cluster center. In this example, cluster center 1 only contains node A, and cluster center 2 contains nodes B, C, and D. The connectivity inside cluster center 2, that is, between nodes B, C, and D, is analyzed. According to the data flow and path efficiency, the connectivity inside cluster center 2 is corrected. Considering that the weight between node B and node D is 5, this is the edge with the highest weight in cluster center 2, indicating that the data flow may be large. If it is found that the data flow from node B to node D is continuously higher than other paths, adjustments are required to reduce congestion. For example, we can consider adding an auxiliary path directly from node C to node D, even if it already exists in the original AOE graph (weight 4), we can also improve its priority or increase its bandwidth. According to the modified connectivity graph, the cluster family sub-data link set is redefined. In this example, the possible adjustments are: cluster center 1 (node ​​A) remains unchanged, and cluster center 2 (nodes B, C, D) is modified to improve the priority of the path from C to D, and may also include the redistribution of the weight of the path from B to D.

[0102] S4: According to the balance index of each cluster family sub-data link set, the corresponding membership function is constructed; the balance index of the network usually includes the load capacity of the node, network delay, bandwidth utilization, etc. Through these indicators, the degree of balance of the data link set in terms of function and performance can be measured. The membership function is used to evaluate whether the nodes of each cluster center are suitable to continue to stay in the current cluster center, or whether they should consider migrating to other cluster centers to improve the overall network performance. The definition of balance index includes the following key indicators. The first is the node load, which represents the ratio of the amount of data currently processed by the node to its processing capacity. The second is the network delay, which represents the average time required for data to be transmitted from one node to another. The third is the bandwidth utilization, which represents the ratio of the current utilization of the connection between nodes to the maximum bandwidth. For each cluster center, a membership function is constructed based on the above indicators. These functions can be simple linear functions or more complex nonlinear functions, depending on the specific needs and strategies of data center management. When calculating the membership of the cluster center, for each node, the membership function is used to calculate the membership score of its corresponding cluster center. If the membership score of the node is lower than a certain threshold, it is considered to be migrated from the current cluster center to other cluster centers.

[0103] S5: Obtain the common energy consumption influencing parameters of each cluster family sub-data link set, and establish a multi-objective optimized sub-virtual machine scheduling model in combination with constraints; in some embodiments, step S5 involves how to use the common energy consumption influencing parameters in these data link sets after obtaining the cluster family sub-data link sets, and establish a multi-objective optimized sub-virtual machine scheduling model in combination with constraints. This step is the key to virtual machine deployment optimization in data center management, and the purpose is to ensure that the allocation of virtual machines on physical servers can maximize efficiency and meet various operating requirements. In a data center environment, common energy consumption influencing parameters generally include CPU utilization rate, virtual machine demand for CPU resources, the amount of memory required for virtual machine operation, the storage space required by the virtual machine, and the network bandwidth required for communication between virtual machines and between virtual machines and the outside world; when constructing a sub-virtual machine scheduling model, the constraints that need to be considered include the upper limit of the CPU, memory and storage resources of each physical server; the minimum standard of virtual machine performance guaranteed by the service level agreement (SLA); and the overall energy consumption target or limit of the data center. Collect the above energy consumption influencing parameters from each cluster family sub-data link set, analyze the deployment status of virtual machines in each cluster to determine the current resource usage, and establish a multi-objective optimization model using the collected data and constraints. This model may aim to minimize energy consumption while maximizing virtual machine performance and resource utilization, use mathematical optimization methods such as linear programming, integer programming or heuristic algorithms to express these goals, select appropriate optimization algorithms (such as genetic algorithms, simulated annealing, etc.) to solve multi-objective optimization problems, and perform algorithm tuning and testing to ensure the effectiveness and efficiency of the model.

[0104] S6: Solve the multi-objective optimization problem based on the improved particle swarm algorithm to obtain the optimal sub-solution, which is the corresponding virtual machine scheduling solution; in some embodiments, an improved particle swarm algorithm (PSO) is used to solve the multi-objective optimization problem constructed in the previous step S5, so as to find the optimal virtual machine scheduling solution. This step is to actually perform the optimization calculation to ensure that the resource allocation of the data center maximizes efficiency and meets different operating requirements. Particle swarm optimization is a swarm intelligence algorithm that simulates the hunting behavior of bird flocks. Each particle represents a potential solution, and by following its own and the group's optimal solution, the particle searches for the optimal solution in the solution space. In the improved PSO version, some strategies are usually introduced to improve the algorithm's search efficiency and the quality of the solution, such as mutation operations, local search strategies, or dynamic adjustment of parameters. The optimization goal is to minimize the total energy consumption and maximize the resource utilization; the constraint is not to exceed the resource upper limit of the server (CPU, memory, bandwidth) and meet the service level agreement (SLA); each particle represents a virtual machine deployment plan, that is, the distribution of virtual machines on the physical server; the initial particle swarm is randomly generated, and the position of each particle corresponds to a specific virtual machine deployment configuration; the PSO parameters are set including the learning factor (usually two: individual learning factor and social learning factor), inertia weight (a parameter that affects the particle speed and controls the balance between global search and local search), the maximum number of iterations and the particle swarm size; in each iteration, the fitness value of each particle is calculated based on the energy consumption and resource utilization of its corresponding solution; the individual optimal solution and the global optimal solution of each particle are updated; the speed and position of the particle are adjusted to explore new possible solutions; if necessary, mutation operations are performed to avoid local optimality, and the iteration is performed until the termination condition is met (such as reaching the maximum number of iterations or the quality of the solution is no longer significantly improved), and the virtual machine deployment plan represented by the particle with the best fitness value is selected as the final solution.

[0105] In one embodiment, a data center has 50 virtual machines that need to be deployed on 10 physical servers. We use the improved PSO algorithm to determine the best physical server location for each virtual machine to optimize energy consumption and increase resource utilization. The initial particle swarm consists of 100 particles, and each particle randomly assigns 50 virtual machines to the distribution of 10 servers. The optimization process iterates 200 times, using individual and social learning factors of 2.0 each, and the inertia weight is linearly reduced from 0.9 to 0.4. At the end of the iteration, an optimal virtual machine deployment plan is obtained, which achieves the best balance between energy consumption and resource utilization while meeting all constraints.

[0106] S7: Fuse the sub-models of different data sets, use the information fusion algorithm to integrate the optimal sub-solutions of different data sets to construct a detection chain for the sub-detection models of different data sets, and achieve the optimal detection of the final global virtual machine scheduling solution.

[0107] Step S7 involves fusing sub - models of different data sets, using information fusion algorithms to synthesize the optimal sub - solutions of these data sets, and constructing a detection chain to achieve the optimal detection of the global final virtual machine scheduling scheme. In some embodiments, this step aims to integrate multiple optimized sub - solutions, ensure the consistency and efficiency of the overall virtual machine scheduling scheme, while improving the reliability and robustness of the system. The information fusion algorithm is a technology that improves the quality of the decision - making process by integrating information from different sources or subsystems. In the virtual machine scheduling of a data center, information fusion can help merge solutions from different optimization models, ensuring that all relevant system constraints and objectives are comprehensively considered. In one embodiment, there are multiple sub - models, each corresponding to a different data center or data set, and each sub - model has independently solved its local multi - objective optimization problem. For example, one sub - model focuses on energy efficiency, and another may focus on maximizing performance. The optimal solutions are collected from each sub - model. These solutions may include the specific deployment plans of virtual machines, involving the optimized configuration of CPU, memory, storage, and network resources; using an information fusion algorithm, such as weighted average, decision fusion, or rule - based method, to merge these solutions; the fusion algorithm will consider the quality, reliability, and relevance of each sub - solution to determine their weights in the final decision; constructing a detection chain to verify whether the fused virtual machine scheduling scheme meets the requirements of all data sets; the detection chain will check the consistency of resource configuration, the satisfaction of performance metrics, and any possible security vulnerabilities.

[0108] In step S1, according to the connection relationship and weights of each device node in the data center, a symmetric matrix of the data link set is constructed and represented by an AOE directed acyclic graph network, including:

[0109] S11: For each data node I in the data center i , through the data transfer program, it is judged whether there is a connected path from the sending point to the receiving point of the data, where i is the node serial number, 0 < i ≤ N, and N is the total number of nodes;

[0110] S12: For the nodes with a connected path, 1 is used to represent them, and for the nodes without a connected path, 0 is used to represent them, and an N - order symmetric matrix B is constructed;

[0111] S13: According to the symmetric matrix B, the transmission routes passed by the connected paths are screened, and an AOE directed acyclic graph network is constructed in combination with the said weights.

[0112] In step S3, in combination with the AOE directed acyclic graph network, multiple cluster centers are respectively corrected to obtain the corresponding sub - data link sets of the cluster families, including:

[0113] S31: Obtain the neighborhood set P(pi)=[p1…pk] of each cluster center data set, where k is the number of nodes in the neighborhood set and o is the cluster center;

[0114] S32: Calculate the connected distance Lp-o between the data point pi in the neighborhood set and the cluster center o, and calculate the connected distance Lp-p' between the data point pi in the neighborhood set and the adjacent point pi';

[0115] S33: Calculate the local anomaly identification index LOF p-p'(pi) of the data point pi, and the calculation formula is:

[0116]

[0117] in:

[0118] P(pi) represents the neighborhood set of data point pi;

[0119] Lp-p' represents the distance from the data point pi to its neighboring points;

[0120] Lp-o represents the distance from data point pi to cluster center o;

[0121] LOF pp′(pi) represents the local anomaly identification index of data point pi;

[0122] S34: If the local anomaly recognition index exceeds the specified threshold, the data point pi is judged as an abnormal data point and is not added to the corresponding cluster family sub-data link set; if the local anomaly recognition index does not exceed the specified threshold, the data point pi is judged as a non-abnormal data point and is added to the corresponding cluster family sub-data link set.

[0123] In the step S4, a corresponding membership function is constructed according to the balance index of each cluster group sub-data link set, including:

[0124] S41: If a connected path exists between every two vertices in the AOE directed acyclic graph network, determine the consumption cost p corresponding to the edge between every two vertices according to the weight g(i, j) between every two storage nodes. cov ;

[0125] Wherein, if a connected path between every two vertices in the AOE directed acyclic graph network exists, the corresponding occupancy coefficient p(i, j) is 1; if a connected path between every two vertices in the AOE directed acyclic graph network does not exist, the corresponding occupancy coefficient p(i, j) is 0;

[0126] S42: Calculate the cost p of the edge between each two vertices cov ,include:

[0127] Determine the corresponding sub-symmetric matrix Z according to the cluster family sub-data link set m×n , where the position of each site is s ij , the calculation formula is:

[0128] s ij =(x ij ,y ij );

[0129]

[0130] Among them, m and n are the sub-symmetric matrices Z m×n The number of rows and columns;

[0131] S43: According to the consumption cost p cov Determine the type of sub-model and construct the corresponding evaluation function, including:

[0132]

[0133] Among them, A1 and A2 are the types of models;

[0134] S44: Determine the type of each sub-model according to the evaluation function, and then process the multi-objective function to optimize each sub-model, including: if the sub-model belongs to the A1 type, increase the importance of each indicator of the sub-model; if the sub-model belongs to the A2 type, reduce the importance of each indicator of the sub-model.

[0135] In step S6, the multi-objective optimization problem is solved based on the improved particle swarm algorithm to obtain the optimal sub-solution, including:

[0136] S61: Initialize the population and its parameters, including population size, particle dimension, number of iterations, indicator membership, indicator weight, determine the encoding method, use binary encoding, and represent the physical host where each virtual machine is placed as a binary string;

[0137] S62: Optimizing the system objective function using the evaluation function;

[0138] S63: Calculate the fitness function value of the particle corresponding to the path through the statistical information of the link, and calculate the aggregation degree of the current particle from the fitness value;

[0139] S64: Calculating the individual historical optimal solution of each particle according to the aggregation degree;

[0140] S65: Take the position in the particle swarm as a candidate solution, and select a part of the particles to generate the next generation of population and particle swarm according to the similarity between the candidate solution and the individual historical optimal solution;

[0141] S66: For multi-objective problems, the fitness value of each particle is calculated using a multi-objective fitness function;

[0142] S67: Compare the fitness value of the current particle with the optimal fitness value of the individual particle of the previous generation. If the fitness value of the current particle is less than the optimal fitness value of the individual particle of the previous generation, update the optimal solution of the individual particle, otherwise do not update.

[0143] S68: Determine whether the termination condition is met: Determine whether the algorithm has converged based on whether the updated particle individuals are stable; if it has converged.

[0144] In step S6, the multi-objective optimization problem is solved based on the improved particle swarm algorithm to obtain an optimal sub-solution, which is the corresponding virtual machine scheduling solution, including:

[0145] S61: Obtaining the objective function of each sub-model;

[0146] S62: Perform crossover and mutation operations on each sub-model data set based on the improved particle swarm algorithm;

[0147] S63: sort each objective function, and then calculate a new speed and a new position respectively;

[0148] S64: Based on the fact that each particle position corresponds to a flow scheduling scheme, the particle position corresponds to the scheduling path of the network flow in the flow scheduling scheme, and the best path is obtained, that is, the optimal sub-solution and the corresponding virtual machine scheduling scheme are obtained.

[0149] In the step S7, the sub-models of different data sets are fused, and the optimal sub-solutions of different data sets are integrated with the information fusion algorithm to construct a detection chain for the sub-detection models of different data sets, so as to achieve the optimal detection of the final global virtual machine scheduling solution, including:

[0150] S71: Fusing the sub-model data sets of different data sets to obtain a fused matrix data set again, and constructing a fused AOE directed acyclic graph network;

[0151] S72: according to the fused AOE directed acyclic graph network, taking the sum of the weights corresponding to all the edges in each feasible link as the consumption cost corresponding to each feasible link;

[0152] S73: Obtain the optimal link according to the consumption cost corresponding to each feasible link, the CPU utilization rate and the transmission load of each feasible link;

[0153] S74: Based on the node path on the optimal link, final global virtual machine scheduling is implemented.

[0154] Embodiment 2:

[0155] like Figure 2 FIG. 1 is a schematic diagram of the structure of an electronic device of a method provided by an embodiment of the present invention.

[0156] The electronic device 1 may include a processor 10 , a memory 11 , a communication interface 13 and a bus, and may also include a computer program stored in the memory 11 and executable on the processor 10 , such as a program 12 .

[0157] The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a magnetic memory, a disk, an optical disk, etc. The memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, such as a mobile hard disk of the electronic device 1. The memory 11 may also be an external storage device of the electronic device 1 in other embodiments, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device. The memory 11 may be used not only to store application software and various types of data installed in the electronic device 1, such as the code of the program 12, etc., but also to temporarily store data that has been output or is to be output.

[0158] The processor 10 may be composed of an integrated circuit in some embodiments, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips, etc. The processor 10 is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect various components of the entire electronic device, and executes or executes programs or modules stored in the memory 11 (such as the program 12 for implementing the construction of the adaptive database detection data link), and calls the data stored in the memory 11 to execute various functions of the electronic device 1 and process data.

[0159] The communication interface 13 may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices, and to achieve connection and communication between internal components of the electronic device.

[0160] The bus may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize connection and communication between the memory 11 and at least one processor 10, etc.

[0161] Figure 2 Only an electronic device with components is shown, and those skilled in the art will understand that Figure 2 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0162] For example, although not shown, the electronic device 1 may also include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include any components such as one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, etc. The electronic device 1 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0163] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device. The display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.

[0164] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0165] The program 12 stored in the memory 11 in the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve:

[0166] According to the connectivity and weight of each device node in the data center, a symmetric matrix of the data link set is constructed and represented by an AOE directed acyclic graph network;

[0167] The data link set is divided into multiple cluster center data sets through a clustering algorithm;

[0168] Combined with AOE directed acyclic graph network, multiple cluster center data sets are modified respectively to obtain the corresponding cluster family sub-data link sets;

[0169] According to the balance index of each cluster family sub-data link set, the corresponding membership function is constructed;

[0170] Obtain the common energy consumption influencing parameters of each cluster family sub-data link set, and establish a multi-objective optimization sub-virtual machine scheduling model based on constraint conditions;

[0171] Based on the improved particle swarm algorithm, the multi-objective optimization problem is solved to obtain the optimal sub-solution, which is the corresponding virtual machine scheduling solution;

[0172] The sub-models of different data sets are fused, and the optimal sub-solutions of different data sets are integrated using the information fusion algorithm to construct a detection chain for the sub-detection models of different data sets, thereby achieving the optimal detection of the final global virtual machine scheduling solution.

[0173] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to Figure 1 to Figure 2 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0174] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0175] 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 invention is essentially 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, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0176] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A one-click data center intelligent layout method based on particle swarm algorithm, characterized in that: The method comprises: S1: According to the connectivity and weight of each device node in the data center, a symmetric matrix of the data link set is constructed and represented by an AOE directed acyclic graph network; S2: Divide the data link set into multiple cluster center data sets through clustering algorithm; S3: Combine the AOE directed acyclic graph network to modify multiple cluster center data sets respectively to obtain the corresponding cluster family sub-data link sets; S4: construct the corresponding membership function according to the balance index of each cluster family sub-data link set; S5: obtaining the common energy consumption influencing parameters of the sub-data link sets of each cluster family, and establishing a sub-virtual machine scheduling model with multi-objective optimization in combination with constraint conditions; S6: Solve the multi-objective optimization problem based on the improved particle swarm algorithm to obtain the optimal sub-solution, which is the corresponding virtual machine scheduling solution; S61: Initialize the population and its parameters, including population size, particle dimension, number of iterations, indicator membership, indicator weight, determine the encoding method, use binary encoding, and represent the physical host where each virtual machine is placed as a binary string; S62: Optimizing the system objective function using the evaluation function; S63: Calculate the fitness function value of the particle corresponding to the path through the statistical information of the link, and calculate the aggregation degree of the current particle from the fitness value; S64: Calculating the individual historical optimal solution of each particle according to the aggregation degree; S65: Take the position in the particle swarm as a candidate solution, and select a part of the particles to generate the next generation of population and particle swarm according to the similarity between the candidate solution and the individual historical optimal solution; S66: For multi-objective problems, the fitness value of each particle is calculated using a multi-objective fitness function; S67: Compare the fitness value of the current particle with the optimal fitness value of the individual particle of the previous generation. If the fitness value of the current particle is less than the optimal fitness value of the individual particle of the previous generation, update the optimal solution of the individual particle, otherwise do not update. S68: Determine whether the termination condition is met: Determine whether the algorithm has converged based on whether the updated particle individuals are stable; S7: Fuse the sub-models of different data sets, use the information fusion algorithm to integrate the optimal sub-solutions of different data sets to construct a detection chain for the sub-detection models of different data sets, and achieve the optimal detection of the final global virtual machine scheduling solution.

2. The one-click data center intelligent layout method based on particle swarm algorithm according to claim 1, characterized in that: In the step S1, a symmetric matrix of a data link set is constructed according to the connectivity relationship and weight of each device node in the data center, and is represented by an AOE directed acyclic graph network, including: S11: For each data node I in the data center i , determine whether there is a connected path for the data from the sending point to the data receiving point through the data transfer program, where i is the node number, 0 < i ≤ N, and N is the total number of nodes; S12: Use 1 to represent nodes with connected paths, and use 0 to represent nodes without connected paths, and construct an N-order symmetric matrix B; S13: Filter the transmission paths through which the connected paths pass according to the symmetric matrix B, and construct an AOE directed acyclic graph network in combination with the weights.

3. The one-click data center intelligent layout method based on particle swarm algorithm according to claim 1, characterized in that: In the step S3, multiple cluster centers are modified respectively in combination with the AOE directed acyclic graph network to obtain corresponding cluster family sub-data link sets, including: S31: Obtain the neighborhood set P(pi)=[p1…pk] of each cluster center data set, where k is the number of nodes in the neighborhood set and o is the cluster center; S32: Calculate the connected distance Lp-o between the data point pi in the neighborhood set and the cluster center o, and calculate the connected distance Lp-p' between the data point pi in the neighborhood set and the adjacent point pi'; S33: Calculate the local anomaly identification index LOF p-p'(pi) of the data point pi, and the calculation formula is: in: P(pi) represents the neighborhood set of data point pi; Lp-p' represents the distance from the data point pi to its neighboring points; Lp-o represents the distance from data point pi to cluster center o; LOF pp′(pi) represents the local anomaly identification index of data point pi; S34: If the local anomaly recognition index exceeds the specified threshold, the data point pi is judged as an abnormal data point and is not added to the corresponding cluster family sub-data link set; if the local anomaly recognition index does not exceed the specified threshold, the data point pi is judged as a non-abnormal data point and is added to the corresponding cluster family sub-data link set.

4. The one-click data center intelligent layout method based on particle swarm algorithm as claimed in claim 2, characterized in that: In the step S4, a corresponding membership function is constructed according to the balance index of each cluster group sub-data link set, including: S41: If a connected path exists between every two vertices in the AOE directed acyclic graph network, determine the consumption cost p corresponding to the edge between every two vertices according to the weight g(i,j) between every two storage nodes. cov ; Wherein, if a connected path between every two vertices in the AOE directed acyclic graph network exists, the corresponding occupancy coefficient p(i, j) is 1; if a connected path between every two vertices in the AOE directed acyclic graph network does not exist, the corresponding occupancy coefficient p(i, j) is 0; S42: Calculate the cost p of the edge between each two vertices cov ,include: Determine the corresponding sub-symmetric matrix Z according to the cluster family sub-data link set m×n , where the position of each site is s ij , the calculation formula is: s ij =(x ij ,y ij ); Among them, m and n are the sub-symmetric matrices Z m×n The number of rows and columns; S43: According to the consumption cost p cov Determine the type of sub-model and construct the corresponding evaluation function, including: Among them, A1 and A2 are the types of models; S44: Determine the type of each sub-model according to the evaluation function, and then process the multi-objective function to optimize each sub-model, including: if the sub-model belongs to the A1 type, increase the importance of each indicator of the sub-model; if the sub-model belongs to the A2 type, reduce the importance of each indicator of the sub-model.

5. The one-click data center intelligent layout method based on particle swarm algorithm as claimed in claim 4, characterized in that: In step S6, the multi-objective optimization problem is solved based on the improved particle swarm algorithm to obtain an optimal sub-solution, which is the corresponding virtual machine scheduling solution, including: S61: Obtaining the objective function of each sub-model; S62: Perform crossover and mutation operations on each sub-model data set based on the improved particle swarm algorithm; S63: sort each objective function, and then calculate a new speed and a new position respectively; S64: Based on the fact that each particle position corresponds to a flow scheduling scheme, the particle position corresponds to the scheduling path of the network flow in the flow scheduling scheme, and the best path is obtained, that is, the optimal sub-solution and the corresponding virtual machine scheduling scheme are obtained.

6. The one-click data center intelligent layout method based on particle swarm algorithm according to claim 1, characterized in that: In the step S7, the sub-models of different data sets are fused, and the optimal sub-solutions of different data sets are integrated with the information fusion algorithm to construct a detection chain for the sub-detection models of different data sets, so as to achieve the optimal detection of the final global virtual machine scheduling solution, including: S71: Fusing the sub-model data sets of different data sets to obtain a fused matrix data set again, and constructing a fused AOE directed acyclic graph network; S72: according to the fused AOE directed acyclic graph network, taking the sum of the weights corresponding to all the edges in each feasible link as the consumption cost corresponding to each feasible link; S73: Obtaining the best link according to the consumption cost corresponding to each feasible link, the CPU utilization rate of each feasible link, and the transmission load; S74: Based on the node path on the best link, the final global virtual machine scheduling is implemented.

7. A one-click data center intelligent layout method system based on particle swarm algorithm, characterized in that: include: Original data set module: According to the connectivity and weight of each device node in the data center, a symmetric matrix of the data link set is constructed and represented by an AOE directed acyclic graph network; Sub-dataset module: divides the data link set into multiple cluster center data sets through clustering algorithm; Screening module: Combined with the AOE directed acyclic graph network, multiple cluster center data sets are corrected to obtain the corresponding cluster family sub-data link set; Evaluation module: construct the corresponding membership function according to the balance index of each cluster group sub-data link set; Correction module: obtain the common energy consumption influencing parameters of each cluster family sub-data link set, and establish a multi-objective optimization sub-virtual machine scheduling model based on constraint conditions; Optimization module: solves the multi-objective optimization problem based on the improved particle swarm algorithm to obtain the optimal sub-solution, which is the corresponding virtual machine scheduling solution; Initialize the population and its parameters, including population size, particle dimension, number of iterations, indicator membership, indicator weight, determine the encoding method, use binary encoding, and represent the physical host where each virtual machine is placed as a binary string; Use the evaluation function to optimize the system objective function; The fitness function value of the particle's corresponding path is calculated through the link's statistical information, and the aggregation degree of the current particle is calculated from the fitness value; Calculate the individual historical optimal solution of each particle according to the aggregation degree; The positions in the particle swarm are taken as candidate solutions, and according to the similarity between the candidate solutions and the individual historical optimal solutions, a part of the particles are selected to generate the next generation of population and particle swarm; For multi-objective problems, the fitness value of each particle is calculated using a multi-objective fitness function; Compare the fitness value of the current particle with the optimal fitness value of the individual particle of the previous generation. If the fitness value of the current particle is less than the optimal fitness value of the individual particle of the previous generation, update the optimal solution of the individual particle, otherwise do not update. Determine whether the termination condition is met: Determine whether the algorithm has converged based on whether the updated particle individuals are stable; if it has converged; Detection module: fuses the sub-models of different data sets, uses the information fusion algorithm to integrate the optimal sub-solutions of different data sets to build a detection chain for the sub-detection models of different data sets, and realizes the optimal detection of the final global virtual machine scheduling solution.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the one-click data center intelligent layout method based on particle swarm algorithm as described in any one of claims 1 to 6 are implemented.

9. A computer storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the one-click data center intelligent layout method based on particle swarm algorithm as described in any one of claims 1 to 6 are implemented.

Images