Virtual network resource allocation method in multi-server scene

Abstract

A virtual network resource allocation method in a multi-server scene is based on a network communication system of a virtual network and a physical network, resources needed by virtual nodes comprise computing power, memory and storage, and the demand of a virtual link is bandwidth; the physical network is composed of physical nodes and physical links, and the core of virtual network resource mapping is that virtual nodes and links are reasonably distributed to the physical nodes and the links; in a multi-server environment, communication among virtual nodes can cross a plurality of physical servers, so that the number and efficiency of cross-host links become the key to influence the system performance; a mapping algorithm based on the adaptation degree is adopted; judging whether the physical node can meet the requirement of the virtual node or not by calculating the adaptation degree between the virtual node and the physical node; the adaptation degree is calculated according to resource requirements of the virtual nodes and available resources of the physical nodes, and only the physical nodes with the adaptation degree reaching a set threshold value are regarded as effective mapping targets.

Description

Technical Field

[0001] The invention relates to a network communication technology, in particular to a multi-server network virtualization resource allocation method. Background Art

[0002] With the rapid development of network function virtualization and cloud computing technology, broadband operators are facing increasingly complex challenges in network resource management and optimization. The emergence of network function virtualization allows traditional network functions to run on general hardware platforms through software, thereby improving the flexibility and scalability of the network. However, this also brings new problems, namely how to efficiently allocate virtual network resources in a multi-server environment to achieve optimal resource utilization and improve network performance.

[0003] In multi-server scenarios, the allocation of virtual network resources becomes a core issue. The communication needs between virtual nodes are frequent. How to reduce the number of cross-host links and optimize the data transmission paths between physical hosts is the key to improving overall network performance. Especially in large-scale network deployments, unoptimized cross-host links will cause network bottlenecks, increase data transmission delays, and significantly affect user experience.

[0004] Existing virtual network resource allocation methods mainly focus on how to map virtual nodes to physical servers to achieve resource optimization and network performance improvement. However, these methods have the following shortcomings:

[0005] 1. Insufficient cross-host link optimization: In a multi-server environment, frequent communications between virtual nodes need to go through different physical hosts. Current methods fail to fully optimize the number of cross-host links, resulting in large data transmission delays, which in turn affects the overall performance of the network.

[0006] 2. Unbalanced resource utilization: The existing resource allocation method lacks effectiveness in load balancing. Some physical servers are overloaded, while others are idle, and resources are not fully utilized. Optimizing resource allocation can not only improve the economic benefits of the system, but also reduce unnecessary energy consumption.

[0007] 3. Lack of real-time dynamic adjustment mechanism: Many existing allocation strategies are only aimed at initial allocation and fail to dynamically adjust virtual node allocation according to the real-time load of the physical server, resulting in system performance degradation and serious waste of resources in high-load scenarios.

[0008] In summary, the current virtual network resource allocation method has obvious deficiencies in cross-host link optimization, resource utilization balance and dynamic allocation strategy, and cannot fully meet the needs of broadband operators in a multi-server environment. In order to solve these problems, there is an urgent need for a virtual network resource allocation method that can optimize cross-host links between physical hosts, improve resource utilization and support real-time dynamic adjustment to adapt to the actual application scenarios of broadband operators in network function virtualization and cloud computing environments. Summary of the invention

[0009] The purpose of the present invention is to provide a virtual network resource allocation method in a multi-server scenario, covering two parts: initial allocation and real-time dynamic adjustment, aiming to optimize the number of cross-host links between multiple physical hosts, and improve the efficiency, real-time and overall system performance of resource allocation. The present invention not only effectively optimizes the resource utilization and cross-host communication of virtual nodes through the initial allocation method, but also makes intelligent adjustments according to the real-time load conditions of physical nodes during system operation through the dynamic allocation module, ensuring that the system still maintains stable performance under high load conditions.

[0010] The technical solution of the present invention is: a method for allocating virtual network resources in a multi-server scenario, a network communication system based on a virtual network and a physical network, wherein the virtual nodes and links in the virtual network are characterized by certain resource requirements, the resources required by the virtual nodes include computing power, memory, storage, etc., while the requirement of the virtual links is bandwidth; the physical network is composed of physical nodes and physical links, each physical node has a certain resource capacity including processor, memory, storage, etc., and the physical links provide corresponding bandwidth resources.

[0011] Use G V Represents a virtual network, which consists of a set of virtual nodes and virtual links. Each virtual node n V Has specific resource requirements V , virtual link l V With bandwidth requirement B V ; Physical Network G P Contains physical nodes and physical links, physical node n P With available resources R P , physical link l P Provide available bandwidth B P ;

[0012] The core of virtual network resource mapping is to reasonably allocate virtual nodes and links to physical nodes and links. During the allocation process, the following conditions must be guaranteed:

[0013] Each virtual node must be mapped to a physical node that can provide sufficient resources, that is, R P ≥R V, to meet the computing and storage requirements of virtual nodes;

[0014] Each virtual link must be mapped to a physical link that can provide sufficient bandwidth, that is, B P ≥B V To ensure network communication needs;

[0015] In a multi-server environment, the communication between virtual nodes may span multiple physical servers, so the number and efficiency of cross-host links become the key to system performance. In order to optimize the link configuration, the present invention designs a link optimization strategy to reduce the number of cross-host links, thereby reducing data transmission delay and improving resource utilization efficiency. This strategy analyzes the communication requirements between virtual nodes and maps virtual nodes with frequent communication to the same physical node as much as possible to avoid unnecessary cross-host communication. When cross-host links cannot be avoided, try to select physical links with higher link capacity and lower load for mapping to ensure the stability and efficiency of data transmission.

[0016] In order to achieve efficient resource allocation, the present invention adopts a mapping algorithm based on fitness. The algorithm determines whether the physical node can meet the needs of the virtual node by calculating the fitness between the virtual node and the physical node. The fitness is calculated based on the resource requirements of the virtual node and the available resources of the physical node. Only physical nodes whose fitness reaches the set threshold are considered to be valid mapping targets. During the mapping process, this method gives priority to physical nodes with high fitness and low load to ensure balanced resource utilization. In addition, the optimization of the link is considered simultaneously during the initial allocation. By evaluating the communication needs of the virtual nodes, nodes that communicate frequently are preferentially mapped to the same physical node, reducing the demand for cross-host links.

[0017] The allocation mechanism of virtual network resources consists of the following parts: network resource judgment, network resource quantification, link strategy optimization, initial resource allocation, and dynamic resource allocation.

[0018] (1) The process of network resource judgment is achieved through a comprehensive analysis of the resource requirements of virtual nodes and the available resources of physical nodes. First, the system collects the resource requirement information of virtual nodes, such as computing power, memory, storage, and bandwidth, to form a virtual node requirement matrix. At the same time, the system monitors the resource status of each physical node, including available computing power, memory, storage space, and bandwidth, to form a physical resource matrix. Next, by calculating the resource adaptability between virtual nodes and physical nodes, it is determined whether the physical nodes can meet the needs of virtual nodes, providing a reliable basis for subsequent allocation and link optimization.

[0019] (2) The process of network resource quantification is to express the resource requirements and resource status of virtual nodes and physical nodes with specific numerical values ​​so as to achieve accurate matching in the allocation process. First, the resource requirements of virtual nodes (such as CPU, memory, storage, and bandwidth) are quantified into demand vectors, which represent the specific resource requirements of each virtual node. Secondly, the available resources of physical nodes are also quantified into corresponding resource vectors, which record the amount of resources they can currently provide. By establishing the virtual node demand matrix and the physical node available resource matrix, the demand and supply of network resources are clearly quantified. This process provides a numerical basis for the subsequent virtual node mapping, helps to accurately calculate the fitness, and optimize the allocation of virtual nodes.

[0020] (3) The process of link strategy optimization is to improve data transmission efficiency and reduce latency by reducing the number of links between virtual nodes that communicate across physical hosts. First, the system analyzes the communication needs between virtual nodes and identifies virtual node pairs that communicate frequently. Then, during resource allocation, these frequently communicating virtual nodes are mapped to the same physical host as much as possible to avoid unnecessary cross-host links. When cross-host communication is unavoidable, the system will give priority to mapping physical links with sufficient bandwidth and low load to ensure the stability and efficiency of data transmission. Through this link optimization strategy, network performance is improved and resource utilization is more reasonable.

[0021] (4) Dynamic resource allocation. The final process of dynamic resource allocation will be carried out according to the results of these preliminary steps. Based on the allocation results obtained in the preliminary calculation, the system will generate a set of configuration files corresponding to a set of physical nodes. Each file corresponds to a physical node and contains the virtual node mapping relationship of the node. These files usually use a standardized format (such as YAML) and include parameters such as the resource requirements of the virtual node, the allocated computing power, memory, storage space, and bandwidth. Use automated operation and maintenance tools (such as Kubernetes) to push these configuration files to the corresponding physical nodes. The automated tool reads the configuration file and starts and configures the corresponding virtual machine instance on each physical node according to the file content to ensure that the resource requirements of the virtual node match the resources of the physical node. The entire initial allocation process ensures the reasonable allocation of resources between virtual nodes and physical nodes, optimizes the data transmission path, and maximizes the utilization of physical resources.

[0022] Beneficial effects: The present invention proposes a method for allocating virtual network resources, aiming to solve the problems of low physical resource utilization and excessive number of cross-host links when virtual nodes are mapped to physical nodes in a multi-server environment. The method effectively judges and allocates virtual network and physical network resources to ensure that cross-host link configuration is optimized while satisfying resource requirements.

[0023] The virtual network resource allocation scheme of the present invention provides a highly efficient, flexible and low-cost solution for broadband operators, which can meet the actual needs in today's complex network environment and significantly improve the overall performance and economic benefits of the system:

[0024] The method of the present invention is particularly suitable for broadband operators facing the needs of large-scale virtual network deployment and management in the network function virtualization and cloud computing environment. It can effectively solve the problems of insufficient cross-host link optimization, unbalanced resource utilization, and untimely dynamic response in the existing resource allocation methods, and provide operators with efficient and flexible resource allocation solutions in dynamic and changeable network environments.

[0025] The specific advantages are:

[0026] 1. Improve resource utilization: By quantifying and matching the resources of virtual nodes and physical nodes, the adaptation algorithm designed in the present invention ensures that the resource demand of virtual nodes is highly matched with the resource supply of physical nodes, thereby maximizing the resource utilization of physical servers and reducing resource waste.

[0027] 2. Reduce cross-host link communication: This technical solution pays special attention to optimizing the link configuration across physical hosts. By preferentially mapping virtual nodes that frequently communicate to the same physical node, the number of cross-host links is effectively reduced, thereby reducing data transmission latency and improving network performance.

[0028] 3. Dynamic adjustment capability: The present invention not only covers the initial resource allocation, but also introduces a real-time dynamic adjustment mechanism, which can adaptively adjust the resource allocation according to the load of the physical node. This ensures that the system still maintains high stability and efficiency when the load changes, and is suitable for dynamic and changeable network environments.

[0029] 4. Support large-scale network deployment: This method is particularly suitable for the deployment of large-scale virtual networks. By introducing and optimizing link degree information, it can reduce system complexity while maintaining network performance and ensure high scalability and adaptability in large-scale environments.

[0030] 5. Reduce operating costs: By improving resource utilization and reducing ineffective cross-host communication links, the present invention can help operators significantly reduce operating and maintenance costs in large-scale network deployment. At the same time, the optimization of resource allocation also reduces energy consumption, further improving the economic benefits of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a schematic diagram of a virtual network resource allocation scenario provided by the present invention;

[0032] Figure 2It is a schematic diagram of the virtual network resource allocation process provided by the present invention;

[0033] Figure 3 It is a schematic diagram of a node allocation method based on fitness of the present invention;

[0034] Figure 4 It is a schematic diagram of the algorithm flow of the initial allocation module of the present invention;

[0035] Figure 5 It is a schematic diagram of the algorithm flow of the dynamic distribution module of the present invention. DETAILED DESCRIPTION

[0036] The present invention is mainly composed of a resource analysis module, an initial allocation module, a mapping operation module and a dynamic allocation module. The resource analysis module is responsible for comprehensively evaluating and judging the resource requirements of virtual nodes and the available resources of physical nodes, ensuring that each virtual node can be accurately matched with its corresponding physical node during the resource allocation process. By quantifying the resource status of virtual nodes and physical nodes, a clear resource requirement matrix and resource availability matrix are constructed, providing a reliable basis for subsequent mapping.

[0037] The initial allocation module uses a genetic algorithm, combined with the link degree information of the virtual node, to optimize the mapping process of the virtual node to the physical node through operations such as selection, crossover and mutation. In this module, the adaptability not only considers the reasonable allocation of resources, but also introduces the link degree as an important evaluation index, thereby effectively reducing the number of cross-host links and improving the overall network performance. Through the collaborative work of these two modules, the present invention can achieve efficient and flexible initial allocation of virtual network resources to meet the actual needs in a multi-server environment.

[0038] The mapping operation module is responsible for actually applying the virtual node link allocation results output by the algorithm mapping module to the physical host. The main task of this module is to generate configuration files and use automated tools to allocate virtual nodes to physical nodes.

[0039] The dynamic allocation module is designed to dynamically adjust the allocation of virtual nodes based on the real-time load of physical nodes and the resource requirements of virtual nodes. By real-time monitoring the resource usage of physical nodes, the module will trigger resource reallocation operations when the load exceeds the set threshold. By introducing the particle swarm optimization algorithm, the module can adaptively adjust the allocation scheme of virtual nodes to achieve system load balancing and resource optimization.

[0040] 1. Resource analysis module: judge and quantify the resource requirements of virtual network nodes and the resource status of physical nodes through modeling, and conduct preliminary analysis on the compatibility of the two.

[0041] Step 1: Modeling virtual node resource requirements

[0042] Each virtual node V in the virtual network i The resource requirements can be characterized by the following key indicators:

[0043] Computing power requirements Indicated by the number of CPU cores.

[0044] Memory requirements Expressed in GB.

[0045] Storage requirements Expressed in GB.

[0046] Bandwidth requirements Expressed in Mbps.

[0047] The resource requirements of each virtual node can be combined into a resource requirement vector:

[0048]

[0049] Step 2: Modeling physical node resource status

[0050] Each server node P in the physical network j The resource status can also be characterized by similar key indicators:

[0051] Available computing power Indicated by the number of CPU cores.

[0052] Available memory Expressed in GB.

[0053] Available storage Expressed in GB.

[0054] Available bandwidth Expressed in Mbps.

[0055] The available resources of the server can be summarized into a resource status vector:

[0056]

[0057] Step 3: Establishment of resource demand and resource status matrix

[0058] By organizing the resource requirements of all virtual nodes and the resource status of all physical nodes into a matrix, the resource matching process can be effectively managed.

[0059] Virtual Node Requirements Matrix:

[0060]

[0061] Physical node resource status matrix R P :

[0062]

[0063] Step 4: Calculate resource suitability

[0064] After quantifying the resources of virtual nodes and physical nodes, the next step is to calculate the Compatibility Fit Rating (CFR) ij The adaptability is used to evaluate whether the physical node can meet the resource requirements of the virtual node. The calculation formula is as follows:

[0065]

[0066] Step 5: Setting the fitness threshold

[0067] In order to ensure the rationality of the mapping, a fitness threshold CFR is set Th , only when CFR ij ≥CFR Th When the physical node P j To be considered as satisfying the virtual node V i Valid mapping targets.

[0068] Summary: Through the above modeling process, the resource requirements of virtual nodes and the resource status of physical nodes are effectively quantified and analyzed, laying a solid foundation for the subsequent virtual node mapping. This process ensures accurate matching during resource allocation, maximizes resource utilization efficiency, and provides a scientific basis for the management and allocation of the entire network resources.

[0069] 2. Initial allocation module: The mapping process of virtual nodes to physical nodes is optimized by combining genetic algorithm with virtual node link degree. The following is the detailed algorithm design of integrating link degree and genetic algorithm.

[0070] Step 1: Algorithm Input

[0071] Virtual node demand matrix R V : Resource requirements for each virtual node.

[0072] Physical node resource status matrix R P : Available resources for each physical node.

[0073] ·Fitness threshold T: The fitness threshold used to screen valid mappings.

[0074] Link degree matrix D: The link degree of each virtual node.

[0075] Population size P: The population size in the genetic algorithm.

[0076] Crossover probability C p : The probability of the crossover operation.

[0077] Mutation probability M p : The probability of mutation operation.

[0078] Iteration number G: The number of iterations of the algorithm.

[0079] Step 2: Initialize the population

[0080] Generate an initial solution:

[0081] Randomly generate P mapping solutions, each of which contains the mapping relationship between virtual nodes and physical nodes.

[0082] Ensure that each virtual node is mapped to a physical node with a fitness greater than a threshold T.

[0083] Step 3: Goodness of fit evaluation

[0084] For each solution S in the population k (represents the relationship between a set of virtual nodes mapped to physical nodes), calculate the fitness F(S k ). The adaptability not only considers the adaptability, but also combines the link degree. The formula is:

[0085]

[0086] Among them, CFR ij is the degree of fit, is the link degree of the virtual node, and λ is the weight coefficient of the link degree’s contribution to the fitness.

[0087] Step 4: Choose Action - Tournament Selection

[0088] In the selection phase, a tournament selection method is used to decide which individuals will enter the next generation population.

[0089] 4.1: Setting Tournament Parameters

[0090] Determine the size T of the tournament, that is, the number of individuals randomly selected from the population each time (e.g., T=3).

[0091] 4.2: Conducting a Tournament

[0092] Repeat the following process until the required number of adapted individuals N is selected:

[0093] Randomly select individuals: randomly select T individuals from the current population.

[0094] Calculate the fitness: Calculate the fitness F(S) of these T individuals ti ).

[0095] Select the fittest individual: Select the individual with the highest fitness as the winner.

[0096] 4.3: Forming a new population

[0097] The selected adaptive individuals are added to the new population S' until the size of the new population reaches a predetermined value.

[0098] Step 5: Crossover Operation

[0099] Perform a crossover operation on the selected solution to generate a new solution. The crossover operation uses single-point crossover or multi-point crossover:

[0100] Select crossover individuals: Randomly select two solutions S i and S j .

[0101] Crossover point selection: The crossover point is determined randomly.

[0102] Generate a new solution: keep the part before the intersection, exchange the part after the intersection, and generate a new solution.

[0103] Step 6: Mutation Operation

[0104] The new solution is mutated to increase the diversity of the population. The mutation operation can be achieved by randomly selecting a virtual node and reallocating it to another physical node.

[0105] Select mutation individuals: according to the mutation probability M p , randomly select individuals for mutation.

[0106] ·Execute mutation: Randomly change the mapping of a virtual node in the individual.

[0107] Step 7: Generate a new generation of population

[0108] The solution after crossover and mutation is combined with the original population to form a new population S", and the optimal solution is selected based on the fitness.

[0109] Step 8: Iteration process

[0110] Repeat steps 3 to 7 until the set number of iterations G is reached or the degree of adaptation reaches the desired level.

[0111] Step 9: Output the mapping results

[0112] After the algorithm ends, the solution with the highest fitness is selected as the final virtual node and physical node mapping solution, and the mapping relationship is output.

[0113] Summary: By combining the virtual node link degree with the genetic algorithm, the present invention realizes efficient virtual node mapping, which not only takes into account the resource adaptability, but also effectively utilizes the link degree information of the virtual node to optimize resource allocation, reduce the number of cross-host links, and ultimately improve the performance and resource utilization of the entire network.

[0114] 3. Mapping operation module: Generate configuration files and use automated tools to allocate virtual nodes to physical nodes.

[0115] Step 1: Generate a configuration file:

[0116] Based on the output of the algorithm allocation module, a set of configuration files is generated, each corresponding to a physical node. The configuration file records in detail the mapping relationship of each virtual node, including its resource requirements, required computing power, memory, storage space, bandwidth and other parameters. These configuration files are usually in a standardized format (such as YAML) to facilitate subsequent processing and automated deployment.

[0117] Step 2: Push the configuration file:

[0118] Use automated operation and maintenance tools (such as Kubernetes) to push the generated configuration files to the corresponding physical nodes. The automated tools will start and configure the corresponding virtual machine instances on each physical node based on the information in the configuration files. This process can greatly reduce manual intervention and improve deployment efficiency.

[0119] Step 3: Resource Verification:

[0120] After the virtual node instance is started, the system will perform resource verification to ensure that the actual resources of each virtual node match the available resources of the physical node. At the same time, the system will also check whether the configuration of the network link is as expected to ensure the validity of the data transmission path. If a resource mismatch or link problem is found, the system will record the abnormal information and automatically adjust the mapping to solve the problem.

[0121] Step 4: Monitoring and dynamic adjustment:

[0122] After the mapping is completed, the mapping operation module will continue to monitor the operation status and resource usage of the virtual nodes. Through real-time monitoring of the system, the performance data of each virtual node is analyzed to ensure the stability of network services. If the resource utilization rate of some physical nodes is found to be too high or too low during operation, the module will make necessary adjustments based on real-time feedback information to optimize resource allocation and network performance. For example, some virtual nodes can be reallocated to physical nodes with lighter loads, or additional virtual instances can be started to meet demand when necessary.

[0123] Step 5: Feedback and Optimization:

[0124] The module also includes a feedback mechanism that can regularly evaluate the effectiveness of the mapping strategy based on the operation results. By analyzing historical data and resource usage, the mapping operation module will provide optimization suggestions for future resource allocation and mapping strategies, thereby continuously improving the overall efficiency of the system.

[0125] Summarize:

[0126] Through the implementation of this mapping operation module, the present invention can ensure efficient matching of virtual nodes and physical nodes, optimize data transmission paths, maximize the utilization of physical resources, and simultaneously achieve dynamic adjustment and optimization, so that the network system has higher flexibility and stability.

[0127] 4. Dynamic allocation module: Through the particle swarm optimization algorithm, the module can effectively find the global optimal allocation plan during the migration of virtual nodes, reduce the risk of overload of physical nodes, reduce the communication overhead across host links, and improve system performance.

[0128] 1. Definition:

[0129] Particle xi: represents an allocation scheme from virtual node V to physical node P.

[0130] Particle velocity v i : represents particle x i Update direction and movement speed.

[0131] Fitness function f(x i ): used to evaluate the quality of the particle allocation scheme. By minimizing the system

[0132] The overall load difference.

[0133] Global optimal solution g * : The globally best virtual node allocation scheme so far.

[0134] Individual optimal solution p i : The optimal solution found by particle i itself.

[0135] 2. Fitness function:

[0136] f(x i ) It is necessary to measure the load balancing status of each physical node, which is represented by the load index of the physical node:

[0137]

[0138] in:

[0139] · is the physical node P j The current load index of

[0140] · is the average load of all physical nodes.

[0141] The goal is to minimize the deviation of each physical node's load from the average load, ensuring the overall load balance of the system.

[0142] 3. Dynamically adjust algorithm steps

[0143] Step 1: Initialization

[0144] (1) Initialize the particle swarm: Randomly generate a particle swarm {x 1 ,x 2 ,…,x n}, where each particle represents a mapping from a virtual node to a physical node.

[0145] (2) Initialization speed: Assign an initial speed v to each particle i , the initial value can be a random decimal.

[0146] (3) Calculate the initial fitness: Through the fitness function f(x i ) Calculate the fitness of each particle.

[0147] (4) Set the initial individual optimal solution and the global optimal solution: take the current state of each particle as its individual optimal solution p i , and select the particle with the best fitness as the global optimal solution g * .

[0148] Step 2: Particle Swarm Update

[0149] In each iteration, the particle is updated according to its velocity and position as follows:

[0150] (1) Update speed:

[0151] The particle velocity update formula is:

[0152] v i(t+1) =w·v i (t)+c 1 ·r 1 ·(p i -x i (t))+c 2 ·r 2 ·(g * -x i (t))

[0153] in:

[0154] w is the inertia weight, which is used to control the smoothness of velocity updates;

[0155] c1 and c 2 is the learning factor, controlling the influence of the individual and global optimal solutions;

[0156] r 1 and r 2 is a random number used to increase exploration diversity;

[0157] p i is the individual optimal solution of the particle itself;

[0158] g * It is the global optimal solution.

[0159] (2) Update location:

[0160] The particle position (i.e., allocation scheme) is updated as:

[0161] x i (t+1)=x i (t)+v i (t+1)

[0162] The new position of the particle represents a new mapping scheme of the virtual nodes.

[0163] (3) Boundary constraints:

[0164] Ensure that the mapping of each virtual node is within the resource limit of the physical node. If the new position exceeds the resource limit of the physical node, the position of the particle is corrected to avoid invalid allocation schemes.

[0165] Step 3: Fitness evaluation and update

[0166] Calculate the new fitness: according to the new particle position x i (t+1), recalculate the fitness f(x i ).

[0167] Update individual optimal solution: If the current fitness of the particle is better than its individual optimal solution, update its individual optimal solution p i .

[0168] Update the global optimal solution: If the fitness of the current particle is better than the global optimal solution g * , then update the global optimal solution.

[0169] Step 4: Iteration and stopping condition

[0170] Iteration: Repeat steps 2 and 3 to continuously update the particle's velocity, position, and fitness.

[0171] Stop condition: The algorithm stops when the maximum number of iterations G is reached or the fitness of the global optimal solution no longer improves.

[0172] 4. Dynamically adjust strategies

[0173] Migration strategy: According to the global optimal solution found by the particle swarm, the virtual nodes that need to be migrated are moved from the overloaded physical nodes P j Redistribute to physical nodes P with lower load k .

[0174] Real-time adjustment: The real-time dynamic allocation module regularly runs the particle swarm optimization algorithm to analyze and adjust the mapping scheme of virtual nodes to ensure that the system maintains stable performance under high load.

[0175] Summarize:

[0176] The dynamic allocation module monitors the load status of physical nodes in real time and uses the particle swarm optimization algorithm to make quick and effective adjustments in the resource allocation process to ensure that the resources of physical nodes are used evenly and the system load is optimized. This module demonstrates its efficiency and flexibility in high-load and dynamic network environments, especially in large-scale network scenarios for broadband operators.

[0177] In summary, the four modules of the present invention, namely, resource analysis module, initial allocation module, mapping operation module and dynamic allocation module, are closely combined to form a complete set of dynamic allocation schemes for virtual network resources. This scheme not only provides a scientific basis for the mapping of virtual nodes and physical nodes of broadband operators in a multi-server environment through accurate resource demand analysis and effective fitness evaluation, but also optimizes resource allocation by using genetic algorithms and link degree information, significantly reducing the number of cross-host links, thereby improving network performance.

[0178] In actual application scenarios, broadband operators are faced with growing user demands and complex network environments. The present invention can ensure that operators maintain network stability and service quality when dealing with peak traffic through an efficient resource allocation mechanism and a dynamic adjustment strategy. At the same time, the optimized resource utilization efficiency and reduced operating costs will bring higher economic benefits and competitiveness to broadband operators.

[0179] Therefore, the present invention not only has high technical feasibility, but also shows significant commercial value in practical applications, and can provide broadband operators with practical solutions to meet their urgent needs in network management and resource optimization.

[0180] Example:

[0181] A group of 10 virtual network nodes are assigned to two physical hosts. The link topology connectivity rate of the virtual network nodes is 0.5. The topology structure is randomly generated. The resource matrix is ​​as follows:

[0182]

[0183] Calculate the number of links between each virtual node and its adjacent nodes in the virtual link as the link degree of the node:

[0184] D i =[3,6,1,2,5,3,2,3,4,1]

[0185] Take virtual node V 1 and physical node P 1 Take the example to calculate the fitness:

[0186]

[0187] Here, the reasonable fitness threshold CFR Th =5, corresponding to the situation where the virtual network nodes are evenly divided by the physical nodes. The initial allocation fitness matrix is ​​calculated as above:

[0188]

[0189] Visible CFR ij Both are greater than CFR Th , multiply the fitness by the link degree weight and optimize it through genetic algorithm to complete the initial allocation:

[0190]

[0191] The initial allocation results are as follows: the physical node P1 is allocated with virtual network nodes V2, V4, V5, V9, and V10; the physical node P2 is allocated with virtual network nodes V1, V3, V6, V7, and V8.

[0192] The resource status of the virtual network node changes, the link topology remains unchanged, and the dynamic allocation process begins.

[0193] After Update

[0194]

[0195] The initial allocation result is used as the initial position of the particle, and the fitness matrix is ​​passed in as the speed. The dynamic allocation result is iterated through the particle swarm algorithm: [2,2,2,1,1,2,1,2,2,1]. That is, the physical node P1 is allocated with virtual network nodes V4, V5, V7, and V10;

[0196] The physical node P2 is assigned virtual network nodes V1, V2, V3, V6, V8, and V9.

[0197] Simulation Performance Evaluation

[0198] 1. Test scenario: Set the number of virtual network nodes to 25, 50, 75, 100, 125, 150, 175, 200, the link connection rate to 0.5, and the topology to be randomly generated; set two physical hosts, and the total amount of resources to 150% of the total amount of virtual network node resources.

[0199] 2. Test indicators:

[0200] Number of cross-physical host resource links: records the number of cross-physical server links generated by the virtual network topology after allocation. Since the total amount of resources is fixed at 150% of the virtual network resources and the occupancy rate is fixed, the number of cross-host resource links can reflect the degree of node aggregation and link optimization after algorithm allocation. The fewer the number, the better.

[0201] Algorithm running time: Record the execution time of each algorithm on nodes of different scales to evaluate its real-time performance and efficiency. The faster the time, the better.

[0202] 3. Test algorithms: genetic algorithm, particle swarm optimization algorithm, sequential algorithm, greedy algorithm.

[0203] 4. Test results:

[0204]

[0205]

[0206]

[0207] From the above simulation test data, in terms of resource allocation, the genetic algorithm shows the least number of cross-physical host resource links and the highest resource utilization under all node numbers, while the particle swarm optimization algorithm is second. Both are significantly better than the greedy algorithm and the sequential algorithm. In terms of running time, the genetic algorithm time increases significantly as the number of nodes increases due to the computational complexity. Although the particle swarm optimization algorithm time is inferior to the greedy algorithm and the sequential algorithm, the gap is not large. In summary: In the initial allocation scenario, the use of genetic algorithms can make full use of its global search capabilities to find the optimal resource and link allocation scheme and improve system efficiency. In the dynamic allocation scenario, the particle swarm optimization algorithm takes into account both real-time performance and resource optimization effects, ensuring that it can adjust quickly when the load changes, and maintaining the stability and performance of the system. This combined optimization can not only improve resource utilization, but also effectively reduce the number of cross-host links, ensuring that the system can maintain the best state under different load conditions.

[0208] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for allocating virtual network resources in a multi-server scenario, characterized in that: A network communication system based on virtual networks and physical networks, in which virtual nodes and links in the virtual network are characterized by certain resource requirements. The resources required by virtual nodes include computing power, memory and storage, while the requirements of virtual links are bandwidth. A physical network is composed of physical nodes and physical links. Each physical node has a certain resource capacity including processor, memory and storage, and the physical link provides corresponding bandwidth resources. Use G V Represents a virtual network, which consists of a set of virtual nodes and virtual links. Each virtual node n V Has specific resource requirements V , virtual link l V With bandwidth requirement B V ; Physical Network G P Contains physical nodes and physical links, physical node n P With available resources R P , physical link l P Provide available bandwidth B P ; The core of virtual network resource mapping is to reasonably allocate virtual nodes and links to physical nodes and links. During the allocation process, the following conditions must be guaranteed: Each virtual node must be mapped to a physical node that can provide sufficient resources, that is, R P ≥R V , to meet the computing and storage requirements of virtual nodes; Each virtual link must be mapped to a physical link that can provide sufficient bandwidth, that is, B P ≥B V To ensure network communication needs; In a multi-server environment, the communication between virtual nodes can span multiple physical servers, so the number and efficiency of cross-host links become the key to system performance; In order to achieve efficient resource allocation, a mapping algorithm based on fitness is adopted. The algorithm determines whether the physical node can meet the needs of the virtual node by calculating the fitness between the virtual node and the physical node. The fitness is calculated based on the resource requirements of the virtual node and the available resources of the physical node. Only physical nodes whose fitness reaches the set threshold are considered to be valid mapping targets. During the mapping process, physical nodes with high adaptability and low load are selected to ensure balanced resource utilization. In addition, link optimization is considered simultaneously during initial allocation. By evaluating the communication needs of virtual nodes, nodes that communicate frequently are preferentially mapped to the same physical node, reducing the need for cross-host links. The allocation mechanism of virtual network resources consists of the following parts: network resource judgment, network resource quantification, link strategy optimization, initial resource allocation, and dynamic resource allocation; (1) The process of network resource judgment is achieved through a comprehensive analysis of the resource requirements of virtual nodes and the available resources of physical nodes. First, the system collects the resource requirement information of virtual nodes, including computing power, memory, storage, and bandwidth, to form a virtual node requirement matrix. At the same time, the system monitors the resource status of each physical node, including available computing power, memory, storage space, and bandwidth, to form a physical resource matrix. Next, by calculating the resource fit between virtual nodes and physical nodes, it is determined whether the physical nodes can meet the requirements of virtual nodes, providing a reliable basis for subsequent allocation and link optimization. (2) The process of network resource quantification is to express the resource requirements and resource status of virtual nodes and physical nodes with specific numerical values ​​so as to achieve accurate matching in the allocation process. First, the resource requirements of virtual nodes (such as CPU, memory, storage and bandwidth, etc.) are quantified into demand vectors, which represent the specific resource requirements of each virtual node. Second, the available resources of physical nodes are also quantified into corresponding resource vectors, which record the amount of resources they can currently provide. By establishing the virtual node demand matrix and the physical node available resource matrix, the demand and supply of network resources are clearly quantified. This process provides a numerical basis for the subsequent virtual node mapping, helps to accurately calculate the fitness and optimize the allocation of virtual nodes.

2. The method for allocating virtual network resources in a multi-server scenario according to claim 1, characterized in that: Optimize link strategy. The process of link strategy optimization is to improve data transmission efficiency and reduce latency by reducing the number of links between virtual nodes that communicate across physical hosts. First, the system analyzes the communication needs between virtual nodes and identifies virtual node pairs that communicate frequently. Then, during resource allocation, these frequently communicating virtual nodes are mapped to the same physical host as much as possible to avoid unnecessary cross-host links. When cross-host communication is unavoidable, the system will give priority to mapping physical links with sufficient bandwidth and low load to ensure the stability and efficiency of data transmission. Through this link optimization strategy, network performance is improved and resource utilization is more reasonable.

3. The method for allocating virtual network resources in a multi-server scenario according to claim 1 or 2, characterized in that: Dynamic resource allocation. The final process of dynamic resource allocation will be based on the results of these previous steps. Based on the allocation results obtained in the previous calculations, the system will generate a set of configuration files corresponding to a set of physical nodes. Each file corresponds to a physical node and contains the virtual node mapping relationship of the node. These files usually use a standardized format (such as YAML), and their contents include parameters such as the resource requirements of the virtual nodes, the allocated computing power, memory, storage space, and bandwidth. These configuration files are pushed to the corresponding physical nodes using automated operation and maintenance tools (such as Kubernetes). The automated tools read the configuration files and start and configure the corresponding virtual machine instances on each physical node based on the file contents, ensuring that the resource requirements of the virtual nodes match the resources of the physical nodes. The entire initial allocation process ensures the reasonable allocation of resources between virtual and physical nodes, optimizes data transmission paths, and maximizes the utilization of physical resources.

4. The method for allocating virtual network resources in a multi-server scenario according to claim 2, characterized in that: To optimize link configuration, the strategy analyzes the communication needs between virtual nodes and maps virtual nodes with frequent communication to the same physical node to avoid unnecessary cross-host communication. When cross-host links cannot be avoided, try to select physical links with higher link capacity and lower load for mapping to ensure the stability and efficiency of data transmission.

5. The method for allocating virtual network resources in a multi-server scenario according to claim 2, characterized in that: When optimizing the link strategy, a genetic algorithm is combined with the virtual node link degree to optimize the mapping process of virtual nodes to physical nodes; the link degree and genetic algorithm are integrated to design the algorithm: Step 1: Algorithm Input Virtual node demand matrix R V : Resource requirements of each virtual node; Physical node resource status matrix R R : Available resources of each physical node; ·Fitness threshold T: the fitness threshold used to screen valid mappings; Link degree matrix D: the link degree of each virtual node; Population size P: the size of the population in the genetic algorithm; Crossover probability C p : The probability of crossover operation; Mutation probability M p : The probability of mutation operation; Iteration number G: the number of iterations of the algorithm; Step 2: Initialize the population Generate an initial solution: Randomly generate P mapping solutions, each of which contains the mapping relationship between virtual nodes and physical nodes; Ensure that each virtual node is mapped to a physical node with a fitness greater than a threshold T; Step 3: Goodness of fit evaluation For each solution S in the population k Both represent the relationship between a set of virtual nodes and physical nodes, and calculate the fitness F(S k ); The adaptability not only considers the adaptability, but also combines the link degree, and the formula is: Among them, CFR ij is the degree of fit, is the link degree of the virtual node, λ is the weight coefficient of the link degree’s contribution to the fitness; Step 4: Choose Action - Tournament Selection In the selection phase, a tournament selection method is used to decide which individuals will enter the next generation of the population; 4.1: Setting Tournament Parameters Determine the size T of the tournament, that is, the number of individuals randomly selected from the population each time (e.g. T = 3); 4.2: Conducting a Tournament Repeat the following process until the required number of adapted individuals N is selected: Randomly select individuals: randomly select T individuals from the current population; Calculate the fitness: Calculate the fitness F(S) of these T individuals ti ); ·Select the fittest individual: select the individual with the highest fitness as the winner; 4.3: Forming a new population Add the selected adapted individuals to the new population S' until the size of the new population reaches a predetermined value; Step 5: Crossover Operation Perform a crossover operation on the selected solutions to generate new solutions; the crossover operation uses single-point crossover or multi-point crossover: Select crossover individuals: Randomly select two solutions S i and S j ; Crossover point selection: randomly determine the crossover point; Generate a new solution: keep the part before the intersection, exchange the part after the intersection, and generate a new solution; Step 6: Mutation Operation Perform mutation operations on the new solution to increase the diversity of the population; the mutation operation can be achieved by randomly selecting a virtual node and reallocating it to another physical node; Select mutation individuals: according to the mutation probability M p , randomly select individuals for mutation; ·Execute mutation: randomly change the mapping of a virtual node in the individual; Step 7: Generate a new generation of population Combine the crossover and mutation solutions with the original population to form a new population S", and select the optimal solution based on the fitness; Step 8: Iteration process Repeat steps 3 to 7 until the set number of iterations G is reached or the degree of adaptation reaches the expected level; Step 9: Output the mapping results After the algorithm ends, the solution with the highest fitness is selected as the final virtual node and physical node mapping solution, and the mapping relationship is output.

6. The method for allocating virtual network resources in a multi-server scenario according to claim 5, characterized in that: Mapping operation module: generates configuration files and uses automated tools to allocate virtual nodes to physical nodes; Step 1: Generate a configuration file: Based on the output of the algorithm allocation module, a set of configuration files is generated, each corresponding to a physical node. The configuration files record in detail the mapping relationship of each virtual node, including its resource requirements, required computing power, memory, storage space, bandwidth and other parameters. These configuration files usually adopt a standardized format (such as YAML) to facilitate subsequent processing and automated deployment. Step 2: Push the configuration file: Use automated operation and maintenance tools (such as Kubernetes) to push the generated configuration files to the corresponding physical nodes. The automated tools will start and configure the corresponding virtual machine instances on each physical node based on the information in the configuration files. This process can greatly reduce manual intervention and improve deployment efficiency. Step 3: Resource Verification: After the virtual node instance is started, the system will perform resource verification to ensure that the actual resources of each virtual node match the available resources of the physical node. At the same time, the system will also check whether the configuration of the network link is as expected to ensure the validity of the data transmission path. If a resource mismatch or link problem is found, the system will record the abnormal information and automatically adjust the mapping to solve the problem. Step 4: Monitoring and dynamic adjustment: After the mapping is completed, the mapping operation module will continue to monitor the operation status and resource usage of the virtual node; through the real-time monitoring system, the performance data of each virtual node is analyzed to ensure the stability of the network service; If the resource utilization of some physical nodes is found to be too high or too low during operation, the module will make necessary adjustments based on real-time feedback information to optimize resource allocation and network performance; for example, it can reallocate some virtual nodes to physical nodes with lighter loads, or start additional virtual instances to meet demand when necessary.

7. The method for allocating virtual network resources in a multi-server scenario according to claim 5, characterized in that: Dynamic allocation module: Through the particle swarm optimization algorithm, the module can effectively find the global optimal allocation solution during the migration of virtual nodes, reduce the risk of physical node overload, and reduce the communication overhead of cross-host links, thereby improving system performance; 1) Definition: Particle x i : represents an allocation scheme from virtual node V to physical node P; Particle velocity v i : represents particle x i Update direction and movement speed; Fitness function f(x i ): used to evaluate the quality of the particle allocation scheme; minimize the overall load difference of the system through the fitness function; Global optimal solution g * : The best virtual node allocation scheme so far; Individual optimal solution p i : The optimal solution found by particle i itself; 2) Fitness function: f(x i ) It is necessary to measure the load balancing status of each physical node, which is represented by the load index of the physical node: in: · is the physical node P j The current load index of · is the average load of all physical nodes; The goal is to minimize the deviation of each physical node's load from the average load, ensuring the overall load balance of the system; 3) Dynamically adjust algorithm steps Step 1: Initialization (1) Initialize the particle swarm: Randomly generate a particle swarm {x1, x2, .., x n }, where each particle represents a mapping from a virtual node to a physical node; (2) Initialization speed: Assign an initial speed v to each particle i , the initial value can be a random decimal; (3) Calculate the initial fitness: Through the fitness function f(x i ) Calculate the fitness of each particle; (4) Set the initial individual optimal solution and the global optimal solution: take the current state of each particle as its individual optimal solution p i , and select the particle with the best fitness as the global optimal solution g * ; Step 2: Particle Swarm Update In each iteration, the particle is updated according to its velocity and position as follows: (1) Update speed: The particle velocity update formula is: v i(t+1) =w·v i (t)+c1·r1·(p i -x i (t))+c2·r2·(g * -x i (t)) in: w is the inertia weight, which is used to control the smoothness of velocity updates; c1 and c2 are learning factors that control the influence of individual and global optimal solutions; r1 and r2 are random numbers used to increase exploration diversity; p i is the individual optimal solution of the particle itself; g * is the global optimal solution; (2) Update location: The particle position (i.e., allocation scheme) is updated as: x i (t+1)=x i (t)+v i (t+1) The new position of the particle represents a new mapping scheme of the virtual nodes; (3) Boundary constraints: Ensure that the mapping of each virtual node is within the resource limit of the physical node; if the new position exceeds the resource limit of the physical node, the position of the particle is corrected to avoid invalid allocation schemes; Step 3: Fitness evaluation and update Calculate the new fitness: according to the new particle position x i (t+1), recalculate the fitness f(x i ); Update individual optimal solution: If the current fitness of the particle is better than its individual optimal solution, update its individual optimal solution p i ; Update the global optimal solution: If the fitness of the current particle is better than the global optimal solution g * , then update the global optimal solution; Step 4: Iteration and stopping condition Iteration: Repeat steps 2 and 3 to continuously update the particle’s speed, position, and fitness; Stop condition: The algorithm stops when the maximum number of iterations G is reached or the fitness of the global optimal solution no longer improves; 4) Dynamically adjust strategies Migration strategy: According to the global optimal solution found by the particle swarm, the virtual nodes that need to be migrated are moved from the overloaded physical nodes P j Redistribute to physical nodes P with lower load k ; Real-time adjustment: The real-time dynamic allocation module regularly runs the particle swarm optimization algorithm to analyze and adjust the mapping scheme of virtual nodes to ensure that the system maintains stable performance under high load; The dynamic allocation module monitors the load status of physical nodes in real time and uses the particle swarm optimization algorithm to make fast and effective adjustments in the resource allocation process to ensure that the resources of physical nodes are evenly utilized and the system load is optimized.

Images