Service resource allocation method for multi-dimensional heterogeneous environment of intelligent computing fusion network

Abstract

The invention discloses a service resource allocation method oriented to a multi-dimensional heterogeneous environment of an intelligent computing fusion network, which comprises the following steps of: receiving a computing network fusion service request ASR in the multi-dimensional heterogeneous environment, and deploying virtual nodes and virtual links to physical nodes and physical links meeting constraint conditions in the intelligent computing fusion network to obtain a plurality of allocation schemes; taking each distribution scheme as a chromosome, and taking all chromosomes as an initial population; randomly selecting a parameter from the segment length set, and randomly pairing the chromosomes in pairs; dividing each pair of chromosomes into a plurality of gene segments according to the parameters, and calculating the fitness value of each gene segment; performing segmented crossover and mutation operation on each pair of chromosomes to obtain a new individual; updating the population by adopting the new individuals, then judging whether the number of iterations is equal to a preset number of iterations, and if so, selecting the individual with the highest fitness value in the population as a resource allocation scheme of the computing network convergence service request ASR; otherwise, returning to the parameter selection step.

Description

Technical Field

[0001] The present invention relates to network resource allocation technology, and in particular to a service resource allocation method for a multi-dimensional heterogeneous environment of an intelligent computing fusion network. Background Art

[0002] With the rapid development of the interconnection of various heterogeneous networks and artificial intelligence (AI) technologies, especially the emergence of new computing-network fusion services represented by AIGC (AI Generated Content), users' demands for high-computing power, low-latency computing and network integration capabilities such as real-time audio and video generation, large-scale model training and inference have shown explosive growth. To meet this demand, the intelligent computing fusion network, as a new network architecture that deeply integrates computing and network resources, has emerged. By efficiently scheduling distributed computing power and coordinating multi-dimensional heterogeneous resources such as computing, network, and storage, it realizes the deep integration of computing and network. Based on the service mechanism of resource allocation on demand and dynamic scheduling, the intelligent computing fusion network can effectively support diverse intelligent service demands from individual users, enterprise customers to intelligent terminal devices in fields such as social media, marketing, education, and healthcare. It is of great significance for building a new digital infrastructure system and promoting the high-quality development of the digital economy.

[0003] However, the current mainstream centralized cloud architecture faces severe challenges when deploying computing-network fusion services. Users obtain services by invoking generative AI models or functional entities located in the cloud data center through the network. This architecture exposes significant latency problems in practical applications, which not only affect the user experience but also restrict the further development of latency-sensitive services. Another emerging solution is the distributed architecture based on edge computing. By using edge servers for ASP (computing-network fusion service requests), such as edge nodes equipped with high-performance computing devices (such as GPUs), the intelligent computing fusion network can shorten the physical distance between users and services by sinking computing resources near users, thereby achieving ultra-low latency, protecting private data, reducing bandwidth consumption, and improving overall energy efficiency, as Figure 1 shown.

[0004] Some methods systematically traverse edge servers through methods such as exhaustive search or dynamic programming to ensure finding the global optimal solution for the resource allocation of computing-network convergence services. Such algorithms are usually based on mathematical models and are solved strictly according to optimization theory, capable of providing optimal solutions. Although this method can find the global optimal solution for the resource allocation of computing-network convergence services, its computational complexity is extremely high because service resource allocation often involves a large number of heterogeneous networks (with flexible and variable topological structures and scales), multi-dimensional heterogeneous resources (such as computing, storage, bandwidth, etc.), and complex constraint conditions (such as latency limitations, etc.). The time complexity of solving by exhaustive search or dynamic programming usually increases exponentially, and the computing time and resource consumption increase rapidly, making it difficult to meet the real-time requirements in practical applications. In addition, this method may not be able to actually run in large-scale scenarios due to excessive memory consumption or long computing time, weakening the response ability of computing-network convergence services in the intelligent computing convergence network and reducing the service acceptance rate and user experience.

[0005] To reduce the solution time, another type of method adopts technical solutions based on experience and rules to quickly provide feasible solutions or sub-optimal solutions for resource allocation. The core of these methods lies in using heuristic rules or evaluation functions to sort or evaluate the priority of edge servers and resource requirements, thereby guiding the search of the solution space and avoiding the comprehensive traversal of all possible solutions by traditional methods.

[0006] Firstly, this type of method highly depends on the design of heuristic rules, and heuristic rules often require the support of domain experts' experience. If the designers do not comprehensively understand the requirements of computing-network convergence services or the characteristics of the intelligent computing convergence network, the rules may not be reasonable, resulting in a decline in the performance of the resource allocation scheme and affecting resource utilization. Secondly, the quality of the solution lacks stability. In simple scenarios, relatively good solutions may be found, but in complex scenarios, large fluctuations are likely to occur, and even deviate far from the global optimal solution. In addition, this type of method usually lacks mathematical theory support and is difficult to provide clear performance guarantees or theoretical bounds for the quality of the solution, increasing the uncertainty of the application. Summary of the Invention

[0007] Aiming at the above deficiencies in the prior art, the service resource allocation method for the multi-dimensional heterogeneous environment of the intelligent computing convergence network provided by the present invention solves the problems of low resource utilization rate and low service reception rate of the existing methods.

[0008] To achieve the above invention purpose, the technical solution adopted by the present invention is:

[0009] Provide a service resource allocation method for the multi-dimensional heterogeneous environment of the intelligent computing convergence network, which includes the steps:

[0010] S1. Receive the computing-network convergence service request ASR in a multi-dimensional heterogeneous environment. The computing-network convergence service request ASR includes multiple virtual nodes and multiple virtual links;

[0011] S2. Deploy the virtual nodes and virtual links to the physical nodes and physical links that meet the constraint conditions in the intelligent computing convergence network to obtain multiple allocation schemes including node mapping and link mapping;

[0012] S3. Take each allocation scheme as an initial chromosome, and all chromosomes as the initial population;

[0013] S4. Randomly select a parameter from the set of segment lengths, and randomly pair the chromosomes in pairs; according to the parameter, divide each pair of chromosomes into multiple gene segments, and calculate the fitness value of each gene segment;

[0014] S5. According to the fitness of the gene segments, perform segmented crossover and mutation operations on each pair of chromosomes to obtain new individuals;

[0015] S6. Update the population with the new individuals, and then determine whether the number of iterations is equal to the preset number of iterations. If so, go to step S7; otherwise, return to step S4;

[0016] S7. Select the allocation scheme composed of the node mapping and link mapping corresponding to the individual with the highest fitness value in the population as the resource allocation scheme for the computing-network convergence service request ASR.

[0017] Furthermore, step S2 further includes:

[0018] S21. Randomly select physical nodes that meet the constraint conditions in the intelligent computing convergence network to perform one-to-one mapping with each virtual node to obtain a node mapping scheme;

[0019] S22. According to the physical network topology composed of the selected physical nodes, use the k-shortest paths algorithm to calculate k shortest paths based on the physical network topology for each virtual link,

[0020] S23. Among the k shortest paths, filter out all candidate paths whose bandwidth meets the requirements of the virtual link, and randomly select one candidate path as the link mapping of the virtual link;

[0021] S24. Take the node mapping scheme and the corresponding link mapping as an allocation scheme, and repeat steps S21 to S23 until multiple allocation schemes are obtained.

[0022] Furthermore, the expression for calculating the fitness value of the gene segment is:

[0023]

[0024] Among them, S i is a gene segment of any chromosome of the i-th computing-network convergence service request ASR; (S i ) is the fitness value of S i ; α is a predetermined weight for mapping the reward and link cost ratio; β is a predetermined weight for resource coordination efficiency; is the mapping reward of S i ; is the link cost of S i ; is the resource coordination efficiency of the resource allocation scheme ; n v and n p are virtual node and physical node respectively; is the set composed of all node mappings of the i-th computing-network convergence service request ASR; is the resource dimension set, and C, G, M, S, B are all resource dimensions; is the total sum of all resources required by n v ; is the total sum of all resource capacities of n p ; A(n p ) is the resource capacity of the physical node on A; A is any dimension in the resource dimension set; A(n v ) is the resource demand of the virtual node n v on A; δ is a constant.

[0025] Furthermore, the method for performing segment crossover on chromosomes includes:

[0026] S51. Identify the best and worst gene segments in each chromosome according to the fitness values of the gene segments;

[0027] S52. Judge whether to perform crossover operation according to the preset crossover probability. If so, go to step S53; otherwise, end the crossover operation;

[0028] S53. In each pair of chromosomes, each chromosome exchanges its best gene segment with the corresponding segment of the other chromosome and exchanges its worst gene segment with the corresponding segment of the other chromosome;

[0029] S54. Judge whether there are duplicate physical nodes in the new individuals obtained after crossover. If so, go to step S55; otherwise, output the new individuals;

[0030] S55. Search whether there are non-duplicate physical nodes in its parent generation. If so, go to step S56; otherwise, delete the current new individual;

[0031] S56. Randomly select a non-repeated physical node to replace any repeated physical node in the new individual, and output the updated new individual.

[0032] Furthermore, the method for performing mutation operation on the chromosome after crossover operation includes:

[0033] A1. Take the non-optimal gene fragment part of the chromosome after mutation operation as the mutation selection range;

[0034] A2. Select mutation points within the mutation selection range of each chromosome according to the preset mutation probability;

[0035] A3. Adopt the method of random remapping, and in the physical network composed of physical nodes in the intelligent computing and fusion network, select physical nodes and physical links that meet the resource constraints and deploy them at the mutation points to obtain new individuals.

[0036] Furthermore, between step S5 and step S6, it also includes: among all the obtained new individuals, adopt the elite selection strategy, and take the K new individuals with the largest fitness value as the final new individuals.

[0037] Furthermore, the constraint conditions include:

[0038] Resource demand constraint of the computing and network fusion service request ASR:

[0039]

[0040] where n v and n p are virtual node and physical node respectively; is the set composed of all node mappings of the i-th computing and network fusion service request ASR; C(n v ) and C(n p ) are the CPU resource capacities corresponding to n v and n p respectively;

[0041] GPU resource capacity constraint of physical node and virtual node:

[0042]

[0043] where G(n v ) and G(n p ) are the GPU resource capacities corresponding to n v and n p respectively;

[0044] Memory resource demand constraint of physical node and virtual node:

[0045]

[0046] Among them, M(n v ) and M(n p ) are the memory resource requirements corresponding to n v and n p respectively;

[0047] Storage resource requirement constraints for physical nodes and virtual nodes:

[0048]

[0049] Among them, S(n v ) and S(n p ) are the storage resource requirements corresponding to n v and n p respectively;

[0050] Bandwidth resource capacity constraints for physical links and virtual links:

[0051]

[0052] Among them, e p and e v are physical links and virtual links respectively; E p is the set of all physical links in the intelligent computing fusion network; is the set of all virtual links mapped to the physical link e p ; B(e v ) and B(e p ) are the bandwidth resource requirements corresponding to e v and e p respectively;

[0053] Delay constraints for all physical links mapped by virtual links:

[0054]

[0055] Among them, is the set of all physical links mapped to e v ; D(e p ) and D(e v ) are the delay of e p and the maximum allowable delay of e v respectively.

[0056] Furthermore, the constraint conditions also include a reward function constraint and a link overhead constraint, which are respectively:

[0057]

[0058]

[0059] Among them, is the i-th computing-network integrated service request ASR; is 's reward function; ω c , ω g , ω m , ω s and ω b are the weights of C(n v ), G(n v ), M(n v ), S(n v ) and B(e v ), respectively; is the virtual node set of the i-th computing-network integrated service request ASR; N p is the physical node set in the intelligent computing integrated network; is the virtual communication link set of the i-th computing-network integrated service request ASR; is 's mapping solution set to the undirected graph G p formed by the intelligent computing integrated network; is 's link overhead function. The mapping solution includes node mapping and link mapping. The solutions or mapping solutions mentioned below are all a complete chromosome, that is, an allocation scheme.

[0060] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0061] 1. Low deployment and operation and maintenance costs. This solution is applicable to general server clusters and GPU clusters, does not depend on specific hardware or software platforms, and its implementation process can make full use of existing software and hardware resources and simply dock with existing edge computing systems. It only needs to receive service requests and sense the topology and resource information of the underlying network, without changing server configurations and network configurations, without additional investment in hardware facilities, and the optimized results output by the method can be connected to the intelligent computing integrated network controller for execution, thus significantly reducing the complexity of deployment and operation and maintenance.

[0062] 2. High solution efficiency. This solution significantly improves the exploration efficiency and convergence speed of the algorithm and can quickly generate high-quality solutions. Through the combination of chromosome solution space segmentation technology, crossover and mutation strategies, and elite selection mechanisms, this solution shows excellent performance in solving complex problems. This solution can comprehensively consider link overhead, rewards, and resource utilization rates to ensure that the solution is comprehensively optimal in all key indicators. At the same time, its efficient optimization mechanism also enables it to handle complex problems with multiple constraints and multiple objectives, and it is a solution method with both rapidity and efficiency.

[0063] 3. Strong scalability. Whether it is a small edge node cluster or a large distributed server cluster, this solution can adapt to it without additional complex environment configuration. In addition, this solution allows flexible adjustment of parameters (such as segment length) according to different application scenarios to adapt to optimization tasks of various scales and can be easily integrated into existing edge computing systems.

[0064] 4. High resource utilization. This solution can minimize resource waste and prevent resource fragmentation problems, thus improving the overall utilization rate of physical resources. This efficient resource management ability is particularly important for scenarios with limited resources and dynamically changing demands in the intelligent computing and networking integration network.

[0065] 5. Wide range of applicable scenarios. This solution is not only applicable to service deployment and resource allocation problems in the intelligent computing and networking integration network, but also can be extended to other complex optimization scenarios, such as large-scale task scheduling, network function virtualization (NFV) mapping, cloud computing resource scheduling, etc. Its framework and method ideas have wide generality and portability. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a schematic diagram of computing and networking integration service deployment.

[0067] Figure 2 It is a flowchart of a service resource allocation method for an intelligent computing and networking integration network's multi-dimensional heterogeneous environment.

[0068] Figure 3 It is an example diagram of node mapping, edge mapping, and chromosome segmentation.

[0069] Figure 4 It is a schematic diagram of crossover operation based on segmentation.

[0070] Figure 5 It is a schematic diagram of crossover and mutation operations. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0071] The following describes the specific embodiments of the present invention to facilitate those skilled in the art of this technology to understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art of this technology, as long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions and creations using the concept of the present invention are within the scope of protection.

[0072] In this solution, the intelligent computing and networking integration network is represented as an undirected graph G p =(N p , E p ), where N p represents the set of edge physical servers, and E p represents the set of physical links. Np Any physical node n in p has a set of resource attributes {C(n p ), G(n p ), M(n p ), S(n p ), B(n p )}, representing its CPU, GPU, memory, storage, and bandwidth resource capacities respectively; E p Any physical link e in p has a bandwidth capacity of B(e p ) and a link delay of D(e p ). These resource dimensions are the main dimensions that the computing-network convergence service focuses on.

[0073] Since the computing-network convergence service needs to be distributedly deployed on edge servers, therefore, the upcoming i-th ASR is also represented as a virtual undirected graph The set of virtual nodes and the set of virtual links represent micro-services or service function entities and the internal network topology of the service respectively.

[0074] Each virtual node n in v has resource requirements {C(u v ), G(u v ), M(u v ), S(u v ), B(u v )}. Each virtual link e in v has a bandwidth requirement B(e v ) and an allowable maximum delay D(e v ). The resource allocation of the edge AIGC service can be regarded as a sequential online decision-making problem, where ASRs continuously arrive and leave, occupying the resources of the intelligent computing convergence network within a specific time. We define as the mapping solution from p to G The node mapping is one-to-one, one virtual node corresponds to one physical node, and the link mapping is The link mapping is one-to-many. One virtual link may need to be composed of multiple physical links connected end to end. Define as the set of physical links that e v is mapped to, and define as the set of virtual links that are all mapped to the physical link e p .

[0075] Refer to Figure 2 ,Figure 2 shows a service resource allocation method for the multi-dimensional heterogeneous environment of the intelligent computing fusion network; as Figure 2 shown, the method S includes steps S1 to S7.

[0076] In step S1, receive the computing network fusion service request ASR in the multi-dimensional heterogeneous environment, and the computing network fusion service request ASR includes multiple virtual nodes and multiple virtual links.

[0077] In step S2, deploy the virtual nodes and virtual links to the physical nodes and physical links that meet the constraint conditions in the intelligent computing fusion network to obtain multiple allocation schemes including node mapping and link mapping.

[0078] In this solution, node mapping indicates which physical node each functional entity (such as a virtual machine or a service image container) in the virtual network should be deployed to, and link mapping indicates which physical link paths the logical links in the virtual network should consist of, which can refer to the legend shown in the upper part. The rule of node mapping is that each virtual node is mapped one-to-one with a physical node, and the requirements of the virtual node must be met. The rule of link mapping requires that the physical path mapped by each virtual link can meet the topological connectivity and bandwidth delay constraints. For the node numbers of the physical network and the virtual network, they can be in any order, but once determined in the initialization stage, they will not be changed in the subsequent process. Figure 3 In an embodiment of the present invention, step S2 further includes:

[0079] S21. Randomly select physical nodes that meet the constraint conditions in the intelligent computing fusion network to map one-to-one with each virtual node to obtain a node mapping scheme;

[0080] S22. According to the physical network topology formed by the selected physical nodes, use the k-shortest paths algorithm to calculate k shortest paths based on the physical network topology for each virtual link,

[0081] S23. Among the k shortest paths, screen out all candidate paths whose bandwidth meets the requirements of the virtual link, and randomly select one candidate path as the link mapping of the virtual link;

[0082] S24. Take the node mapping scheme and the corresponding link mapping as an allocation scheme, and repeat steps S21 to S23 until multiple allocation schemes are obtained.

[0083] S24. Take the node mapping scheme and the corresponding link mapping as an allocation scheme, and repeat steps S21 to S23 until multiple allocation schemes are obtained.

[0084] In step S3, each allocation scheme is used as an initial chromosome, and all chromosomes form an initial population. Then, the segment length set and other hyperparameters (hyperparameters related to the genetic algorithm) are initialized. The segment length set is a constant set, such as {3, 5, 7, 9}, which is a newly added hyperparameter for this algorithm. The value of the segment length should refer to the size of the virtual network. For example, when the virtual network is small, the scale of the segment length set can be smaller, such as {3, 4, 5}; when the virtual network is large, it can take values {3, 6, 9, 12}. The specific value can be determined empirically or randomly.

[0085] In step S4, a parameter is randomly selected from the segment length set, and chromosomes are randomly paired in pairs; according to the parameter, each pair of chromosomes is divided into multiple gene segments, and the fitness value of each gene segment is calculated;

[0086] The individual fitness value can evaluate the quality of a complete solution (resource allocation scheme), while the fitness based on segmentation evaluates the quality of some genes in a chromosome. To quantify the advantages and disadvantages of the multi-dimensional resource allocation scheme, this scheme defines the total amount of resources required for a virtual node as: where represents the set of different resource dimensions. Similarly, the total resource capacity of a physical node is:

[0087] In implementation, the preferred expression for calculating the fitness value of a gene segment in this scheme is:

[0088]

[0089] where S i is a gene segment of any chromosome of the i-th computing-network fusion service request ASR; (S i ) is the fitness value of S i ; α is the predetermined weight of the mapping reward and link overhead ratio; β is the predetermined weight of the resource coordination efficiency; is the mapping reward of S i ; is the link overhead of S i ; is the resource coordination efficiency of the resource allocation scheme ; n v and n p are the virtual node and the physical node respectively; is the set composed of all node mappings of the i-th computing-network fusion service request ASR; is the resource dimension set, and C, G, M, S, B are all resource dimensions; is the total sum of all resources required by n v ; is np The total sum of all resource capacities; A(n p ) is the resource capacity of a physical node on A; A is any dimension in the set of resource dimensions; A(n v ) is the resource demand of virtual node n v on A; δ is a constant.

[0090] This solution can optimize the reward of the current mapping by maximizing the fitness value, while minimizing the link overhead and resource imbalance, thereby improving the long-term benefit and resource utilization rate.

[0091] To facilitate the understanding of "divide each pair of chromosomes into multiple gene segments according to parameters", the following is combined with Figure 3 for detailed description:

[0092] The basic idea of chromosome segmentation is to map related nodes and links into the same segment. In the invention, it is assumed that the segments are continuous and uniform. All virtual nodes are treated equally, allowing any sorting within the chromosome. Given an ASR with a node number N, a set of discrete segment lengths L = {l 1 , l 2 , …, l k} is defined in the initial stage. In each iteration, a segment length l t ∈L is randomly selected. Starting from the first virtual node, this solution divides the continuous node mapping sequence that meets the length requirement in each chromosome and its associated link mapping into segments Any remaining sequence with insufficient length is also classified as a chromosome segment. For example:

[0093] A chromosome has 9 nodes, numbered from 1, which is: 123456789. The edge can be regarded as a tuple, such as (1,2) representing the edge between virtual nodes 1 and 2. Assuming that the selected parameter (i.e., the segment length) is 4, then the first gene segment includes the mapping of nodes 1234 and the edges associated with these nodes; similarly, the second gene segment includes 5678 and their associated edges; although the length of the last segment is insufficient, it is still regarded as a chromosome segment, that is, node 9 and its associated edge.

[0094] It should be noted that the node mappings in the chromosome segmentation of this solution do not overlap with each other, but they may share some link mappings, because the nodes at both ends of the link can be assigned to different segments. Subsequently, the quality of these segments is evaluated to guide subsequent crossover and mutation operations.

[0095] Figure 3The middle and lower part shows that a chromosome includes nodes abcdef, sets its segment length to 2, its segment recombination and the edges associated with these nodes. Under this segmentation rule, the segmentation is uniquely determined. However, since a virtual edge has two nodes, these two nodes may be assigned to different segments. At this time, the edge mappings need to be added to the corresponding segments.

[0096] In step S5, according to the fitness of the gene segments, perform segmentation crossover and mutation operations on each pair of chromosomes to obtain new individuals.

[0097] In an embodiment of the present invention, the method for performing segmentation crossover on chromosomes includes:

[0098] S51. According to the fitness values of the gene segments, identify the optimal and worst gene segments in each chromosome;

[0099] S52. Determine whether to perform a crossover operation according to a preset crossover probability. If so, enter step S53; otherwise, end the crossover operation;

[0100] S53. In each pair of chromosomes, each chromosome exchanges its best gene segment with the corresponding segment in the other chromosome and exchanges its worst gene segment with the corresponding segment in the other chromosome;

[0101] S54. Determine whether there are duplicate physical nodes in the new individuals obtained after crossover. If so, enter step S55; otherwise, output the new individuals;

[0102] S55. Check whether there are non-duplicate physical nodes in its parent generation. If so, enter step S56; otherwise, delete the current new individual;

[0103] S56. Randomly select a non-duplicate physical node to replace any duplicate physical node in the new individual and output the updated new individual.

[0104] This solution performs crossover in the above manner, which can improve the exploration efficiency by using high-quality genes in complex optimization scenarios. The determination of the crossover range depends on the quality assessment of different chromosome segmentations. To facilitate understanding of this crossover process, Figure 4 An example of a crossover operation is given. In the figure, each chromosome includes 10 nodes, the segment length is 2, and each chromosome is divided into 5 gene segments.

[0105] During implementation, the method preferably used by this solution for performing mutation operations on the chromosomes after crossover includes:

[0106] A1. Use the non-optimal gene segment part of the chromosome after the mutation operation as the mutation selection range;

[0107] A2. Select mutation points within the mutation selection range of each chromosome according to a preset mutation probability;

[0108] A3. Adopt the method of random remapping. In the physical network composed of physical nodes of the intelligent computing and fusion network, select physical nodes and physical links that meet the resource constraints and deploy them at the mutation points to obtain new individuals.

[0109] For the convenience of understanding crossover, Figure 5 a mutation example is shown, which adds a crossover schematic diagram on the basis of Figure 4 In this figure, the two chromosomes at the top are the initial chromosomes before crossover, the two in the middle are the schematic diagrams of the chromosomes after crossover using steps S51 - S56 of this solution, and the last two are the chromosome structures obtained after mutation on the basis of crossover.

[0110] To improve the efficiency of crossover and mutation operations, the quality of gene fragments is evaluated (calculate the fitness value of gene fragments) before the execution of this crossover and mutation. This quality evaluation process comprehensively considers the rewards of fragments, link overhead, and resource coordination efficiency to ensure that high-quality gene fragments are preferentially retained in genetic operations.

[0111] To ensure a high-quality resource allocation scheme obtained subsequently, between step S5 and step S6, it also includes: among all the obtained new individuals, adopt the elite selection strategy, and use the K new individuals with the largest fitness value as the final new individuals.

[0112] In step S6, update the population with the new individuals, and then judge whether the number of iterations is equal to the preset number of iterations. If so, enter step S7; otherwise, return to step S4;

[0113] In step S7, select the allocation scheme composed of the node mapping and link mapping corresponding to the individual with the highest fitness value in the population as the resource allocation scheme for the computing and network fusion service request ASR.

[0114] In an embodiment of the present invention, the constraint conditions include:

[0115] Resource demand constraint of the computing and network fusion service request ASR:

[0116]

[0117] where n v and n p are the virtual node and the physical node respectively; is the set composed of all node mappings of the i-th computing and network fusion service request ASR; C(n v ) and C(n p ) are C(n v ) and np The corresponding CPU resource capacity;

[0118] GPU resource capacity constraints for physical and virtual nodes:

[0119]

[0120] where G(n v ) and G(n p ) are the GPU resource capacities corresponding to n v and n p respectively;

[0121] Memory resource demand constraints for physical and virtual nodes:

[0122]

[0123] where M(n v ) and M(n p ) are the memory resource demands corresponding to n v and n p respectively;

[0124] Storage resource demand constraints for physical and virtual nodes:

[0125]

[0126] where S(n v ) and S(n p ) are the storage resource demands corresponding to n v and n p respectively;

[0127] Bandwidth resource capacity constraints for physical and virtual links:

[0128]

[0129] where e p and e v are the physical and virtual links respectively; E p is the set of all physical links in the intelligent computing fusion network; is the set of all virtual links mapped to the physical link e p ; B(e v ) and B(e p ) are the bandwidth resource demands corresponding to e v and e p respectively;

[0130] Delay constraints for all physical links mapped by virtual links:

[0131]

[0132] Among them, is the set of all physical links mapped by e; D(e v ) and D(e p ) are the delay of e v and the maximum allowable delay of e p respectively; v

[0133] The constraint conditions also include the reward function constraint and the link overhead constraint, which are respectively:

[0134]

[0135] Among them, is the i-th computing-network convergence service request ASR; is 's reward function; ω c , ω g , ω m , ω s and ω b are the weights of C(n v ), G(n v ), M(n v ), S(n v ) and B(e v ) respectively; is the set of virtual nodes of the i-th computing-network convergence service request ASR; N p is the set of physical nodes in the intelligent computing convergence network; is the set of virtual communication links of the i-th computing-network convergence service request ASR; is 's mapping solution set to the undirected graph G p formed by the intelligent computing convergence network; is 's link overhead function.

[0136] Under the above constraints, this solution performs resource allocation, which can further improve the service reception rate, system reward, reduce system overhead, and reduce resource fragmentation.

[0137] ​When the resource allocation solution provided by this solution is applied, its deployment environment is simple and can run on general servers without relying on specific hardware environments or dedicated devices, with strong environmental adaptability. It allows for deployment at any location in the intelligent computing and networking integration network. Nodes in the intelligent computing and networking integration network can be interconnected according to any physical topology structure. As long as the communication interoperability between all nodes is ensured, the operation requirements of the method can be met. At the same time, it is necessary to be able to perceive the topology structure and resource information of the intelligent computing and networking integration network, including the status of key resources such as computing power, storage capacity, and network bandwidth, to provide comprehensive data support for the method.

[0138] In addition, it is also necessary to have the function of receiving service requests for computing and networking integration, which can handle service requests from users or applications, including information such as the resource demand logical topology structure of the computing and networking integration service. After the method runs, its output result, that is, the optimal resource mapping solution for the computing and networking integration service, can be directly connected to the intelligent computing and networking integration network controller. The controller performs specific service image deployment operations according to this mapping solution, including mapping virtual service images to physical nodes and mapping virtual links to physical links, so as to achieve decoupling from specific deployment operations.

[0139] In summary, the resource allocation method for computing and networking integration services provided by this solution minimizes resource fragmentation, reduces service deployment costs, and increases long-term average revenue through reasonable resource scheduling and load balancing strategies. At the same time, this solution can adapt to the heterogeneous and changing service requirements and resource status in the edge environment, achieve a fast solution, improve efficiency, and has good scalability, capable of adapting to the growth of network nodes, changes in service requirements, and future technology iterations, thus providing long-term stable technical support and operation optimization capabilities for computing and networking integration services.

Claims

1. A service resource allocation method for a multi-dimensional heterogeneous environment of an intelligent computing fusion network, characterized in that: Includes steps: S1. Receive a computing network integration service request ASR in a multi-dimensional heterogeneous environment, where the computing network integration service request ASR includes multiple virtual nodes and multiple virtual links; S2. Deploy virtual nodes and virtual links to physical nodes and physical links that meet the constraints in the intelligent computing fusion network, and obtain multiple allocation schemes including node mapping and link mapping; S3, each allocation scheme is regarded as an initial chromosome, and all chromosomes are regarded as the initial population; S4, randomly select a parameter from the segment length set, and randomly pair the chromosomes in pairs; according to the parameter, divide each pair of chromosomes into multiple gene segments, and calculate the fitness value of each gene segment; S5. According to the fitness of the gene fragments, perform segmented crossover and mutation operations on each pair of chromosomes to obtain new individuals; S6, using the new individual to update the population, and then judging whether the number of iterations is equal to the preset number of iterations, if so, proceeding to step S7, otherwise returning to step S4; S7. Select an allocation scheme consisting of node mapping and link mapping corresponding to the individual with the highest fitness value in the population as the resource allocation scheme for the computing network convergence service request ASR.

2. The service resource allocation method for a multi-dimensional heterogeneous environment of an intelligent computing fusion network according to claim 1 is characterized in that: Step S2 further comprises: S21, randomly select a physical node that meets the constraint conditions in the intelligent computing fusion network and perform one-to-one mapping with each virtual node to obtain a node mapping solution; S22, according to the physical network topology composed of the selected physical nodes, the k-shortest paths algorithm is used to calculate k shortest paths based on the physical network topology for each virtual link. S23, among the k shortest paths, screen out all candidate paths whose bandwidths meet the virtual link requirements, and randomly select a candidate path as the link mapping of the virtual link; S24: Take the node mapping scheme and the corresponding link mapping as an allocation scheme, and repeat steps S21 to S23 until multiple allocation schemes are obtained.

3. The service resource allocation method for a multi-dimensional heterogeneous environment of an intelligent computing fusion network according to claim 1 is characterized in that: The expression for calculating the fitness value of the gene fragment is: Among them, S i A gene segment of any chromosome of the i-th computing network fusion service request ASR; (S i ) is S i The fitness value of ; α is the predetermined weight of the mapping reward and link overhead ratio; β is the predetermined weight of the resource coordination efficiency; For S i Mapping rewards; For S i Link overhead; Allocate resources to the solution Resource coordination efficiency; n v and n p They are virtual nodes and physical nodes respectively; The set consisting of all node mappings for the i-th computing-network fusion service request ASR; is a resource dimension set, C, G, M, S, B are all resource dimensions; n v The sum of all resources required; n p The sum of all resource capacities; A(n p ) is the resource capacity of the physical node on A; A is any dimension in the resource dimension set; A(n v ) is the virtual node n v The resource demand on A; δ is a constant.

4. The service resource allocation method for a multi-dimensional heterogeneous environment of an intelligent computing fusion network according to claim 1 is characterized in that: Methods for performing segmented crossover on chromosomes include: S51, identifying the best and worst gene segments in each chromosome according to the fitness values ​​of the gene segments; S52, judging whether to perform a crossover operation according to a preset crossover probability, if so, proceeding to step S53, otherwise ending the crossover operation; S53. In each pair of chromosomes, each chromosome exchanges the best gene segment of the other chromosome with its corresponding segment, and exchanges its worst gene segment with the corresponding segment of the other chromosome; S54, determine whether there are repeated physical nodes in the new individual obtained after the crossover, if so, proceed to step S55, otherwise output the new individual; S55, check whether there is a non-duplicate physical node in its parent generation, if so, proceed to step S56, otherwise delete the current new individual; S56. Randomly select a non-duplicate physical node to replace any duplicate physical node in the new individual, and output the updated new individual.

5. The service resource allocation method for a multi-dimensional heterogeneous environment of an intelligent computing fusion network according to claim 1 is characterized in that: Methods for performing mutation operations on chromosomes after crossover operations include: A1. Use the non-optimal gene fragments in the chromosome after the mutation operation as the mutation selection range; A2. Select a mutation point within the mutation range of each chromosome according to the preset mutation probability; A3. Using the random remapping method, in the physical network composed of the physical nodes of the intelligent computing fusion network, physical nodes and physical links that meet the resource constraints are selected and deployed at the mutation points to obtain new individuals.

6. The service resource allocation method for a multi-dimensional heterogeneous environment of an intelligent computing fusion network according to claim 1 is characterized in that: Also included between step S5 and step S6: Among all the new individuals obtained, an elite selection strategy is adopted to select the K new individuals with the largest fitness values ​​as the final new individuals.

7. The method for allocating service resources for a multi-dimensional heterogeneous environment of an intelligent computing fusion network according to any one of claims 1 to 6, characterized in that: The constraints include: Resource requirement constraints for computing-network converged service requests ASR: Among them, n v and n p They are virtual nodes and physical nodes respectively; is the set of all node mappings of the i-th computing-network fusion service request ASR; C(n v ) and C(n p ) are n v and n p Corresponding CPU resource capacity; GPU resource capacity constraints for physical nodes and virtual nodes: Among them, G(n v ) and G(n p ) are n v and n p Corresponding GPU resource capacity; Memory resource requirements for physical nodes and virtual nodes: Among them, M(n v ) and M(n p ) are n v and n p The corresponding memory resource requirements; Storage resource requirements for physical nodes and virtual nodes: Among them, S(n v ) and S(n p ) are n v and n p The corresponding storage resource requirements; Bandwidth resource capacity constraints of physical links and virtual links: Among them, e p and e v They are physical link and virtual link respectively; E p It is the collection of all physical links in the intelligent computing fusion network; For all mappings to physical links e p The set of virtual links of v ) and B(e p ) are e v and e p The corresponding bandwidth resource demand; Delay constraints of all physical links to which the virtual link is mapped: in, for e v The set of all physical links mapped to; D(e p ) and D(e v ) are e p The delay and e v The maximum allowed delay.

8. The service resource allocation method for a multi-dimensional heterogeneous environment of an intelligent computing fusion network according to claim 7 is characterized in that: The constraints also include reward function constraints and link cost constraints, which are: in, Request ASR for the i-th computing-network fusion service; for The reward function of c ,ω g ,ω m ,ω s and ω b C(n v )、G(n v )、M(n v )、S(n v ) and B(e v )’s weight; N is the virtual node set for the i-th computing-network fusion service request ASR; p It is a collection of physical nodes in the intelligent computing fusion network; A set of virtual communication links for the i-th computing-network fusion service request ASR; for To the undirected graph G composed of the intelligent computing fusion network p The mapping solution set of ; for The link cost function.

Images