Cloud computing resource management data transmission system and method based on artificial intelligence

Abstract

The invention discloses a cloud computing resource management data transmission system and method based on artificial intelligence, and relates to the technical field of data transmission. The system comprises a data acquisition and correlation modeling module, an influence evaluation and simulation module, a dynamic collaborative strategy generation module and a model optimization and threshold updating module. The data acquisition and correlation modeling module acquires related data, extracts features and constructs a correlation model of resource topology and a transmission path; the influence evaluation and simulation module simulates the network state and evaluates the influence degree on data transmission based on the resource operation request features and the association model; the dynamic collaborative strategy generation module generates and executes a corresponding collaborative strategy according to the evaluation result, and records real-time network state data at the same time; and the model optimization and threshold updating module optimizes the correlation model through deviation analysis and updates a preset threshold, is applied to a subsequent process, realizes intelligent cooperation of resource management and data transmission, and improves system efficiency and stability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data transmission, in particular to a cloud computing resource management data transmission system and method based on artificial intelligence. BACKGROUND

[0002] In the cloud computing environment, resource management operations (such as virtual machine migration, container scaling, bandwidth adjustment, etc.) are key means to ensure efficient operation of the system. However, in traditional cloud computing resource management systems, resource management and data transmission management are independent modules, lacking a collaborative mechanism. When performing resource operations, the network topology, traffic pattern and data distribution will change accordingly, but the traditional system often does not evaluate the impact of such changes on key data transmission tasks (such as cross-region backup, real-time synchronization, cross-availability zone communication, etc.) in advance. This leads to problems such as data transmission path deterioration, bandwidth bottleneck, delay surge or cost out of control caused by resource operations. For example, after virtual machine migration, cross-region data flow may increase significantly, but congestion may occur due to the failure to adjust the transmission strategy in advance; if the impact on associated business data synchronization is not predicted when scaling containers, data consistency deviation may occur. At the same time, the traditional system lacks a feedback optimization mechanism after the operation, and cannot iterate the resource decision logic based on the actual transmission effect, making it difficult to adapt to the dynamic changes of the cloud computing environment. SUMMARY

[0003] The present application aims to provide a cloud computing resource management data transmission system and method based on artificial intelligence to solve the problems raised in the background.

[0004] To solve the above technical problems, the present application provides the following technical solutions: A cloud computing resource management data transmission method based on artificial intelligence, comprising the following steps: S1. Collect resource operation basic data and data transmission task data in the cloud computing environment, analyze the resource operation basic data and data transmission task data, and extract corresponding resource operation basic features and data transmission task features; based on the resource operation basic features and data transmission task features, construct an association model of resource topology and transmission path; S2. When receiving a resource operation request, extract the key features corresponding to the resource operation request, simulate the network state after the execution of the resource operation based on the key features and in combination with the association model, and obtain simulation network state data; analyze the simulation network state data and evaluate the impact of the resource operation request on data transmission; S3. According to the evaluation result of the influence degree, a corresponding dynamic coordination strategy is generated, the coordination strategy is executed, and real-time network state data in the operation process is recorded; wherein, if the influence degree is within a preset threshold, a data transmission path fine-tuning scheme is generated; if the influence degree exceeds the threshold, the coordination adjustment of resource operation and transmission task is triggered; S4. Deviation analysis is performed on the real-time network state data and the simulated network state data, thereby obtaining a corresponding deviation coefficient, and the association model of the resource topology and the transmission path is optimized based on the deviation coefficient; the preset threshold of the influence degree is updated according to the distribution characteristics of the deviation coefficient, and the optimized association model and the updated preset threshold are applied to the subsequent resource operation request processing flow.

[0005] Further, S1 includes the following content: Resource operation basic data and data transmission task data in a cloud computing environment are collected, wherein the resource operation basic data refers to core data related to the state, allocation and operation behavior of cloud computing resources (such as virtual machines and physical hosts), specifically including CPU usage rate, memory occupancy, storage IOPS, network interface bandwidth of virtual machines, total resource capacity and allocated resource amount of physical hosts, type (migration / capacity expansion / bandwidth adjustment) of resource operation, execution time, involved resource ID, and resource allocation difference before and after operation; the data transmission task data refers to core data related to the attributes, path and performance of data transmission tasks, specifically including source node ID, target node ID, data total amount, transmission protocol (TCP / UDP / QUIC), real-time transmission rate, one-way delay, packet loss rate, task priority, planned start time and planned end time of the transmission task; the collected resource operation basic data and data transmission task data are respectively subjected to corresponding feature extraction, thereby obtaining resource operation basic features and data transmission task features, and the extracted resource operation basic features and data transmission task features are respectively subjected to normalization processing, thereby forming resource operation basic feature vector Z and data transmission task feature vector S; wherein the resource operation basic features include resource load rate, resource load balancing degree, operation resource scale, operation frequency and resource dependency, and the data transmission task features include transmission efficiency, delay sensitivity, packet loss influence degree, path length coefficient and task time pressure; Take the resource operation basic feature vector Z as the core input, abstract the physical nodes in the cloud computing environment as the vertex set V of the topology graph, wherein the physical nodes include computing nodes, storage nodes, gateway nodes; abstract the network link between nodes as the edge set E, construct the resource topology graph G=(V, E, W), wherein each vertex vi in the vertex set V corresponds to a physical node, and the attribute vector is composed of the feature components related to the node in the resource operation basic feature vector Z, that is: Attr(vi)=[Zi1, Zi2,..., Zik], wherein k is the feature dimension related to the node attribute, Zi1 to Zik are feature components of the node; each edge eij in the edge set E corresponds to a physical link connecting the vertices vi and vj, and the weight vector Wij is composed of the feature components related to the link in the data transmission task feature vector S, that is: Wij=[Sij1, Sij2,..., Sijm], wherein m is the feature dimension related to the link performance, and Sij1 to Sijm are feature components of the link; For the data transmission task, based on the resource topology graph G and the data transmission task feature vector S, a transmission path feature vector P is generated, and the specific process is as follows: Obtain the source node vs and the target node vd of data transmission, search the optimal transmission path from vs to vd in the resource topology graph G based on Dijkstra algorithm, denoted as Path={vs, v1,..., vn, v_d}, wherein v1 to vn are intermediate nodes in the optimal transmission path from vs to vd; extract the feature parameters of the path Path and perform normalization processing to form the transmission path feature vector P, and P=[p1, p2, p3, p4], wherein p1 is the number of edges contained in the path; p2 is the sum of the bandwidths of all edges in the path; p3 is the maximum value of the delay of all edges in the path; p4 is the minimum value of the transmission stability of all edges in the path; An associated model M of resource topology and transmission path is constructed by using a graph neural network, and the associated model M takes the resource topology graph G and the resource operation basic feature vector Z as input, and outputs the transmission path prediction feature vector P'.

[0006] Further, S2 includes the following contents: Resource operation requests are parsed to extract key features related to the operation type, intensity, and scope of impact, forming a key feature vector R, where R = [r1, r2, r3, r4]. Here, r1 is the operation type identifier, represented using one-hot encoding; for example, virtual machine migration corresponds to r1 = [1, 0, 0], container scaling corresponds to r1 = [0, 1, 0], and bandwidth adjustment corresponds to r1 = [0, 0, 1]. r2 is the resource change amount, quantifying the resource allocation changes caused by the operation, such as changes in CPU core count or memory capacity during virtual machine migration. r3 is the set of nodes involved in the operation, recording the UUID list of physical nodes affected by the operation; for example, the source node vs and target node vd involved in virtual machine migration, i.e., r3 = {UUID_vs, UUID_vd}. r4 is the operation duration, estimating the time from start to finish of the operation, determined based on the average duration of similar historical operations. For the numerical features in the feature vector R = [r1, r2, r3, r4]... Features (such as r2 and r4) are normalized (using min-max normalization, mapped to the [0,1] interval); based on r3 and r2 in the key feature vector R, the vertex attributes of the resource topology graph G are pre-adjusted. For each node vi contained in r3, the feature components related to resource changes in its attribute vector Attr(vi) are corrected according to r2. The correction formula is: Attr'(vi)=Attr(vi)+α·r2, where α is the resource change influence coefficient, determined according to the operation type r1, such as 0.15 for virtual machine migration, 0.1 for container scaling, etc.; Attr'(vi) is the pre-adjusted node attribute vector; based on the pre-adjusted node attribute vector Attr'(vi), the pre-adjusted resource topology graph G' is obtained. The adjusted resource topology graph G' and the resource operation basic feature vector Z are input into the association model M, thereby outputting the transmission path prediction feature vector P', and P'=[p'1,p'2,p'3,p'4]; Based on the transmission path prediction feature vector P' and the pre-adjusted resource topology graph G', simulated network state data is generated, including simulated path topology, link performance parameters, and node resource load. The simulated network state data is compared with the baseline state data before operation to calculate the impact quantification index, which includes the path performance change rate δp and the service level agreement compliance risk value R_SLA, where δp = ∑ a∈[1,4] (wa·|p'a-pa| / pa), where p'a represents the a-th feature value in the transmission path prediction feature vector P', pa represents the a-th feature value in the transmission path feature vector P, and wa represents the weight of the a-th feature value, and ∑ a∈[1,4]wa=1; Service Level Agreement (SLA) compliance risk value R_SLA is calculated based on the simulated maximum latency p'3 and the SLA-required latency threshold Lth. If p'3≤Lth, then R_SLA=0; if p'3>Lth, then R_SLA=(p'3-Lth) / Lth; Based on the path performance change rate δp and the SLA compliance risk value R_SLA, the comprehensive impact index I is calculated, and I=β1·δp+β2·R_SLA, where β1 and β2 both represent weighting coefficients, and β1+β2=1.

[0007] Furthermore, S3 includes the following: Set a preset threshold I0 for the comprehensive impact index, and the preset threshold I0 is pre-configured based on historical operation and maintenance data and service level agreement requirements; compare the comprehensive impact index I with the preset threshold I0 to determine the range of the impact level. If I≤I0, it is determined that the impact level is within the preset threshold; if I>I0, it is determined that the impact level exceeds the threshold. If the impact is determined to be within a preset threshold, a data transmission path fine-tuning scheme is generated, specifically including: based on the pre-adjusted resource topology graph G' and the transmission path prediction feature vector P', searching for alternative paths within the neighborhood of the original transmission path Path, where the neighborhood is defined as a path with a difference of ≤1 in the number of edges from the original path; calculating the fine-tuning coefficient F for each alternative path, where F = γ1·p'2 + γ2·p'3 + γ3·p'4, where γ1, γ2, and γ3 represent the fine-tuning weights corresponding to p'2, p'3, and p'4, respectively, and γ1 + γ2 + γ3 = 1; selecting the alternative path with the largest fine-tuning coefficient F as the adjusted transmission path, and generating a path switching instruction; If the impact is determined to exceed the threshold, a coordinated adjustment of resource operations and transmission tasks is triggered, thereby generating a dynamic coordination strategy. The specific process is as follows: Calculate the transmission load index T for the current time period, where T = current total transmission bandwidth / maximum available network bandwidth; if T > 0.7, delay the resource operation to a time period where T ≤ 0.5, and the delay duration Δt = min{3600s, (T-0.5) / 0.2×300s}; if the resource operation involves ≥ 5 nodes or the resource change r2 ≥ 20% of the total system resources, then the original operation is split into n sub-operations, where n = ceil(I / I0), and ceil is the floor function; the resource change of each sub-operation is r2 / n, and the interval between adjacent sub-operations is ≥ 60s; adjust the duration of the resource operation to r'4 = r4×(I / I0) to reduce the resource change rate from r2 / r4 to r2 / r'4; perform hierarchical scheduling according to the priority of data transmission tasks; execute the dynamic coordination strategy and synchronously record real-time network status data.

[0008] Furthermore, S4 includes the following: Deviation analysis is performed between real-time network status data and simulated network status data. The node attribute deviation ΔAttr(vi) and link performance deviation ΔWij are calculated separately. Based on these two deviations, the corresponding comprehensive deviation coefficient C is calculated, and the corresponding calculation formula is as follows: C=[∑ vi∈V w_vi·ΔAttr(vi)+∑ eij∈E w_eij·ΔWij] / (∑ vi∈V w_vi+∑ eij∈E w_eij), Where w_vi and w_eij are the weights of node vi and link eij, respectively; the comprehensive deviation coefficient C is compared with the preset threshold C0. When C≤C0, it means that the change in network status is within the expected range, and the system can continue to manage data transmission according to the current association model; if C>C0, it means that the change in network status exceeds the acceptable range, and the association model of resource topology and transmission path needs to be optimized. The optimization of the correlation model based on the deviation coefficient is as follows: The input path feature vector P_real=[p1_real,p2_real,p3_real,p4_real] from the real-time network state data is used as the true label. The corresponding pre-adjusted resource topology graph G' and resource operation basic feature vector Z are used as input samples to form a deviation training sample pair {(G',Z),P_real}. Gradient descent is used to update the parameters of the correlation model M with the deviation loss function L as the optimization objective. The optimized model is validated using the last three historical deviation sample sets. If the average comprehensive deviation coefficient C_avg ≤ C0 on the validation set, the model optimization is confirmed to be effective; otherwise, the fine-tuning process is repeated until the condition is met. Collect the comprehensive deviation coefficients {C1, C2, ..., CN} calculated during the most recent N resource operation processes, where N ≥ 30, and construct a deviation coefficient distribution sequence. Calculate the deviation distribution statistics, including the deviation mean μ and the deviation standard deviation σ. When the deviation standard deviation σ is less than or equal to the preset standard deviation threshold σ0, it indicates that the deviation coefficient distribution is relatively concentrated, and the deviation mean μ can stably reflect the overall deviation trend. In this case, the preset threshold for the degree of influence is directly updated based on μ. The corresponding preset threshold update formula for the degree of influence is: I0_new = I0 × (1 + h·μ), where k is the threshold adjustment coefficient, which can be calibrated according to the system's sensitivity to deviation. When the deviation standard deviation σ is greater than the preset standard deviation threshold σ0, it indicates that the deviation coefficient fluctuates greatly, and the overall deviation trend is not stable enough. The deviation coefficient sequence needs to be smoothed first. Calculate the smoothing... The mean deviation after smoothing is μ' = e·μ + (1-e)·μ_prev, where e is the smoothing coefficient (e.g., e takes the value 0.6-0.8), and μ_prev is the mean deviation at the time of the last update. Then, based on the smoothed μ', the preset threshold of the degree of influence is updated, and the corresponding update formula is: I0_new = I0 × (1+h·μ'). To avoid the preset threshold of the degree of influence from exceeding a reasonable range after the update, the threshold constraint condition is set: I0_min ≤ I0_new ≤ I0_max, where I0_min is the minimum threshold specified by the service level agreement, and I0_max is the maximum threshold that the system can withstand. If the calculated I0_new < I0_min, then I0_new is forcibly set to I0_min; if I0_new > I0_max, then I0_new is forcibly set to I0_max. The optimized resource topology and transmission path association model and the updated impact threshold are deployed to the cloud resource management system to replace the original association model and threshold. When processing resource operation requests, the system automatically calls the optimized association model to predict the transmission path and judges the impact of resource operations on data transmission based on the updated threshold, and generates corresponding dynamic coordination strategies.

[0009] An artificial intelligence-based cloud computing resource management data transmission system includes: a data acquisition and correlation modeling module, an impact assessment and simulation module, a dynamic collaborative strategy generation module, and a model optimization and threshold update module; The data acquisition and correlation modeling module collects basic resource operation data and data transmission task data in the cloud computing environment, analyzes the basic resource operation data and data transmission task data to extract the corresponding basic resource operation features and data transmission task features; based on the basic resource operation features and data transmission task features, it constructs a correlation model between resource topology and transmission path. When the impact assessment and simulation module receives a resource operation request, it extracts the key features corresponding to the resource operation request. Based on the key features and combined with the correlation model, it simulates the network state after the resource operation is executed, thereby obtaining simulated network state data. The simulated network state data is analyzed to assess the degree of impact of the resource operation request on data transmission. The dynamic coordination strategy generation module generates corresponding dynamic coordination strategies based on the assessment results of the impact level, executes the coordination strategies, and records real-time network status data during the operation process. Specifically, if the impact level is within a preset threshold, a data transmission path fine-tuning scheme is generated; if the impact level exceeds the threshold, coordinated adjustment of resource operations and transmission tasks is triggered. The model optimization and threshold update module performs deviation analysis between real-time network status data and simulated network status data to obtain the corresponding deviation coefficients. Based on the deviation coefficients, it optimizes the correlation model between resource topology and transmission path. According to the distribution characteristics of the deviation coefficients, it updates the preset threshold of the degree of influence and applies the optimized correlation model and the updated preset threshold to the subsequent resource operation request processing flow.

[0010] Furthermore, the data acquisition and association modeling module includes a data acquisition unit, a feature extraction and topology construction unit, and an association model training unit; The data acquisition unit is responsible for collecting basic resource operation data and data transmission task data in the cloud computing environment; the feature extraction and topology construction unit extracts features from the collected data to generate basic resource operation feature vectors and data transmission task feature vectors; it abstracts physical nodes as vertices of the topology graph and links as edges to construct a resource topology graph containing node attributes and link weights; the association model training unit uses a graph neural network to train an association model between the resource topology and the transmission path, using the resource topology graph and the basic resource operation feature vectors as input.

[0011] Furthermore, the impact assessment and simulation module includes an operational characteristic analysis unit, a network state simulation unit, and an impact degree assessment unit; The network state simulation unit receives resource operation requests, parses and extracts key features to form a key feature vector; the network state simulation unit pre-adjusts the resource topology map based on the key feature vector, calls the correlation model to predict the transmission path characteristics after the resource operation, and generates simulated network state data; the impact assessment unit compares the simulated network state with the baseline state before the operation, and comprehensively assesses the impact of the resource operation on data transmission.

[0012] Furthermore, the dynamic collaborative strategy generation module includes a strategy hierarchical determination unit, a path fine-tuning unit, and a collaborative adjustment unit; The strategy classification and judgment unit compares the comprehensive impact index with a preset threshold to determine the range of impact level; for impact levels within the threshold, the path fine-tuning unit searches for alternative paths in the neighborhood of the original transmission path, selects the path with the optimal fine-tuning coefficient as the adjusted transmission path, and generates a path switching instruction; for impact levels exceeding the threshold, the collaborative adjustment unit triggers the collaborative adjustment of resource operations and transmission tasks, generates a dynamic collaborative strategy, and executes it.

[0013] Furthermore, the model optimization and threshold update module includes a deviation analysis unit, an association model optimization unit, and a threshold dynamic update unit; The deviation analysis unit compares the real-time network status data with the simulated network status data during the operation process, calculates the node attribute deviation and link performance deviation, and generates a comprehensive deviation coefficient. The association model optimization unit, based on the comprehensive deviation coefficient, uses the gradient descent method and takes the real-time transmission path characteristics as the true label to update the parameters of the association model. The threshold dynamic update unit dynamically updates the preset threshold of the degree of influence according to the distribution characteristics of the deviation coefficient.

[0014] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention constructs a correlation model between resource topology and transmission paths, deeply binding resource operation characteristics with data transmission characteristics, thus creating a linkage mechanism between resource adjustment and transmission optimization. This fundamentally solves problems such as path deterioration and bandwidth bottlenecks caused by the lack of coordination between the two. Addressing the deficiency of traditional systems that do not assess the transmission impact before performing resource operations, this invention simulates the network state after resource operations, calculates the path performance change rate and SLA compliance risk value, and proactively identifies risks such as transmission congestion and latency spikes that may be caused by virtual machine migration, container scaling, etc., providing a forward-looking basis for decision-making. According to this invention, it avoids passive responses; for low-impact operations, it only fine-tunes the transmission path, reducing efficiency losses caused by excessive adjustments; for high-impact operations, it triggers coordinated adjustments of resource operations and transmission tasks, ensuring data transmission stability while taking into account the timeliness of resource operations, flexibly adapting to different scenario requirements; through deviation analysis between real-time network status and simulated status, this invention continuously optimizes the parameters of the correlation model and dynamically updates the impact assessment threshold based on the deviation distribution characteristics, forming a closed loop of "prediction-execution-feedback-iteration". Compared with traditional static systems, it can autonomously adapt to dynamic factors such as resource load fluctuations and changes in business scenarios, maintaining high-efficiency operation in the long term. Attached Figure Description

[0015] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a cloud computing resource management data transmission method based on artificial intelligence according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1 The present invention provides the following technical solution: A cloud computing resource management data transmission method based on artificial intelligence includes the following steps: S1. Collect basic resource operation data and data transmission task data in the cloud computing environment, analyze the basic resource operation data and data transmission task data to extract the corresponding basic resource operation features and data transmission task features; based on the basic resource operation features and data transmission task features, construct a correlation model between resource topology and transmission path; S2. When a resource operation request is received, extract the key features corresponding to the resource operation request. Based on the key features and combined with the correlation model, simulate the network state after the resource operation is executed to obtain simulated network state data. Analyze the simulated network state data to evaluate the degree of impact of the resource operation request on data transmission. S3. Based on the assessment results of the impact level, generate corresponding dynamic coordination strategies, execute the coordination strategies, and record real-time network status data during the operation process; if the impact level is within a preset threshold, generate a data transmission path fine-tuning scheme; if the impact level exceeds the threshold, trigger coordinated adjustment of resource operations and transmission tasks. S4. Perform deviation analysis between real-time network status data and simulated network status data to obtain the corresponding deviation coefficients. Optimize the correlation model between resource topology and transmission path based on the deviation coefficients. Update the preset threshold of influence based on the distribution characteristics of the deviation coefficients, and apply the optimized correlation model and the updated preset threshold to the subsequent resource operation request processing flow.

[0018] S1 includes the following: This involves collecting basic resource operation data and data transmission task data from the cloud computing environment. Basic resource operation data refers to core data related to the status, allocation, and operational behavior of cloud computing resources (such as virtual machines and physical hosts). Specifically, this includes virtual machine CPU utilization, memory usage, storage IOPS, and network interface bandwidth; total resource capacity and allocated resources of physical hosts; type of resource operation (migration / scaling / bandwidth adjustment), execution time, involved resource IDs, and resource allocation differences before and after the operation. Data transmission task data refers to core data related to the attributes, path, and performance of data transmission tasks. Specifically, this includes the source node ID, target node ID, total data volume, and transmission protocol (TCP / UDP). The data includes data such as QUIC (real-time transmission rate), one-way latency, packet loss rate, task priority, planned start time, and planned end time. Corresponding features are extracted from the collected resource operation basic data and data transmission task data to obtain resource operation basic features and data transmission task features. These features are then normalized to form the resource operation basic feature vector Z and the data transmission task feature vector S. The resource operation basic features include resource load rate, resource load balancing, operation resource scale, operation frequency, and resource dependency, while the data transmission task features include transmission efficiency, latency sensitivity, packet loss impact, path length coefficient, and task time pressure. Using the resource operation basic feature vector Z as the core input, the physical nodes in the cloud computing environment are abstracted into a vertex set V of the topology graph, where physical nodes include compute nodes, storage nodes, and gateway nodes; the network links between nodes are abstracted into an edge set E, constructing a resource topology graph G=(V,E,W), where each vertex vi in ​​vertex set V corresponds to a physical node, and its attribute vector is composed of the feature components related to that node in the resource operation basic feature vector Z, i.e.: Attr(vi)=[Zi1,Zi2,...,Zik], where k is the feature dimension related to the node attribute, and Zi1 to Zik are the feature components of that node respectively; each edge eij in edge set E corresponds to the physical link connecting vertices vi and vj, and its weight vector Wij is composed of the feature components related to that link in the data transmission task feature vector S, i.e.: Wij=[Sij1,Sij2,...,Sijm], where m is the feature dimension related to the link performance, and Sij1 to Sijm are the feature components of that link respectively; For data transmission tasks, based on the resource topology graph G and the data transmission task feature vector S, a transmission path feature vector P is generated. The specific process is as follows: Obtain the source node vs and the destination node vd for data transmission. Based on Dijkstra's algorithm, search for the optimal transmission path from vs to vd in the resource topology graph G, denoted as Path={vs,v1,...,vn,v_d}, where v1 to vn are intermediate nodes in the optimal transmission path from vs to vd. Extract the feature parameters of the path Path and normalize them to form the transmission path feature vector P, where P=[p1,p2,p3,p4], where p1 is the number of edges in the path; p2 is the sum of the bandwidths of all edges in the path; p3 is the maximum delay of all edges in the path; and p4 is the minimum transmission stability of all edges in the path. A graph neural network is used to construct a correlation model M between resource topology and transmission path. The correlation model M takes the resource topology graph G and the basic feature vector Z of resource operation as input and outputs the transmission path prediction feature vector P'.

[0019] In this embodiment, the specific construction process of the association model M is as follows: The model input layer receives the vertex attribute matrix, edge weight matrix, and resource operation basic feature vector Z of the resource topology graph G, and converts them into a tensor form suitable for graph neural network processing. The model hidden layer consists of 3 graph convolutional layers and 2 fully connected layers. The graph convolutional layers are used to extract the structural features of the resource topology graph and the attribute features of nodes and edges. The fully connected layers are used to fuse the resource operation basic feature vector Z with the topological structure features. The calculation formula is: H (l+1) =σ(AH (l) W (l) +ZW (l) z, where H (l) Let A be the feature matrix of the l-th layer, and let W be the normalized adjacency matrix of the resource topology graph G. (l) W is the weight matrix of the l-th graph convolutional layer. (l) z is the weight matrix of the basic features of resource operations in layer l, and σ is the activation function (using the ReLU function); the model output layer outputs the predicted value P' of the transmission path feature vector P through a fully connected layer, that is: P'=σ(H (L) W (L) ), where H (L) W is the feature matrix of the last hidden layer. (L) This is the weight matrix of the output layer.

[0020] The association model M is trained using historical data with the goal of minimizing the error between the predicted value P' and the actual value P of the transmission path feature vector. The specific training process is as follows: Historical resource topology G, resource operation basic feature vector Z, and corresponding transmission path feature vector P are collected to form a training sample set {(Gi,Zi,Pi)|i=1,2,...,N}, where N is the number of samples. Mean squared error (MSE) is used as the loss function, and the Adam optimizer is used to optimize the model parameters. The learning rate is set to 0.001, and the number of iterations is 1000. In each iteration, the training dataset is randomly shuffled, and the model parameters are updated using batch gradient descent until the loss function converges (training stops when the decrease in the loss function is less than 1e-5 in 10 consecutive iterations).

[0021] S2 includes the following: Resource operation requests are parsed to extract key features related to the operation type, intensity, and scope of impact, forming a key feature vector R, where R = [r1, r2, r3, r4]. Here, r1 is the operation type identifier, represented using one-hot encoding; for example, virtual machine migration corresponds to r1 = [1, 0, 0], container scaling corresponds to r1 = [0, 1, 0], and bandwidth adjustment corresponds to r1 = [0, 0, 1]. r2 is the resource change amount, quantifying the resource allocation changes caused by the operation, such as changes in CPU core count or memory capacity during virtual machine migration. r3 is the set of nodes involved in the operation, recording the UUID list of physical nodes affected by the operation; for example, the source node vs and target node vd involved in virtual machine migration, i.e., r3 = {UUID_vs, UUID_vd}. r4 is the operation duration, estimating the time from start to finish of the operation, determined based on the average duration of similar historical operations. For the numerical features in the feature vector R = [r1, r2, r3, r4]... Features (such as r2 and r4) are normalized (using min-max normalization, mapped to the [0,1] interval); based on r3 and r2 in the key feature vector R, the vertex attributes of the resource topology graph G are pre-adjusted. For each node vi contained in r3, the feature components related to resource changes in its attribute vector Attr(vi) are corrected according to r2. The correction formula is: Attr'(vi)=Attr(vi)+α·r2, where α is the resource change influence coefficient, determined according to the operation type r1, such as 0.15 for virtual machine migration, 0.1 for container scaling, etc.; Attr'(vi) is the pre-adjusted node attribute vector; based on the pre-adjusted node attribute vector Attr'(vi), the pre-adjusted resource topology graph G' is obtained. The adjusted resource topology graph G' and the resource operation basic feature vector Z are input into the association model M, thereby outputting the transmission path prediction feature vector P', and P'=[p'1,p'2,p'3,p'4]; Based on the transmission path prediction feature vector P' and the pre-adjusted resource topology graph G', simulated network state data is generated, including simulated path topology, link performance parameters, and node resource load. The simulated network state data is compared with the baseline state data before operation to calculate the impact quantification index, which includes the path performance change rate δp and the service level agreement compliance risk value R_SLA, where δp = ∑ a∈[1,4] (wa·|p'a-pa| / pa), where p'a represents the a-th feature value in the transmission path prediction feature vector P', pa represents the a-th feature value in the transmission path feature vector P, and wa represents the weight of the a-th feature value, and ∑ a∈[1,4] wa=1; Service Level Agreement (SLA) compliance risk value R_SLA is calculated based on the simulated maximum latency p'3 and the SLA-required latency threshold Lth. If p'3≤Lth, then R_SLA=0; if p'3>Lth, then R_SLA=(p'3-Lth) / Lth; Based on the path performance change rate δp and the SLA compliance risk value R_SLA, the comprehensive impact index I is calculated, and I=β1·δp+β2·R_SLA, where β1 and β2 both represent weighting coefficients, and β1+β2=1.

[0022] S3 includes the following: Set a preset threshold I0 for the comprehensive impact index, and the preset threshold I0 is pre-configured based on historical operation and maintenance data and service level agreement requirements; compare the comprehensive impact index I with the preset threshold I0 to determine the range of the impact level. If I≤I0, it is determined that the impact level is within the preset threshold; if I>I0, it is determined that the impact level exceeds the threshold. If the impact is determined to be within a preset threshold, a data transmission path fine-tuning scheme is generated, specifically including: based on the pre-adjusted resource topology graph G' and the transmission path prediction feature vector P', searching for alternative paths within the neighborhood of the original transmission path Path, where the neighborhood is defined as a path with a difference of ≤1 in the number of edges from the original path; calculating the fine-tuning coefficient F for each alternative path, where F = γ1·p'2 + γ2·p'3 + γ3·p'4, where γ1, γ2, and γ3 represent the fine-tuning weights corresponding to p'2, p'3, and p'4, respectively, and γ1 + γ2 + γ3 = 1; selecting the alternative path with the largest fine-tuning coefficient F as the adjusted transmission path, and generating a path switching instruction; If the impact is determined to exceed the threshold, a coordinated adjustment of resource operations and transmission tasks is triggered, thereby generating a dynamic coordination strategy. The specific process is as follows: Calculate the transmission load index T for the current time period, where T = current total transmission bandwidth / maximum available network bandwidth; if T > 0.7, delay the resource operation to a time period where T ≤ 0.5, and the delay duration Δt = min{3600s, (T-0.5) / 0.2×300s}; if the resource operation involves ≥ 5 nodes or the resource change r2 ≥ 20% of the total system resources, then the original operation is split into n sub-operations, where n = ceil(I / I0), and ceil is the floor function; the resource change of each sub-operation is r2 / n, and the interval between adjacent sub-operations is ≥ 60s; adjust the duration of the resource operation to r'4 = r4×(I / I0) to reduce the resource change rate from r2 / r4 to r2 / r'4; perform hierarchical scheduling according to the priority of data transmission tasks; execute the dynamic coordination strategy and synchronously record real-time network status data.

[0023] In this embodiment, hierarchical scheduling is performed based on the priority of data transmission tasks. It is assumed that the priorities of data transmission tasks are divided into high / medium / low, with corresponding priority coefficients Pr=3 / 2 / 1. For high-priority tasks, reserved bandwidth B_reserve=task-required bandwidth×(1+I) is allocated. For medium-priority tasks, the transmission rate is limited to (1-μ×I) of the original rate, where μ=0.2. For low-priority tasks, transmission is paused and resumed from high to low priority after resource operations are completed and I≤0.5×I0.

[0024] S4 includes the following: Deviation analysis is performed between real-time network status data and simulated network status data. The node attribute deviation ΔAttr(vi) and link performance deviation ΔWij are calculated separately. Based on these two deviations, the corresponding comprehensive deviation coefficient C is calculated, and the corresponding calculation formula is as follows: C=[∑ vi∈V w_vi·ΔAttr(vi)+∑ eij∈E w_eij·ΔWij] / (∑ vi∈V w_vi+∑ eij∈E w_eij), Where w_vi and w_eij are the weights of node vi and link eij, respectively; the comprehensive deviation coefficient C is compared with the preset threshold C0. When C≤C0, it means that the change in network status is within the expected range, and the system can continue to manage data transmission according to the current association model; if C>C0, it means that the change in network status exceeds the acceptable range, and the association model of resource topology and transmission path needs to be optimized. The optimization of the correlation model based on the deviation coefficient is as follows: The input path feature vector P_real=[p1_real,p2_real,p3_real,p4_real] from the real-time network state data is used as the true label. The corresponding pre-adjusted resource topology graph G' and resource operation basic feature vector Z are used as input samples to form a deviation training sample pair {(G',Z),P_real}. Gradient descent is used to update the parameters of the correlation model M with the deviation loss function L as the optimization objective. The optimized model is validated using the last three historical deviation sample sets. If the average comprehensive deviation coefficient C_avg ≤ C0 on the validation set, the model optimization is confirmed to be effective; otherwise, the fine-tuning process is repeated until the condition is met. Collect the comprehensive deviation coefficients {C1, C2, ..., CN} calculated during the most recent N resource operation processes, where N ≥ 30, and construct a deviation coefficient distribution sequence. Calculate the deviation distribution statistics, including the deviation mean μ and the deviation standard deviation σ. When the deviation standard deviation σ is less than or equal to the preset standard deviation threshold σ0, it indicates that the deviation coefficient distribution is relatively concentrated, and the deviation mean μ can stably reflect the overall deviation trend. In this case, the preset threshold for the degree of influence is directly updated based on μ. The corresponding preset threshold update formula for the degree of influence is: I0_new = I0 × (1 + h·μ), where k is the threshold adjustment coefficient, which can be calibrated according to the system's sensitivity to deviation. When the deviation standard deviation σ is greater than the preset standard deviation threshold σ0, it indicates that the deviation coefficient fluctuates greatly, and the overall deviation trend is not stable enough. The deviation coefficient sequence needs to be smoothed first. Calculate the smoothing... The mean deviation after smoothing is μ' = e·μ + (1-e)·μ_prev, where e is the smoothing coefficient (e.g., e takes the value 0.6-0.8), and μ_prev is the mean deviation at the time of the last update. Then, based on the smoothed μ', the preset threshold of the degree of influence is updated, and the corresponding update formula is: I0_new = I0 × (1+h·μ'). To avoid the preset threshold of the degree of influence from exceeding a reasonable range after the update, the threshold constraint condition is set: I0_min ≤ I0_new ≤ I0_max, where I0_min is the minimum threshold specified by the service level agreement, and I0_max is the maximum threshold that the system can withstand. If the calculated I0_new < I0_min, then I0_new is forcibly set to I0_min; if I0_new > I0_max, then I0_new is forcibly set to I0_max. The optimized resource topology and transmission path association model and the updated impact threshold are deployed to the cloud resource management system to replace the original association model and threshold. When processing resource operation requests, the system automatically calls the optimized association model to predict the transmission path and judges the impact of resource operations on data transmission based on the updated threshold, and generates corresponding dynamic coordination strategies.

[0025] An artificial intelligence-based cloud computing resource management data transmission system includes: a data acquisition and correlation modeling module, an impact assessment and simulation module, a dynamic collaborative strategy generation module, and a model optimization and threshold update module; The data acquisition and correlation modeling module collects basic resource operation data and data transmission task data in the cloud computing environment, analyzes the basic resource operation data and data transmission task data to extract the corresponding basic resource operation features and data transmission task features; based on the basic resource operation features and data transmission task features, it constructs a correlation model between resource topology and transmission path. When the impact assessment and simulation module receives a resource operation request, it extracts the key features corresponding to the resource operation request. Based on the key features and combined with the correlation model, it simulates the network state after the resource operation is executed, thereby obtaining simulated network state data. The simulated network state data is analyzed to assess the degree of impact of the resource operation request on data transmission. The dynamic coordination strategy generation module generates corresponding dynamic coordination strategies based on the assessment results of the impact level, executes the coordination strategies, and records real-time network status data during the operation process. Specifically, if the impact level is within a preset threshold, a data transmission path fine-tuning scheme is generated; if the impact level exceeds the threshold, coordinated adjustment of resource operations and transmission tasks is triggered. The model optimization and threshold update module performs deviation analysis between real-time network status data and simulated network status data to obtain the corresponding deviation coefficients. Based on the deviation coefficients, it optimizes the correlation model between resource topology and transmission path. According to the distribution characteristics of the deviation coefficients, it updates the preset threshold of the degree of influence and applies the optimized correlation model and the updated preset threshold to the subsequent resource operation request processing flow.

[0026] The data acquisition and association modeling module includes a data acquisition unit, a feature extraction and topology construction unit, and an association model training unit; The data acquisition unit is responsible for collecting basic resource operation data and data transmission task data in the cloud computing environment; the feature extraction and topology construction unit extracts features from the collected data to generate basic resource operation feature vectors and data transmission task feature vectors; it abstracts physical nodes as vertices of the topology graph and links as edges to construct a resource topology graph containing node attributes and link weights; the association model training unit uses a graph neural network to train an association model between the resource topology and the transmission path, using the resource topology graph and the basic resource operation feature vectors as input.

[0027] The impact assessment and simulation module includes an operational characteristic analysis unit, a network state simulation unit, and an impact degree assessment unit; The network state simulation unit receives resource operation requests, parses and extracts key features to form a key feature vector; the network state simulation unit pre-adjusts the resource topology map based on the key feature vector, calls the correlation model to predict the transmission path characteristics after the resource operation, and generates simulated network state data; the impact assessment unit compares the simulated network state with the baseline state before the operation, and comprehensively assesses the impact of the resource operation on data transmission.

[0028] The dynamic collaborative strategy generation module includes a strategy hierarchical determination unit, a path fine-tuning unit, and a collaborative adjustment unit; The strategy classification and judgment unit compares the comprehensive impact index with a preset threshold to determine the range of impact level; for impact levels within the threshold, the path fine-tuning unit searches for alternative paths in the neighborhood of the original transmission path, selects the path with the optimal fine-tuning coefficient as the adjusted transmission path, and generates a path switching instruction; for impact levels exceeding the threshold, the collaborative adjustment unit triggers the collaborative adjustment of resource operations and transmission tasks, generates a dynamic collaborative strategy, and executes it.

[0029] The model optimization and threshold update module includes a deviation analysis unit, an association model optimization unit, and a threshold dynamic update unit; The deviation analysis unit compares the real-time network status data with the simulated network status data during the operation process, calculates the node attribute deviation and link performance deviation, and generates a comprehensive deviation coefficient. The association model optimization unit, based on the comprehensive deviation coefficient, uses the gradient descent method and takes the real-time transmission path characteristics as the true label to update the parameters of the association model. The threshold dynamic update unit dynamically updates the preset threshold of the degree of influence according to the distribution characteristics of the deviation coefficient.

[0030] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0031] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A cloud computing resource management data transmission method based on artificial intelligence, characterized in that: The method includes the following steps: S1. Collect basic resource operation data and data transmission task data in the cloud computing environment, analyze the basic resource operation data and data transmission task data to extract the corresponding basic resource operation features and data transmission task features; based on the basic resource operation features and data transmission task features, construct a correlation model between resource topology and transmission path; S2. When a resource operation request is received, extract the key features corresponding to the resource operation request. Based on the key features and combined with the correlation model, simulate the network state after the resource operation is executed to obtain simulated network state data. Analyze the simulated network state data to evaluate the degree of impact of the resource operation request on data transmission. S3. Based on the assessment results of the impact level, generate corresponding dynamic coordination strategies, execute the coordination strategies, and record real-time network status data during the operation process; if the impact level is within a preset threshold, generate a data transmission path fine-tuning scheme; if the impact level exceeds the threshold, trigger coordinated adjustment of resource operations and transmission tasks. S4. Perform deviation analysis between real-time network status data and simulated network status data to obtain the corresponding deviation coefficients. Optimize the correlation model between resource topology and transmission path based on the deviation coefficients. Update the preset threshold of influence based on the distribution characteristics of the deviation coefficients, and apply the optimized correlation model and the updated preset threshold to the subsequent resource operation request processing flow.

2. The cloud computing resource management data transmission method based on artificial intelligence according to claim 1, characterized in that: S1 includes the following: Collect basic data on resource operations and data transmission tasks in the cloud computing environment. The basic data on resource operations refers to the core data related to the status, allocation, and operation of cloud computing resources. Data transmission task data refers to the core data related to the attributes, path, and performance of data transmission tasks. The collected resource operation basic data and data transmission task data are subjected to corresponding feature extraction to obtain resource operation basic features and data transmission task features. The extracted resource operation basic features and data transmission task features are normalized to form the resource operation basic feature vector Z and the data transmission task feature vector S. Using the resource operation basic feature vector Z as the core input, the physical nodes in the cloud computing environment are abstracted into a vertex set V of the topology graph, and the network links between nodes are abstracted into an edge set E, constructing a resource topology graph G=(V,E,W). In the vertex set V, each vertex vi corresponds to a physical node, and its attribute vector is composed of the feature components related to the node in the resource operation basic feature vector Z, that is: Attr(vi)=[Zi1,Zi2,...,Zik], where k is the feature dimension related to the node attribute, and Zi1 to Zik are the feature components of the node respectively; in the edge set E, each edge eij corresponds to the physical link connecting vertices vi and vj, and its weight vector Wij is composed of the feature components related to the link in the data transmission task feature vector S, that is: Wij=[Sij1,Sij2,...,Sijm], where m is the feature dimension related to the link performance, and Sij1 to Sijm are the feature components of the link respectively; For data transmission tasks, based on the resource topology graph G and the data transmission task feature vector S, a transmission path feature vector P is generated. The specific process is as follows: Obtain the source node vs and the destination node vd for data transmission. Based on Dijkstra's algorithm, search for the optimal transmission path from vs to vd in the resource topology graph G, denoted as Path={vs,v1,...,vn,v_d}, where v1 to vn are intermediate nodes in the optimal transmission path from vs to vd. Extract the feature parameters of the path Path and normalize them to form the transmission path feature vector P, where P=[p1,p2,p3,p4], where p1 is the number of edges in the path; p2 is the sum of the bandwidths of all edges in the path; p3 is the maximum delay of all edges in the path; and p4 is the minimum transmission stability of all edges in the path. A graph neural network is used to construct a correlation model M between resource topology and transmission path. The correlation model M takes the resource topology graph G and the basic feature vector Z of resource operation as input and outputs the transmission path prediction feature vector P'.

3. The data transmission method for cloud resource management based on artificial intelligence according to claim 2, characterized in that: S2 includes the following: The resource operation request is parsed to extract key features related to the operation type, intensity, and scope of impact, forming a key feature vector R, where R = [r1, r2, r3, r4], where r1 is the operation type label, r2 is the resource change amount, r3 is the set of nodes involved in the operation, and r4 is the operation duration. Based on r3 and r2 in the key feature vector R, the vertex attributes of the resource topology graph G are pre-adjusted. For each node vi contained in r3, the feature components related to resource changes in its attribute vector Attr(vi) are... The correction is performed based on r2, and the correction formula is: Attr'(vi)=Attr(vi)+α·r2, where α is the resource change impact coefficient, and Attr'(vi) is the pre-adjusted node attribute vector; the pre-adjusted resource topology graph G' is obtained based on the pre-adjusted node attribute vector Attr'(vi), and the adjusted resource topology graph G' and the resource operation basic feature vector Z are input into the association model M, thereby outputting the transmission path prediction feature vector P', and P'=[p'1,p'2,p'3,p'4]; Based on the transmission path prediction feature vector P' and the pre-adjusted resource topology graph G', simulated network state data is generated, including simulated path topology, link performance parameters, and node resource load. The simulated network state data is compared with the baseline state data before operation to calculate the impact quantification index, which includes the path performance change rate δp and the service level agreement compliance risk value R_SLA, where δp = ∑ a∈[1,4] (wa·|p'a-pa| / pa), where p'a represents the a-th feature value in the transmission path prediction feature vector P', pa represents the a-th feature value in the transmission path feature vector P, and wa represents the weight of the a-th feature value, and ∑ a∈[1,4] wa=1; Service Level Agreement (SLA) compliance risk value R_SLA is calculated based on the simulated maximum latency p'3 and the SLA-required latency threshold Lth. If p'3≤Lth, then R_SLA=0; if p'3>Lth, then R_SLA=(p'3-Lth) / Lth; Based on the path performance change rate δp and the SLA compliance risk value R_SLA, the comprehensive impact index I is calculated, and I=β1·δp+β2·R_SLA, where β1 and β2 both represent weighting coefficients, and β1+β2=1.

4. The data transmission method for cloud resource management based on artificial intelligence according to claim 3, characterized in that: S3 includes the following: Set a preset threshold I0 for the comprehensive impact index, and the preset threshold I0 is pre-configured based on historical operation and maintenance data and service level agreement requirements; The comprehensive impact index I is compared with the preset threshold I0 to determine the range of the impact level. If I≤I0, the impact level is determined to be within the preset threshold. If I > I0, it is determined that the degree of influence exceeds the threshold; If the impact is determined to be within a preset threshold, a data transmission path fine-tuning scheme is generated, specifically including: based on the pre-adjusted resource topology graph G' and the transmission path prediction feature vector P', searching for alternative paths within the neighborhood of the original transmission path Path, where the neighborhood is defined as a path with a difference of ≤1 in the number of edges from the original path; calculating the fine-tuning coefficient F for each alternative path, where F = γ1·p'2 + γ2·p'3 + γ3·p'4, where γ1, γ2, and γ3 represent the fine-tuning weights corresponding to p'2, p'3, and p'4, respectively, and γ1 + γ2 + γ3 = 1; selecting the alternative path with the largest fine-tuning coefficient F as the adjusted transmission path, and generating a path switching instruction; If the impact is determined to exceed the threshold, a coordinated adjustment of resource operations and transmission tasks is triggered, thereby generating a dynamic coordination strategy. The specific process is as follows: Calculate the transmission load index T for the current time period, where T = current total transmission bandwidth / maximum available network bandwidth; if T > 0.7, delay the resource operation to a time period where T ≤ 0.5, and the delay duration Δt = min{3600s, (T-0.5) / 0.2×300s}; if the resource operation involves ≥ 5 nodes or the resource change r2 ≥ 20% of the total system resources, then the original operation is split into n sub-operations, where n = ceil(I / I0), and ceil is the floor function; the resource change of each sub-operation is r2 / n, and the interval between adjacent sub-operations is ≥ 60s; adjust the duration of the resource operation to r'4 = r4×(I / I0) to reduce the resource change rate from r2 / r4 to r2 / r'4; perform hierarchical scheduling according to the priority of data transmission tasks; execute the dynamic coordination strategy and synchronously record real-time network status data.

5. The cloud computing resource management data transmission method based on artificial intelligence according to claim 4, characterized in that: S4 includes the following: Deviation analysis is performed between real-time network status data and simulated network status data. The node attribute deviation ΔAttr(vi) and link performance deviation ΔWij are calculated separately. Based on these two deviations, the corresponding comprehensive deviation coefficient C is calculated, and the corresponding calculation formula is as follows: C=[∑ vi∈V w_vi·ΔAttr(vi)+∑ eij∈E w_eij·ΔWij] / (∑ vi∈V w_vi+∑ eij∈E w_eij), Where w_vi and w_eij are the weights of node vi and link eij, respectively; the comprehensive deviation coefficient C is compared with the preset threshold C0. When C≤C0, it means that the change in network status is within the expected range, and the system can continue to manage data transmission according to the current association model; if C>C0, it means that the change in network status exceeds the acceptable range, and the association model of resource topology and transmission path needs to be optimized. The correlation model is optimized based on the deviation coefficient. The specific optimization content is as follows: The input path feature vector P_real=[p1_real,p2_real,p3_real,p4_real] in the real network state data is used as the true label, and the corresponding pre-adjusted resource topology graph G' and resource operation basic feature vector Z are used as input samples to form deviation training sample pairs {(G',Z),P_real}. The gradient descent method is used to update the parameters of the correlation model M with the deviation loss function L as the optimization objective. The optimized model is validated using the most recent three historical deviation sample sets. If the average comprehensive deviation coefficient C_avg ≤ C0 on the validation set, the model optimization is confirmed to be effective; otherwise, the fine-tuning process is repeated until the condition is met. Collect the comprehensive deviation coefficients {C1, C2, ..., CN} calculated during the most recent N resource operation processes, where N ≥ 30, and construct a deviation coefficient distribution sequence; calculate the deviation distribution statistics, including the deviation mean μ and the deviation standard deviation σ; when the deviation standard deviation σ is less than or equal to the preset standard deviation threshold σ0, the corresponding preset threshold update formula for the degree of influence is: I0_new = I0 × (1 + h·μ), where k is the threshold adjustment coefficient; when the deviation standard deviation σ is greater than the preset standard deviation threshold σ0, the deviation coefficient sequence needs to be smoothed first, and the smoothed deviation mean μ' = e·μ + (1 - e)·μ_prev is calculated, where e is the smoothing coefficient, and μ_prev is the smoothing coefficient. ev is the mean deviation at the time of the last update. Then, based on the smoothed μ', the preset threshold for the degree of influence is updated, and the corresponding update formula is: I0_new = I0 × (1 + h·μ'). To avoid the preset threshold for the degree of influence exceeding a reasonable range after updating, a threshold constraint is set: I0_min ≤ I0_new ≤ I0_max, where I0_min is the minimum threshold specified in the service level agreement, and I0_max is the maximum threshold that the system can withstand. If the calculated I0_new < I0_min, then I0_new is forcibly set to I0_min; if I0_new > I0_max, then I0_new is forcibly set to I0_max. The optimized resource topology and transmission path association model and the updated impact threshold are deployed to the cloud resource management system to replace the original association model and threshold. When processing resource operation requests, the system automatically calls the optimized association model to predict the transmission path and judges the impact of resource operations on data transmission based on the updated threshold, and generates corresponding dynamic coordination strategies.

6. An artificial intelligence-based cloud computing resource management data transmission system, applied to the artificial intelligence-based cloud computing resource management data transmission method according to any one of claims 1-5, characterized in that: The system includes: a data acquisition and correlation modeling module, an impact assessment and simulation module, a dynamic collaborative strategy generation module, and a model optimization and threshold update module; The data acquisition and correlation modeling module collects basic resource operation data and data transmission task data in the cloud computing environment, analyzes the basic resource operation data and data transmission task data to extract the corresponding basic resource operation features and data transmission task features; and constructs a correlation model between resource topology and transmission path based on the basic resource operation features and data transmission task features. When the impact assessment and simulation module receives a resource operation request, it extracts the key features corresponding to the resource operation request. Based on the key features and combined with the correlation model, it simulates the network state after the resource operation is executed, thereby obtaining simulated network state data. The simulated network state data is analyzed to assess the degree of impact of the resource operation request on data transmission. The dynamic coordination strategy generation module generates a corresponding dynamic coordination strategy based on the assessment results of the impact level, executes the coordination strategy, and records real-time network status data during the operation process; wherein, if the impact level is within a preset threshold, a data transmission path fine-tuning scheme is generated; if the impact level exceeds the threshold, a coordinated adjustment of resource operation and transmission task is triggered. The model optimization and threshold update module performs deviation analysis between real-time network status data and simulated network status data to obtain the corresponding deviation coefficients. Based on the deviation coefficients, it optimizes the correlation model between resource topology and transmission path. According to the distribution characteristics of the deviation coefficients, it updates the preset threshold of the degree of influence and applies the optimized correlation model and the updated preset threshold to the subsequent resource operation request processing flow.

7. The cloud computing resource management data transmission system based on artificial intelligence according to claim 6, characterized in that: The data acquisition and association modeling module includes a data acquisition unit, a feature extraction and topology construction unit, and an association model training unit; The data acquisition unit is responsible for collecting basic resource operation data and data transmission task data in the cloud computing environment; the feature extraction and topology construction unit extracts features from the collected data to generate basic resource operation feature vectors and data transmission task feature vectors. Physical nodes are abstracted as vertices of the topology graph, and links are abstracted as edges to construct a resource topology graph containing node attributes and link weights; the association model training unit adopts a graph neural network, which uses the resource topology graph and the basic feature vector of resource operation as input to train the association model between resource topology and transmission path.

8. The cloud computing resource management data transmission system based on artificial intelligence according to claim 6, characterized in that: The impact assessment and simulation module includes an operational feature analysis unit, a network state simulation unit, and an impact degree assessment unit. The network state simulation unit receives resource operation requests, parses and extracts key features to form a key feature vector; the network state simulation unit pre-adjusts the resource topology map based on the key feature vector, calls the correlation model to predict the transmission path features after the resource operation, and generates simulated network state data; the impact assessment unit compares the simulated network state with the baseline state before the operation, and comprehensively assesses the impact of the resource operation on data transmission.

9. The cloud computing resource management data transmission system based on artificial intelligence according to claim 6, characterized in that: The dynamic collaborative strategy generation module includes a strategy hierarchical determination unit, a path fine-tuning unit, and a collaborative adjustment unit. The strategy classification and determination unit compares the comprehensive impact index with a preset threshold to determine the range of impact level; for impact levels within the threshold, the path fine-tuning unit searches for alternative paths in the neighborhood of the original transmission path, selects the path with the optimal fine-tuning coefficient as the adjusted transmission path, and generates a path switching instruction; for impact levels exceeding the threshold, the collaborative adjustment unit triggers the collaborative adjustment of resource operations and transmission tasks, generates a dynamic collaborative strategy, and executes it.

10. A cloud computing resource management data transmission system based on artificial intelligence according to claim 6, characterized in that: The model optimization and threshold update module includes a deviation analysis unit, an association model optimization unit, and a threshold dynamic update unit; The deviation analysis unit compares the real-time network status data with the simulated network status data during the operation process, calculates the node attribute deviation and link performance deviation, and generates a comprehensive deviation coefficient. The association model optimization unit, based on the comprehensive deviation coefficient, uses the gradient descent method and real-time transmission path characteristics as the true label to update the parameters of the association model. The threshold dynamic update unit dynamically updates the preset threshold of the degree of influence according to the distribution characteristics of the deviation coefficient.

Images