Multi-region data center integrated management method and system

Abstract

The invention relates to the technical field of multi-region data center management, and discloses a multi-region data center integrated management method and system, and the method comprises the steps: obtaining the physical position, business data and fault information of a data center, and constructing an adjacent matrix and an incidence matrix; obtaining a first resource group by adopting a clustering algorithm; constructing a resource availability model to obtain a resource demand, performing prediction by using a load prediction algorithm to obtain a resource prediction value, and if the resource prediction value exceeds a preset resource threshold value, triggering an allocation mechanism and obtaining an optimal scheme; calculating a fault isolation range according to the fault information, and if the isolation range contains an available data center, updating an association value to obtain an updated association matrix; optimizing by using a network topology optimization algorithm to obtain an optimization result; performing evaluation by using a preset delay evaluation model to obtain a delay influence degree and a second resource group; and generating a service deployment scheme according to the features. The method can improve the management efficiency of the multi-region data center.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-region data center management, and in particular, to an integrated management method and system for multi-region data centers. Background Art

[0002] Currently, with the rapid development of cloud computing and edge computing, the integrated management of multi-region data centers has become the core infrastructure to support globalized services. Such systems need to achieve dynamic allocation of cross-region resources, fault isolation, and low-latency communication to meet the requirements of high availability and real-time services. However, due to the wide distribution of data centers and diverse business types, how to balance resource utilization, network performance, and fault tolerance in high-dynamic load scenarios has become an urgent technical challenge.

[0003] In an existing technology, multi-region data centers usually adopt a grouping management strategy based on business types or load status. For example, data centers with the same type of business or similar load levels are divided into the same management group to achieve resource pooling and unified scheduling. Such methods rely on a centralized controller or a distributed coordination framework to dynamically allocate resources within the group according to preset rules, and trigger an intra-group isolation mechanism in case of faults to limit the scope of influence. In addition, some solutions optimize the cross-region communication path by introducing SDN (Software Defined Network) technology, and preferentially select nodes within the group for data transmission to reduce management complexity.

[0004] Problems in the Existing Technology

[0005] There is a problem of deteriorated network latency caused by the conflict between the grouping strategy and geographical proximity in the existing technology, resulting in low management efficiency of multi-region data centers. Summary of the Invention

[0006] The present invention provides an integrated management method and system for multi-region data centers to improve the management efficiency of multi-region data centers.

[0007] In a first aspect, to solve the above technical problems, the present invention provides an integrated management method for multi-region data centers, including:

[0008] Obtaining the physical location, business data, and fault information of the data center, and constructing an adjacency matrix and an association matrix according to the physical location and the business data;

[0009] Grouping according to the adjacency matrix and the association matrix by using a clustering algorithm to obtain a first resource grouping;

[0010] Construct a resource availability model based on the first resource grouping to obtain resource requirements, and use a load prediction algorithm to predict the resource requirements to obtain a resource prediction value. If the resource prediction value exceeds a preset resource threshold, trigger a resource allocation mechanism to obtain an optimal solution;

[0011] Calculate a fault isolation range based on the fault information, the proximity matrix, and the association matrix. If the isolation range includes an available data center, update the association value to obtain an updated association matrix;

[0012] Optimize according to the first resource grouping and the resource prediction value using a network topology optimization algorithm to obtain an optimization result;

[0013] Evaluate according to the optimization result using a preset delay evaluation model to obtain a delay impact degree, and obtain a second resource grouping based on the delay impact degree and the first resource grouping;

[0014] Generate a service deployment plan according to the second resource grouping, the optimal solution, the proximity matrix, and the updated association matrix.

[0015] In an alternative embodiment, the obtaining the physical locations, service data, and fault information of the data centers, and constructing a proximity matrix and an association matrix according to the physical locations and service data includes:

[0016] Obtain the physical locations, service data, and fault information of the data centers, extract features from the physical locations and the service data to obtain physical location features and service data features;

[0017] Calculate the geographical distances between the data centers according to the physical location features and construct a proximity matrix;

[0018] Calculate the service association degrees between the data centers according to the service data features and construct an association matrix.

[0019] In an alternative embodiment, the grouping according to the proximity matrix and the association matrix using a clustering algorithm to obtain a first resource grouping includes:

[0020] Perform normalization processing on the proximity matrix and the association matrix to obtain a standardized proximity matrix and a standardized association matrix;

[0021] Perform clustering according to the standardized proximity matrix to obtain proximity clustering centers;

[0022] Perform clustering according to the standardized association matrix to obtain association clustering centers;

[0023] Calculate the Euclidean distance between each data point and the neighboring cluster center according to the neighboring cluster center to obtain a proximity value;

[0024] Calculate the Euclidean distance between each data point and the associated cluster center according to the associated cluster center to obtain an association value

[0025] If the proximity value of a data point is higher than a preset proximity threshold and the association value is higher than a preset association threshold, then classify it into the group where the corresponding cluster center is located to obtain a clustering group;

[0026] If the proximity value of a data point is lower than a preset proximity threshold or the association value is lower than a preset association threshold, then regard it as an independent group;

[0027] Merge the independent group and the clustering group to obtain a first resource grouping.

[0028] In an optional implementation manner, the constructing a resource availability model according to the first resource grouping to obtain a resource requirement, and using a load prediction algorithm to predict the resource requirement to obtain a resource prediction value. If the resource prediction value exceeds a preset resource threshold, then trigger a resource allocation mechanism to obtain an optimal solution, including:

[0029] Construct a resource availability model according to the first resource grouping to obtain a resource requirement;

[0030] Use a load prediction algorithm to predict the resource requirement to obtain a resource prediction value;

[0031] Compare the resource prediction value with a preset resource threshold. If the resource prediction value exceeds the preset resource threshold, then determine that allocation is needed;

[0032] Calculate the difference according to the resource prediction value and the preset resource threshold to obtain an allocated resource amount;

[0033] Optimize according to the allocated resource amount using a linear programming algorithm to obtain an optimal solution for resource allocation.

[0034] In an optional implementation manner, the calculating a fault isolation range according to the fault information, the proximity matrix and the association matrix. If the isolation range includes an available data center, then update the association value to obtain an updated association matrix, including:

[0035] Extract the fault information according to the fault information to obtain a fault point;

[0036] Determine the isolation area of the fault point according to the fault point, the proximity matrix and the association matrix;

[0037] According to the isolation area, if the isolation area contains available points with a proximity less than a preset isolation threshold, calculate the Euclidean distance between the available points and the fault point to obtain a fault isolation range;

[0038] If the isolation range contains available data centers, update the association value to obtain an updated association matrix.

[0039] In an alternative embodiment, the optimizing according to the first resource grouping and the resource prediction value by using a network topology optimization algorithm to obtain an optimization result includes:

[0040] Obtain a network topology structure according to the first resource grouping and the resource prediction value;

[0041] Determine available communication paths according to the network topology structure;

[0042] Optimize according to the network topology structure and the communication paths by using a network topology optimization algorithm to obtain an optimization result.

[0043] In an alternative embodiment, the evaluating according to the optimization result by using a preset delay evaluation model to obtain a delay impact degree, and obtaining a second resource grouping according to the delay impact degree and the first resource grouping includes:

[0044] Extract delay data according to the optimization result;

[0045] Evaluate according to the delay data by using a preset delay evaluation model to obtain a delay impact degree;

[0046] If the delay impact degree is higher than a preset delay threshold, determine that the service is affected, readjust the grouping strategy to obtain a second resource grouping;

[0047] If the delay impact degree is lower than a preset delay threshold, determine that the service is not affected, and use the first resource grouping as the second resource grouping.

[0048] In an alternative embodiment, the generating a service deployment plan according to the second resource grouping, the optimal solution, the proximity matrix, and the updated association matrix includes:

[0049] Extract grouping distribution data according to the second resource grouping;

[0050] Calculate a service dependency according to the optimal solution and the distribution data;

[0051] Generate a service deployment plan according to the service dependency, the proximity matrix, and the updated association matrix.

[0052] In a second aspect, the present invention provides an integrated management system for multi-regional data centers, including:

[0053] A data acquisition module, configured to acquire the physical location, business data, and fault information of the data center, and construct an adjacency matrix and an association matrix according to the physical location and the business data;

[0054] A grouping module, configured to perform grouping by using a clustering algorithm according to the adjacency matrix and the association matrix to obtain a first resource grouping;

[0055] A resource optimization module, configured to construct a resource availability model according to the first resource grouping to obtain resource requirements, and use a load prediction algorithm to predict the resource requirements to obtain a resource prediction value. If the resource prediction value exceeds a preset resource threshold, a resource allocation mechanism is triggered to obtain an optimal solution;

[0056] A fault detection module, configured to calculate a fault isolation range according to the fault information, the adjacency matrix, and the association matrix. If the isolation range includes an available data center, the association value is updated to obtain an updated association matrix;

[0057] A network optimization module, configured to perform optimization by using a network topology optimization algorithm according to the first resource grouping and the resource prediction value to obtain an optimization result;

[0058] A delay evaluation module, configured to perform evaluation by using a preset delay evaluation model according to the optimization result to obtain a delay impact degree, and obtain a second resource grouping according to the delay impact degree and the first resource grouping;

[0059] A service deployment module, configured to generate a service deployment plan according to the second resource grouping, the optimal solution, the adjacency matrix, and the updated association matrix.

[0060] In a third aspect, the present invention further provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the integrated management method for multi-regional data centers described in any one of the above is implemented.

[0061] In a fourth aspect, the present invention further provides a computer-readable storage medium, which includes a stored computer program. When the computer program runs, the device where the computer-readable storage medium is located is controlled to execute the integrated management method for multi-regional data centers described in any one of the above.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] The present invention discloses an integrated management method and system for multi-region data centers. The method includes obtaining the physical location, business data, and fault information of the data centers and constructing an adjacency matrix and an association matrix; using a clustering algorithm to obtain a first resource grouping; constructing a resource availability model to obtain resource requirements, and using a load prediction algorithm to make predictions to obtain resource prediction values. If the preset resource threshold is exceeded, a deployment mechanism is triggered and an optimal solution is obtained; calculating the fault isolation range according to the fault information. If the isolation range includes available data centers, the association value is updated to obtain an updated association matrix; using a network topology optimization algorithm for optimization to obtain an optimization result; using a preset delay evaluation model for evaluation to obtain a delay impact degree and a second resource grouping; generating a business deployment plan according to the above features. This method can improve the management efficiency of multi-region data centers.

[0064] The present invention constructs a multi-dimensional collaborative optimization model, which significantly reduces the network latency of cross-region data centers and improves resource utilization while ensuring business continuity and resource elasticity. The technical derivation process is as follows: First, based on the physical location and business relevance of the data centers, an adjacency matrix and an association matrix are respectively constructed. A first resource grouping is generated through a weighted clustering algorithm, so that the nodes within the group not only satisfy geographical proximity but also have a high business association degree, thereby reducing the cross-group communication requirements. Secondly, the real-time load rates of the CPU, storage, and bandwidth of each node are dynamically calculated through a resource availability model. Combining with a long short-term memory network to predict future resource requirements. When the predicted value exceeds the threshold, a cross-group resource deployment mechanism based on the Hungarian algorithm is triggered, and nodes with geographical proximity and low business conflict are preferentially selected as resource suppliers to avoid local resource exhaustion caused by group barriers under traditional grouping strategies. Further, in the fault isolation stage, the system analyzes the topological influence range of the faulty node. If the isolation boundary includes low-latency available nodes, the weight value of the association matrix is dynamically updated, and such nodes are re-incorporated into the resource pool, which not only realizes fault isolation but also retains the availability of neighboring resources. Finally, based on the network topology optimization algorithm and the delay evaluation model, a second resource grouping and a business deployment plan are generated to ensure that high-latency sensitive services are preferentially deployed in low-latency groups, while high-throughput tasks are assigned to long-distance resource groups. The present invention effectively solves the problems of conflict between grouping strategies and geographical proximity, rigid resource deployment, and mis-isolation of faults in the prior art, and provides systematic support for the efficient management of multi-region data centers. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 is a schematic flowchart of an integrated management method for multi-region data centers provided by the first embodiment of the present invention;

[0066] Figure 2 is a schematic structural diagram of an integrated management system for multi-region data centers provided by the second embodiment of the present invention. Detailed implementation manners

[0067] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part rather than all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0068] Referring to Figure 1 , the first embodiment of the present invention provides an integrated management method for multi-region data centers, including the following steps:

[0069] S11. Obtain the physical location, service data, and fault information of the data center, and construct a proximity matrix and an association matrix according to the physical location and the service data;

[0070] S12. Group according to the proximity matrix and the association matrix by using a clustering algorithm to obtain a first resource group;

[0071] S13. Construct a resource availability model according to the first resource group to obtain resource requirements, and use a load prediction algorithm to predict the resource requirements to obtain a resource prediction value. If the resource prediction value exceeds a preset resource threshold, trigger a resource allocation mechanism to obtain an optimal solution;

[0072] S14. Calculate a fault isolation range according to the fault information, the proximity matrix, and the association matrix. If the isolation range includes an available data center, update the association value to obtain an updated association matrix;

[0073] S15. Optimize according to the first resource group and the resource prediction value by using a network topology optimization algorithm to obtain an optimization result;

[0074] S16. Evaluate according to the optimization result by using a preset delay evaluation model to obtain a delay impact degree, and obtain a second resource group according to the delay impact degree and the first resource group;

[0075] S17. Generate a service deployment plan according to the second resource group, the optimal solution, the proximity matrix, and the updated association matrix.

[0076] In step S11, obtain the physical location, service data, and fault information of the data center, and construct a proximity matrix and an association matrix according to the physical location and the service data.

[0077] In one implementation, obtaining the physical location, service data, and fault information of the data center, and constructing an adjacency matrix and an association matrix based on the physical location and the service data includes:

[0078] Obtaining the physical location, service data, and fault information of the data center, performing feature extraction on the physical location and the service data to obtain physical location features and service data features;

[0079] Calculating the geographical distance between data centers according to the physical location features, and constructing an adjacency matrix;

[0080] Calculating the service association degree between data centers according to the service data features, and constructing an association matrix.

[0081] It should be noted that in the process of feature extraction of the physical location and service data, the system first standardizes the original physical location data such as the longitude and latitude coordinates, network latency, and bandwidth quality of the data center, and converts it into a multi-dimensional vector through geocoding technology; at the same time, feature engineering is performed on the service type labels, resource occupancy rate time series data, and cross-center API call logs in the service data. The TF-IDF is used to statistically weight the frequency of service keywords, and the time series features of the service interaction mode are extracted in combination with the LSTM network. Finally, a 256-dimensional feature vector representing the service characteristics is generated. The fault information is parsed into indicators such as historical fault frequency and average recovery time for subsequent dynamic weight correction of the association matrix.

[0082] It should be noted that the calculation of the geographical distance is based on an improved spherical distance formula, and the network topology hop count and backbone network transmission rate are introduced as correction factors; the service association degree is calculated by the cosine similarity of the feature vectors, and weighted fusion is performed in combination with the service interaction frequency and data dependence intensity.

[0083] In step S12, according to the adjacency matrix and the association matrix, a clustering algorithm is used for grouping to obtain a first resource group.

[0084] In one implementation, the grouping according to the adjacency matrix and the association matrix using a clustering algorithm to obtain a first resource group includes:

[0085] Performing normalization processing on the adjacency matrix and the association matrix to obtain a standardized adjacency matrix and a standardized association matrix;

[0086] Performing clustering according to the standardized adjacency matrix to obtain an adjacency clustering center;

[0087] Performing clustering according to the standardized association matrix to obtain an association clustering center;

[0088] Calculate the Euclidean distance between each data point and the neighboring cluster center according to the neighboring cluster center to obtain a proximity value;

[0089] Calculate the Euclidean distance between each data point and the associated cluster center according to the associated cluster center to obtain an association value

[0090] If the proximity value of a data point is higher than a preset proximity threshold and the association value is higher than a preset association threshold, then classify it into the group where the corresponding cluster center is located to obtain a clustering group;

[0091] If the proximity value of a data point is lower than the preset proximity threshold or the association value is lower than the preset association threshold, then regard it as an independent group;

[0092] Merge the independent group and the clustering group to obtain a first resource grouping.

[0093] It should be noted that in step S12, the clustering algorithm uses an improved K-means algorithm to optimize resource grouping by integrating the dual constraint conditions of the proximity matrix and the association matrix. Specifically, the algorithm first normalizes the proximity matrix (geographical distance, network latency) and the association matrix (business dependency, data interaction frequency) respectively. After eliminating the dimension difference, a parallel clustering strategy is adopted: perform K-means clustering based on geographical density on the standardized proximity matrix, determine the initial number of cluster centers k1 through the elbow method, iteratively calculate the Euclidean distance between each data point and the center point and update the clustering division until the neighboring cluster center is stable; simultaneously perform similarity-driven K-means clustering on the standardized association matrix, use the cosine similarity of the business feature vector as the distance metric, and determine k2 associated cluster centers. Subsequently, the algorithm introduces a dual-threshold decision mechanism, calculates the Euclidean distance between each data point and the two types of cluster centers for each data point (the proximity value reflects geographical closeness, and the association value reflects business coupling degree). If a certain point simultaneously satisfies the proximity value ≤ preset proximity threshold (such as geographical distance ≤ 500 kilometers and latency ≤ 30 ms) and the association value ≤ preset association threshold (such as business similarity ≥ 0.7), then classify it into the corresponding joint clustering group; otherwise, mark it as an independent group (such as a geographically isolated or business-unique node). Finally, the first resource grouping is formed by merging the clustering groups and independent groups that meet the constraints, so that the nodes within the group have both geographical proximity and high business relevance, while retaining the independence of special nodes.

[0094] In step S13, according to the first resource grouping, construct a resource availability model to obtain resource requirements, and use a load prediction algorithm to predict the resource requirements to obtain a resource prediction value. If the resource prediction value exceeds a preset resource threshold, trigger a resource allocation mechanism to obtain an optimal solution.

[0095] In one implementation, building a resource availability model according to the first resource grouping to obtain resource requirements, and using a load prediction algorithm to predict the resource requirements to obtain a resource prediction value. If the resource prediction value exceeds a preset resource threshold, a resource allocation mechanism is triggered to obtain an optimal solution, including:

[0096] Building a resource availability model according to the first resource grouping to obtain resource requirements;

[0097] Using a load prediction algorithm to predict the resource requirements to obtain a resource prediction value;

[0098] Comparing the resource prediction value with the preset resource threshold. If the resource prediction value exceeds the preset resource threshold, it is determined that allocation is required;

[0099] Calculating the difference according to the resource prediction value and the preset resource threshold to obtain the allocated resource quantity;

[0100] Optimizing according to the allocated resource quantity using a linear programming algorithm to obtain an optimal solution for resource allocation.

[0101] It should be noted that in the implementation of the load prediction algorithm, the system adopts a hybrid prediction model based on time series decomposition and deep learning. By analyzing historical resource usage data, combining business cycle fluctuation characteristics and sudden event detection, a long short-term memory network is used to capture long-term dependencies, and an attention mechanism is introduced to dynamically weight key time nodes, and finally the prediction values of each resource index within the future time window are output. The input of the model is the historical load sequence intercepted by a sliding window (such as the data of the past 24 hours, with a sampling interval of 5 minutes). After normalization and feature enhancement (such as Fourier transform to extract periodic components), a multi-dimensional resource demand prediction vector is output.

[0102] It should be noted that the linear programming algorithm optimization takes minimizing the total delay and cost of cross-regional resource allocation as the objective function, and the constraint conditions include the remaining resource upper limit of the supply node, the resource gap of the demand node, network bandwidth limitation, and the maximum delay threshold specified by the SLA (Service Level Agreement). Specifically, the decision variables are defined as the resource quantities allocated by each supply node to the demand node.

[0103] In step S14, according to the fault information, the proximity matrix, and the association matrix, the fault isolation range is calculated. If the isolation range contains an available data center, the association value is updated to obtain an updated association matrix.

[0104] In one implementation, calculating the fault isolation range according to the fault information, the proximity matrix, and the association matrix. If the isolation range contains an available data center, the association value is updated to obtain an updated association matrix, including:

[0105] Extract the fault information based on the said fault information to obtain the fault point;

[0106] Determine the isolation area of the fault point according to the said fault point, the said proximity matrix and the said association matrix;

[0107] According to the said isolation area, if the isolation area contains available points with a proximity less than the preset isolation threshold, calculate the Euclidean distance between the said available points and the said fault point to obtain the fault isolation range;

[0108] If the said isolation range contains available data centers, update the association value to obtain the updated association matrix.

[0109] It should be noted that when determining the isolation area of the fault point, the system calculates the comprehensive isolation influence value between the fault point and other nodes through dynamic weighting based on the geographical distance, network delay index in the proximity matrix and the service association degree in the association matrix. For example, the proximity matrix of fault point A shows that its geographical distance from node B is 200 kilometers and the network delay is 15 ms, and the service association degree in the association matrix is 0.9 (full score 1.0). If the preset isolation threshold is geographical distance ≤ 300 kilometers and service association degree ≥ 0.7, then node B is determined as a high-impact candidate node; the system further calculates the Euclidean distance between fault point A and node B (integrating geographical and service dimensions). If this distance is lower than the preset isolation radius (such as the dynamic threshold generated by K-means clustering), then node B is included in the isolation area, and at the same time, all nodes in the proximity matrix that meet the conditions are traversed to generate the initial isolation range. If the range contains nodes marked as "available" (such as node B is not currently faulty and the load rate is lower than 60%), then trigger the association matrix update logic, and attenuate the association value between node B and fault point A by 50% to reduce subsequent scheduling dependence, so as to avoid fault spread and retain the schedulability of available resources.

[0110] In step S15, optimize according to the said first resource grouping and the said resource prediction value by using the network topology optimization algorithm to obtain the optimization result.

[0111] To facilitate the understanding of the present invention, some preferred embodiments of the present invention will be further described below.

[0112] In one implementation, the said optimizing according to the said first resource grouping and the said resource prediction value by using the network topology optimization algorithm to obtain the optimization result includes:

[0113] Obtain the network topology structure according to the said first resource grouping and the said resource prediction value;

[0114] Determine the available communication paths according to the said network topology structure;

[0115] Optimize according to the network topology structure and the communication path by using a network topology optimization algorithm to obtain an optimization result.

[0116] In another implementation, the optimizing according to the first resource grouping and the resource prediction value by using a network topology optimization algorithm to obtain an optimization result includes:

[0117] Construct a dynamic traffic prediction model according to the first resource grouping and the resource prediction value to obtain time-series correlation data;

[0118] Select a strategy according to the time-series correlation data by using a reinforcement learning algorithm to obtain an available strategy;

[0119] Optimize according to the available strategy and the time-series correlation data by using a network topology optimization algorithm to obtain an optimization result.

[0120] It should be noted that the network topology optimization algorithm is a hybrid optimization method based on graph theory and dynamic programming. By integrating the resource prediction value and the real-time state of the network topology, it dynamically adjusts the weights of communication paths and reconstructs the topology to minimize the global delay and maximize the bandwidth utilization rate. Specifically, the algorithm first maps the nodes in the first resource grouping to the vertices in the graph structure, and assigns initial weights to the edges according to the geographical distance in the proximity matrix, the service coupling degree in the association matrix, and the resource prediction value; subsequently, it traverses all available communication paths by using an improved Dijkstra algorithm, dynamically adjusts the path priorities in combination with the resource prediction value, and introduces a taboo search strategy to avoid local optima, and finally generates a topology optimization scheme including an optimal path set and backup paths.

[0121] In step S16, evaluate according to the optimization result by using a preset delay evaluation model to obtain a delay impact degree, and obtain a second resource grouping according to the delay impact degree and the first resource grouping.

[0122] In one implementation, the evaluating according to the optimization result by using a preset delay evaluation model to obtain a delay impact degree, and obtaining a second resource grouping according to the delay impact degree and the first resource grouping includes:

[0123] Extract delay data according to the optimization result;

[0124] Evaluate according to the delay data by using a preset delay evaluation model to obtain a delay impact degree;

[0125] If the delay impact degree is higher than a preset delay threshold, it is determined that the service is affected, and the grouping strategy is readjusted to obtain a second resource grouping;

[0126] If the delay impact degree is lower than the preset delay threshold, it is determined that the service is not affected, and the first resource group is used as the second resource group.

[0127] It should be noted that the preset delay evaluation model is a multi-factor evaluation system based on service sensitivity weighting and real-time network status. By integrating the optimized network topology delay data, service type priorities (and historical delay impact coefficients (weight parameters trained by a linear regression model)), the delay impact degree of each service flow under the current topology is calculated. Specifically, the model compares the measured delay value of each service flow with the preset SLA threshold. If the measured value exceeds the threshold, the comprehensive impact degree score is generated by superimposing the service weight according to the exceeding ratio. For example, if the measured delay of a financial transaction service is 65 ms (threshold 50 ms), its impact degree = (65 - 50) / 50 × 2 = 0.6. If the preset threshold is 0.5, it is determined that re-grouping is required. The model further combines the geographical redundant node distribution in the proximity matrix to dynamically adjust the grouping strategy for high-impact services, and finally generates a second resource group with priority scheduling for delay-sensitive services.

[0128] In step S17, according to the second resource group, the optimal solution, the proximity matrix, and the updated association matrix, a service deployment plan is generated.

[0129] In one implementation, the generating of the service deployment plan according to the second resource group, the optimal solution, the proximity matrix, and the updated association matrix includes:

[0130] Extract the grouping distribution data according to the second resource group;

[0131] Calculate the service dependency according to the optimal solution and the distribution data;

[0132] Generate a service deployment plan according to the service dependency, the proximity matrix, and the updated association matrix.

[0133] It should be noted that when generating a business deployment plan, the system realizes the optimal matching of services and resource groups through a multi-objective optimization model based on service dependency, geographical proximity metrics of the proximity matrix (such as latency, bandwidth), and the updated association matrix (dynamically adjusted service coupling weights). Specifically, for highly dependent services (such as real-time financial transactions), resource groups with a latency of less than 20 ms and a service coupling weight ≥ 0.8 in the association matrix are preferentially selected for deployment, and the deployment density is dynamically adjusted in combination with the resource allocation results in the optimal plan (such as the capacity of pre-allocated computing nodes); for medium- and low-dependent services (such as offline data analysis), the greedy algorithm is used to allocate them to resource groups with high bandwidth and low coupling according to the remaining bandwidth rate in the proximity matrix and the historical failure impact coefficient of the association matrix (such as nodes with an updated association value ≤ 0.3 are regarded as low-risk). For example, a video stream processing service has a dependency of 0.6, and its deployment plan will select a Southeast Asian resource group with a latency ≤ 50 ms and a bandwidth utilization rate < 70% in the proximity matrix, and avoid nodes with frequent recent failures in the association matrix, and finally generate a global deployment strategy that takes into account performance, disaster tolerance, and resource utilization.

[0134] In summary, the present invention discloses an integrated management method for multi-regional data centers. The method includes obtaining the physical locations, service data, and fault information of the data centers and constructing a proximity matrix and an association matrix; obtaining a first resource grouping by using a clustering algorithm; constructing a resource availability model to obtain resource requirements, and using a load prediction algorithm for prediction to obtain a resource prediction value. If it exceeds a preset resource threshold, a deployment mechanism is triggered to obtain an optimal plan; calculating a fault isolation range according to the fault information. If the isolation range includes an available data center, the association value is updated to obtain an updated association matrix; using a network topology optimization algorithm for optimization to obtain an optimization result; using a preset latency evaluation model for evaluation to obtain a latency impact degree and a second resource grouping; generating a business deployment plan according to the above features. This method can improve the management efficiency of multi-regional data centers.

[0135] The present invention constructs a multi-dimensional collaborative optimization model, which significantly reduces the network latency of cross-regional data centers and improves resource utilization while ensuring business continuity and resource elasticity. The technical derivation process is as follows: First, based on the physical location and business relevance of the data center, a proximity matrix and a correlation matrix are constructed respectively. The first resource grouping is generated through a weighted clustering algorithm, so that the nodes within the group satisfy both geographical proximity and high business correlation, thereby reducing the cross-group communication requirements. Second, the real-time load rates of the CPU, storage, and bandwidth of each node are dynamically calculated through a resource availability model. Combining with a long short-term memory network to predict future resource requirements, when the predicted value exceeds the threshold, a cross-group resource allocation mechanism based on the Hungarian algorithm is triggered, and nodes with geographical proximity and low business conflict are preferentially selected as resource suppliers to avoid local resource exhaustion caused by group barriers under traditional grouping strategies. Further, in the fault isolation stage, the system analyzes the topological influence range of the faulty node. If the isolation boundary includes low-latency available nodes, the weight value of the correlation matrix is dynamically updated, and such nodes are re-incorporated into the resource pool to achieve both fault isolation and the availability of neighboring resources. Finally, based on a network topology optimization algorithm and a latency evaluation model, a second resource grouping and a service deployment plan are generated to ensure that high-latency sensitive services are preferentially deployed in low-latency groups, while high-throughput tasks are allocated to long-distance resource groups. The present invention effectively solves the problems of conflicts between grouping strategies and geographical proximity, rigid resource allocation, and incorrect fault isolation in the prior art, and provides systematic support for the efficient management of multi-regional data centers.

[0136] Referring to Figure 2 , the second embodiment of the present invention provides an integrated management system for multi-regional data centers, including:

[0137] A data acquisition module, configured to acquire the physical location, service data, and fault information of the data center, and construct a proximity matrix and a correlation matrix according to the physical location and service data;

[0138] A grouping module, configured to perform grouping according to the proximity matrix and the correlation matrix by using a clustering algorithm to obtain a first resource grouping;

[0139] A resource optimization module, configured to construct a resource availability model according to the first resource grouping to obtain resource requirements, and use a load prediction algorithm to predict the resource requirements to obtain a resource prediction value. If the resource prediction value exceeds a preset resource threshold, a resource allocation mechanism is triggered to obtain an optimal solution;

[0140] A fault detection module, configured to calculate a fault isolation range according to the fault information, the proximity matrix, and the correlation matrix. If the isolation range includes an available data center, the correlation value is updated to obtain an updated correlation matrix;

[0141] A network optimization module, configured to perform optimization using a network topology optimization algorithm according to the first resource grouping and the resource prediction value, so as to obtain an optimization result;

[0142] A delay evaluation module, configured to perform evaluation using a preset delay evaluation model according to the optimization result, so as to obtain a delay impact degree, and obtain a second resource grouping according to the delay impact degree and the first resource grouping;

[0143] A service deployment module, configured to generate a service deployment plan according to the second resource grouping, the optimal solution, the proximity matrix, and the updated association matrix.

[0144] It should be noted that the multi-region data center integrated management system provided in the embodiment of the present invention is used to execute all the process steps of the multi-region data center integrated management method in the above embodiment. The working principles and beneficial effects of the two correspond one by one, and thus will not be elaborated herein.

[0145] The embodiment of the present invention also provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a data acquisition program. When the processor executes the computer program, the steps in the above embodiments of the multi-region data center integrated management method are implemented, such as Figure 1 the step S11 shown. Alternatively, when the processor executes the computer program, the functions of each module / unit in the above system embodiments are implemented, such as the data acquisition module.

[0146] Exemplarily, the computer program can be divided into one or more modules / units. The one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, and this instruction segment is used to describe the execution process of the computer program in the electronic device.

[0147] The electronic device can be a computing device such as a desktop computer, a notebook, a palm computer, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above components are only examples of the electronic device, and do not constitute a limitation on the electronic device. It may include more or fewer components than the above, or combine certain components, or different components. For example, the electronic device may further include input / output devices, network access devices, a bus, etc.

[0148] The so-called processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the electronic device, connecting various parts of the entire electronic device through various interfaces and circuits.

[0149] The memory can be used to store the computer programs and / or modules. The processor realizes various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory, and by calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created according to the use of the mobile phone (such as audio data, phone book, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, at least one magnetic disk storage device, flash device, or other volatile solid-state storage devices.

[0150] Among them, if the modules / units integrated in the electronic device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above-described embodiment methods of the present invention, it can also be completed by a computer program instructing relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or system, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0151] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the attached drawings of the system embodiments provided by the present invention, the connection relationship between the modules indicates that they have a communication connection, which can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0152] The above-described specific embodiments have further elaborated on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. It is particularly pointed out that for those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A multi-regional data center integrated management method, characterized in that: include: Acquire the physical location, business data and fault information of the data center, and construct a proximity matrix and an association matrix according to the physical location and business data; According to the proximity matrix and the association matrix, a clustering algorithm is used to perform grouping to obtain a first resource grouping; According to the first resource grouping, a resource availability model is constructed to obtain resource demand, and the resource demand is predicted by using a load prediction algorithm to obtain a resource prediction value. If the resource prediction value exceeds a preset resource threshold, a resource allocation mechanism is triggered to obtain an optimal solution; Calculate the fault isolation range according to the fault information, the proximity matrix and the association matrix, and if the isolation range includes an available data center, update the association value to obtain an updated association matrix; According to the first resource grouping and the resource prediction value, optimizing using a network topology optimization algorithm to obtain an optimization result; According to the optimization result, a preset delay evaluation model is used to perform an evaluation to obtain a delay impact, and according to the delay impact and the first resource grouping, a second resource grouping is obtained; A service deployment plan is generated according to the second resource grouping, the optimal plan, the proximity matrix and the updated association matrix.

2. The multi-regional data center integrated management method according to claim 1, characterized in that: The obtaining of the physical location, business data and fault information of the data center, and constructing a proximity matrix and an association matrix according to the physical location and business data, includes: Acquire the physical location, business data, and fault information of the data center, perform feature extraction on the physical location and the business data, and obtain physical location features and business data features; According to the physical location characteristics, the geographical distance between data centers is calculated to construct a proximity matrix; According to the business data characteristics, the business correlation between data centers is calculated and a correlation matrix is ​​constructed.

3. The multi-regional data center integrated management method according to claim 1, characterized in that: The step of grouping the resources using a clustering algorithm according to the proximity matrix and the association matrix to obtain a first resource grouping includes: Performing normalization processing on the proximity matrix and the correlation matrix to obtain a standardized proximity matrix and a standardized correlation matrix; Performing clustering according to the standardized proximity matrix to obtain proximity cluster centers; Performing clustering according to the standardized association matrix to obtain association cluster centers; According to the neighboring cluster center, the Euclidean distance between each data point and the neighboring cluster center is calculated to obtain a neighboring value; According to the associated cluster center, the Euclidean distance between each data point and the associated cluster center is calculated to obtain the associated value If the proximity value of a data point is higher than a preset proximity threshold, and the correlation value is higher than a preset correlation threshold, it is assigned to the group where the corresponding cluster center is located to obtain a cluster group; If the proximity value of a data point is lower than the preset proximity threshold, or the correlation value is lower than the preset correlation threshold, it is treated as an independent group; The independent group and the cluster group are merged to obtain a first resource grouping.

4. The multi-regional data center integrated management method according to claim 1, characterized in that: The method of constructing a resource availability model according to the first resource grouping to obtain resource demand and using a load prediction algorithm to predict the resource demand to obtain a resource prediction value, and triggering a resource allocation mechanism to obtain an optimal solution if the resource prediction value exceeds a preset resource threshold, includes: According to the first resource grouping, a resource availability model is constructed to obtain resource requirements; Using a load prediction algorithm to predict the resource demand to obtain a resource prediction value; Compare the resource prediction value with the preset resource threshold, and if the resource prediction value exceeds the preset resource threshold, determine that allocation is required; Calculate the difference between the resource prediction value and the preset resource threshold value to obtain the allocated resource amount; According to the resource allocation amount, a linear programming algorithm is used for optimization to obtain an optimal solution for resource allocation.

5. The multi-regional data center integrated management method according to claim 1, characterized in that: The calculating the fault isolation range according to the fault information, the proximity matrix and the association matrix, and if the isolation range includes an available data center, updating the association value to obtain an updated association matrix, includes: According to the fault information, extract the fault information to obtain the fault point; Determining an isolation area of ​​the fault point according to the fault point, the proximity matrix and the association matrix; According to the isolation zone, if the isolation zone includes an available point whose proximity is less than a preset isolation threshold, the Euclidean distance between the available point and the fault point is calculated to obtain a fault isolation range; If the isolation range includes an available data center, the association value is updated to obtain an updated association matrix.

6. The multi-regional data center integrated management method according to claim 1, characterized in that: The optimizing using a network topology optimization algorithm according to the first resource grouping and the resource prediction value to obtain an optimization result includes: Obtaining a network topology structure according to the first resource grouping and the resource prediction value; Determining available communication paths according to the network topology; According to the network topology structure and the communication path, a network topology optimization algorithm is used to optimize and obtain an optimization result.

7. The multi-regional data center integrated management method according to claim 1, characterized in that: The step of evaluating according to the optimization result using a preset delay evaluation model to obtain a delay impact, and obtaining a second resource grouping according to the delay impact and the first resource grouping includes: extracting delay data according to the optimization result; According to the delay data, a preset delay evaluation model is used to perform an evaluation to obtain a delay impact; If the delay impact is higher than the preset delay threshold, it is determined that the service is affected, and the grouping strategy is readjusted to obtain a second resource grouping; If the delay impact is lower than the preset delay threshold, it is determined that the service is not affected and the first resource group As a second resource grouping.

8. The multi-regional data center integrated management method according to claim 1, characterized in that: The generating of a service deployment solution according to the second resource grouping, the optimal solution, the proximity matrix and the updated association matrix includes: extracting group distribution data according to the second resource group; Calculating the business dependency according to the optimal solution and the distribution data; A service deployment plan is generated according to the service dependency, the proximity matrix and the updated association matrix.

9. A multi-regional data center integrated management system, characterized in that: include: A data acquisition module, used to acquire the physical location, business data and fault information of the data center, and to construct a proximity matrix and an association matrix according to the physical location and business data; A grouping module, configured to perform grouping using a clustering algorithm according to the proximity matrix and the association matrix to obtain a first resource grouping; A resource optimization module, configured to construct a resource availability model according to the first resource grouping, obtain resource demand, and predict the resource demand using a load prediction algorithm to obtain a resource prediction value. If the resource prediction value exceeds a preset resource threshold, a resource allocation mechanism is triggered to obtain an optimal solution. A fault detection module, configured to calculate a fault isolation range according to the fault information, the proximity matrix and the association matrix, and if the isolation range includes an available data center, update an association value to obtain an updated association matrix; A network optimization module, configured to optimize the network topology according to the first resource grouping and the resource prediction value, and obtain an optimization result; A delay evaluation module, configured to evaluate the optimization result using a preset delay evaluation model to obtain a delay impact, and obtain a second resource grouping according to the delay impact and the first resource grouping; A service deployment module is used to generate a service deployment plan according to the second resource grouping, the optimal plan, the proximity matrix and the updated association matrix.

10. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for integrated management of multi-regional data centers as described in any one of claims 1 to 8 is implemented.

Images