Cloud host operation management method and device, equipment, medium and program product

Abstract

The invention provides a cloud host operation management method, device and equipment, a medium and a program product, the method is applied to a cloud computing management platform, the method comprises the following steps: obtaining operation data information of all cloud hosts, the operation data information comprising control node data and computing node data; inputting the operation data information into a pre-trained cloud host operation evaluation model constructed based on a deep-width network, performing feature extraction on the control node data and the computing node data by the cloud host operation evaluation model, and merging the extracted features of the control node data and the extracted features of the computing node data, and obtaining a corresponding classification state weight, and outputting the running state of the cloud host according to the classification state weight. According to the method, the deep width learning network is utilized to establish the network model for the large-batch cloud hosts, the response speed of the large-batch cloud hosts is improved, the cloud host feature information is screened for weight training classification, and stable operation of the cloud hosts is guaranteed.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of artificial intelligence technology, and in particular, to a cloud host operation management method, device, equipment, medium and program product. Background Art

[0002] At present, machine learning algorithms and fault diagnosis methods are increasingly used in the field of cloud computing and guide the operation and maintenance of cloud computing. However, the following problems still exist:

[0003] The method of using the Transformer model to process fault data to locate the cause of the fault is only applicable to a small amount of data. When faced with a large amount of data, the training time will be greatly increased, resulting in a delay in fault location; the method of using the intelligent device management platform to monitor the device temperature status in real time and set thresholds to determine the real-time operating status of the device fails to process the data in a timely manner when it is collected, affecting the prediction function of the subsequent model; the threshold is set based on experience and lacks precision evaluation, which is very likely to lead to misjudgment of the model operation status; the method of locating the fault level according to the fault data and sending a dialing task to the computing node to locate the cause of the fault when the fault occurs lacks early fault prevention measures. When faced with a large amount of alarm data, the cause of the fault may be repeatedly located and the dialing concurrency is low. Summary of the invention

[0004] The embodiments of the present invention provide a cloud host operation management method, device, equipment, medium and program product to solve the problems of low concurrency and slow response speed in monitoring the operation of a large number of cloud hosts.

[0005] In order to solve the above-mentioned technical problems, the present invention is achieved as follows:

[0006] In a first aspect, an embodiment of the present invention provides a cloud host operation management method, which is applied to a cloud computing management platform, wherein the cloud computing management platform includes at least one computing node for controlling the cloud host and at least one control node for controlling the computing node, and the method includes:

[0007] Obtaining operation data information of all cloud hosts, the operation data information including: control node data and computing node data;

[0008] The operation data information is input into a pre-trained cloud host operation evaluation model based on a deep and wide network. The cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data and the features of the computing node data to obtain corresponding classification status weights, and outputs the cloud host operation status according to the classification status weights.

[0009] Optionally, the cloud host operation evaluation model includes a data input layer, a control node layer, a computing node layer, a feature merging layer and an operation status evaluation layer;

[0010] The cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data and the computing node data to obtain corresponding classification status weights. Outputting the cloud host operation status according to the classification status weights includes:

[0011] The data input layer obtains the operation data information of all cloud hosts, and screens the operation data information to obtain the control node input data and the computing node input data, and randomly generates a first connection weight, wherein the first connection weight is the connection weight between the data input layer and the control node layer;

[0012] The control node layer obtains the control node input data and computing node input data of the data input layer, and performs feature processing on the control node input data based on the deep and wide network to obtain control node output features, wherein the control node data input data includes at least one of the following: the number of virtual machines and the operation cycle of each computing node, and the number of control node running services;

[0013] The computing node layer obtains computing node input data that has passed through the control node layer but has not been processed, performs feature processing on the computing node input data based on the deep and wide network to obtain computing node output features, and randomly generates a second connection weight, wherein the computing node data input data includes at least one of the following: cloud host operation cycle, service quantity, hardware information, disk memory, and the second connection weight is a connection weight between the computing node layer and the feature merging layer;

[0014] The feature merging layer obtains the control node output feature and the computing node output feature, merges the control node output feature and the computing node output feature based on the deep and wide network to obtain a merged feature, and obtains a third connection weight, where the third connection weight is a connection weight between the feature merging layer and the operation status evaluation layer;

[0015] The operation status evaluation layer obtains the combined features, analyzes the combined features, obtains corresponding classification status weights, and outputs the cloud host operation status according to the classification status weights.

[0016] Optionally, also include:

[0017] Constructing the cloud host operation evaluation model;

[0018] Inputting the cloud host operation sample data into the constructed cloud host operation evaluation model, and setting the initialization iteration number and window number of the cloud host operation evaluation model;

[0019] The cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data and the computing node data to obtain corresponding classification state weights, and obtains the optimal weight based on the set number of initialization iterations and number of windows, and uses the optimal weight as the sample classification state weight to obtain the trained cloud host operation evaluation model.

[0020] Optionally, the control node layer obtains the control node input data and the computing node input data of the data input layer, and performs feature processing on the control node input data based on the deep and wide network to obtain control node output features, including:

[0021] The control node layer performs nonlinear transformation on the feature information of the control node input data through an activation function to obtain the transformed feature information, and then retains the useful feature information in the transformed feature information through the singular value decomposition method, discards the useless feature information in the transformed feature information, and obtains the sample control node output feature after dimensionality reduction.

[0022] Optionally, obtaining the optimal weight according to the set initialization iteration number and window number includes:

[0023] An iteration window is set for the merged features through a grid search method, and feature information is filtered within the iteration window according to a randomly generated first connection weight. When the number of iterations is completed, the optimal weight is obtained.

[0024] Optionally, the cloud host operation status includes: excellent status, good status, medium status and poor status;

[0025] After outputting the cloud host running status, the method further includes:

[0026] Processing the cloud host according to the operating status of the cloud host;

[0027] When the cloud host is in an excellent state or a good state, the cloud host is not processed;

[0028] When the cloud host is in a medium state, a dial test script is sent to the cloud host for inspection;

[0029] When the cloud host is in a bad state, a fault reminder is sent to the cloud host.

[0030] In a second aspect, an embodiment of the present invention provides a cloud computing management platform, including:

[0031] An acquisition module is used to acquire the operation data information of all cloud hosts, wherein the operation data information includes: control node data and computing node data;

[0032] The first processing module is used to input the operation data information into a pre-trained cloud host operation evaluation model constructed based on a deep and wide network. The cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data and the features of the computing node data to obtain corresponding classification state weights, and outputs the cloud host operation state according to the classification state weights.

[0033] Optionally, the cloud host operation evaluation model includes a data input layer, a control node layer, a computing node layer, a feature merging layer and an operation status evaluation layer;

[0034] The cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data and the computing node data to obtain corresponding classification status weights. Outputting the cloud host operation status according to the classification status weights includes:

[0035] The data input layer obtains the operation data information of all cloud hosts, and screens the operation data information to obtain the control node input data and the computing node input data, and randomly generates a first connection weight, wherein the first connection weight is the connection weight between the data input layer and the control node layer;

[0036] The control node layer obtains the control node input data and computing node input data of the data input layer, and performs feature processing on the control node input data based on the deep and wide network to obtain control node output features, wherein the control node data input data includes at least one of the following: the number of virtual machines and the operation cycle of each computing node, and the number of control node running services;

[0037] The computing node layer obtains computing node input data that has passed through the control node layer but has not been processed, performs feature processing on the computing node input data based on the deep and wide network to obtain computing node output features, and randomly generates a second connection weight, wherein the computing node data input data includes at least one of the following: cloud host operation cycle, service quantity, hardware information, disk memory, and the second connection weight is a connection weight between the computing node layer and the feature merging layer;

[0038] The feature merging layer obtains the control node output feature and the computing node output feature, merges the control node output feature and the computing node output feature based on the deep and wide network to obtain a merged feature, and obtains a third connection weight, where the third connection weight is a connection weight between the feature merging layer and the operation status evaluation layer;

[0039] The operation status evaluation layer obtains the combined features, analyzes the combined features, obtains corresponding classification status weights, and outputs the cloud host operation status according to the classification status weights.

[0040] Optionally, also include:

[0041] A construction module, used to construct the cloud host operation evaluation model;

[0042] A second processing module is used to input the cloud host operation sample data into the constructed cloud host operation evaluation model, and set the initialization iteration number and window number of the cloud host operation evaluation model;

[0043] A training module is used for the cloud host operation evaluation model to extract features from the control node data and the computing node data, and merge the extracted features of the control node data and the computing node data to obtain corresponding classification state weights, and obtain the optimal weight according to the set initialization iteration number and window number, and use the optimal weight as the sample classification state weight to obtain the trained cloud host operation evaluation model.

[0044] Optionally, the control node layer obtains the control node input data and the computing node input data of the data input layer, and performs feature processing on the control node input data based on the deep and wide network to obtain control node output features, including:

[0045] The first processing submodule is used for the control node layer to perform nonlinear transformation on the feature information of the control node input data through an activation function to obtain the transformed feature information, and then retain the useful feature information in the transformed feature information through the singular value decomposition method, discard the useless feature information in the transformed feature information, and obtain the sample control node output feature after dimensionality reduction.

[0046] Optionally, the running status evaluation layer analyzes the merged features to obtain corresponding classification status weights, and obtains the optimal weights according to the set initialization iteration number and window number, including:

[0047] The second processing submodule is used to set an iteration window for the merged features through a grid search method, filter feature information within the iteration window according to the randomly generated first connection weight, and obtain the optimal weight when the number of iterations ends.

[0048] Optionally, the cloud host operation status includes: excellent status, good status, medium status and poor status;

[0049] After outputting the cloud host running status, the method further includes:

[0050] Processing the cloud host according to the operating status of the cloud host;

[0051] When the cloud host is in an excellent state or a good state, the cloud host is not processed;

[0052] When the cloud host is in a medium state, a dial test script is sent to the cloud host for inspection;

[0053] When the cloud host is in a bad state, a fault reminder is sent to the cloud host.

[0054] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps in the cloud host operation management method described in any one of the first aspects.

[0055] In a fourth aspect, an embodiment of the present invention provides a readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps in the cloud host operation management method as described in any one of the first aspects are implemented.

[0056] In a fifth aspect, an embodiment of the present invention provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the steps in the cloud host operation management method as described in any one of the first aspects.

[0057] In the present invention, a cloud computing management platform is applied, and the cloud computing management platform includes at least one computing node for controlling a cloud host and at least one control node for controlling the computing node. The method includes: obtaining operation data information of all cloud hosts, and the operation data information includes: control node data and computing node data; inputting the operation data information into a pre-trained cloud host operation evaluation model constructed based on a deep and wide network, the cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data with the features of the computing node data to obtain corresponding classification state weights, and outputs the cloud host operation state according to the classification state weights, and establishes a network model for a large number of cloud hosts by using a deep and wide learning network to improve the response speed of a large number of cloud hosts, and at the same time screens the cloud host feature information for weight training classification to ensure the smooth operation of the cloud hosts, thereby solving the problems of low concurrency and slow response speed of operation monitoring of a large number of cloud hosts. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0059] Figure 1 It is a flow chart of a cloud host operation management method provided by an embodiment of the present invention;

[0060] Figure 2 It is a cloud computing management platform architecture diagram of a cloud host operation management method provided by an embodiment of the present invention;

[0061] Figure 3 It is a cloud host operation evaluation model architecture diagram of a cloud host operation management method provided by an embodiment of the present invention;

[0062] Figure 4 is a flow chart of another cloud host operation management method provided by an embodiment of the present invention;

[0063] Figure 5 It is a schematic diagram of a cloud host state of a cloud host operation management method provided by an embodiment of the present invention;

[0064] Figure 6 It is a structural diagram of a cloud computing management platform provided by an embodiment of the present invention;

[0065] Figure 7 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0067] Please refer to Figure 1 The embodiment of the present invention provides a cloud host operation management method, which is applied to a cloud computing management platform. The cloud computing management platform includes at least one computing node for controlling the cloud host and at least one control node for controlling the computing node. The method includes:

[0068] Step 11: Obtaining operation data information of all cloud hosts, the operation data information including: control node data and computing node data;

[0069] In the embodiment of the present invention, Figure 2 As shown, based on the deployment of cloud hosts under the Openstack architecture, the control node generally deploys three high-availability control nodes (K8S-master nodes) so that after one node goes down, another control node can be switched to run, and the control node communicates with the computing node and issues instructions; the computing node (Nova-computer node) is constructed according to the actual number of cloud hosts and is composed of N cloud hosts. The computing node carries the existing running services to ensure the normal operation of the customer's subscription services. After the user logs in to the server remotely, the cloud host running status and business running status can be observed and controlled through the control node (K8S-master node), and the actual running business can be controlled through the computing node (Nova-computer node); the cloud host architecture based on the openstack architecture is modeled by using a deep and wide learning network, and the control node data and computing node data are collected; wherein, the control node data input data includes at least one of the following: the number of virtual machines of each computing node, the operating cycle, and the number of control node running services; the computing node data input data includes at least one of the following: the cloud host operating cycle, the number of services, hardware information, and disk memory, and the second connection weight is the connection weight between the computing node layer and the feature merging layer.

[0070] Step 12: Input the operation data information into a pre-trained cloud host operation evaluation model based on a deep and wide network. The cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data and the features of the computing node data to obtain corresponding classification status weights, and outputs the cloud host operation status according to the classification status weights.

[0071] In the embodiment of the present invention, the cloud host operation evaluation model is as follows: Figure 3 As shown, the cloud host operation evaluation model includes a data input layer, a control node layer, a computing node layer, a feature merging layer and an operation status evaluation layer; by establishing a hierarchical network model, the response speed of a large number of cloud hosts is improved, and at the same time, the idea of ​​operation status evaluation is embedded, and weighted training and classification of filtered cloud host feature information are added to the existing network model, the operation status of the cloud host is judged according to different states, and corresponding processing actions are taken to ensure the smooth operation of the cloud host under the public cloud business.

[0072] In an embodiment of the present invention, it is applied to a cloud computing management platform, the cloud computing management platform includes at least one computing node for controlling a cloud host and at least one control node for controlling the computing node, the method includes: obtaining operation data information of all cloud hosts, the operation data information includes: control node data and computing node data; inputting the operation data information into a pre-trained cloud host operation evaluation model constructed based on a deep and wide network, the cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data with the features of the computing node data to obtain corresponding classification state weights, outputting the cloud host operation state according to the classification state weights, establishing a network model for a large number of cloud hosts by using a deep and wide learning network, improving the response speed of a large number of cloud hosts, and screening the cloud host feature information for weight training classification to ensure the smooth operation of the cloud hosts, thereby solving the problems of low concurrency and slow response speed of operation monitoring of a large number of cloud hosts.

[0073] In the embodiment of the present invention, optionally, the cloud host operation evaluation model includes a data input layer, a control node layer, a computing node layer, a feature merging layer and an operation status evaluation layer;

[0074] The cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data and the computing node data to obtain corresponding classification status weights. Outputting the cloud host operation status according to the classification status weights includes:

[0075] The data input layer obtains the operation data information of all cloud hosts, and screens the operation data information to obtain the control node input data and the computing node input data, and randomly generates a first connection weight, wherein the first connection weight is the connection weight between the data input layer and the control node layer;

[0076] The control node layer obtains the control node input data and computing node input data of the data input layer, and performs feature processing on the control node input data based on the deep and wide network to obtain control node output features, wherein the control node data input data includes at least one of the following: the number of virtual machines and the operation cycle of each computing node, and the number of control node running services;

[0077] The computing node layer obtains computing node input data that has passed through the control node layer but has not been processed, performs feature processing on the computing node input data based on the deep and wide network to obtain computing node output features, and randomly generates a second connection weight, wherein the computing node data input data includes at least one of the following: cloud host operation cycle, service quantity, hardware information, disk memory, and the second connection weight is a connection weight between the computing node layer and the feature merging layer;

[0078] The feature merging layer obtains the control node output feature and the computing node output feature, merges the control node output feature and the computing node output feature based on the deep and wide network to obtain a merged feature, and obtains a third connection weight, where the third connection weight is a connection weight between the feature merging layer and the operation status evaluation layer;

[0079] The operation status evaluation layer obtains the combined features, analyzes the combined features, obtains corresponding classification status weights, and outputs the cloud host operation status according to the classification status weights.

[0080] In the embodiment of the present invention, please refer to Figure 3 , a deep and wide learning network is used to model the cloud host architecture based on the openstack architecture. The data input layer obtains the operating data information of all cloud hosts and filters the operating data information. The data is filtered through features to form the control node layer.

[0081] In the embodiment of the present invention, optionally, the control node layer obtains the control node input data and the computing node input data of the data input layer, and performs feature processing on the control node input data based on the deep and wide network to obtain control node output features, including:

[0082] The control node layer performs nonlinear transformation on the feature information of the control node input data through an activation function to obtain transformed feature information, and then retains useful feature information in the transformed feature information through a singular value decomposition method, discards useless feature information in the transformed feature information, and obtains the sample control node output feature after dimensionality reduction. The formula is as follows:

[0083] T i =φ(X i W1i )

[0084] T i =UZ i V T (1)

[0085] Among them, X i represents the input data of the control node (K8S-master node), φ( ) represents the feature activation function, T i represents the unfiltered feature output in the control node layer, Z i In order to obtain the output features after dimensionality reduction by using the singular value decomposition (SVD), the specific principle of the singular value decomposition (SVD) is to convert T i It is represented as the product of three matrix parameters, Z i The original feature output is reduced in dimension by the SVD method, which can retain the useful information of the original feature to the greatest extent, and the useless redundant information is allocated to U, V T Two parameters; W 1i is the first connection weight, ranging from (-1,1). Since the deep and wide network has a mechanism for randomly generating weights, that is, except for the weight of the output layer, the weights of the other input layers are randomly generated, that is, the first connection weight is randomly generated, wherein the first connection weight is the connection weight between the data input layer and the control node layer. This mechanism can ensure the generalization of the network model; then establish a computing node layer, the computing node layer is composed of N computing nodes (Nova-computer nodes), N nodes represent N cloud hosts, the computing node layer obtains the computing node input data that has passed the control node layer but has not been processed, and performs feature processing on the input data of the computing node based on the deep and wide network to obtain the computing node output feature, and randomly generates the second connection weight, wherein the control node layer first performs feature output dimensionality reduction on the control node cloud host, and then synchronously performs dimensionality reduction processing on the reduced computing node layer data to realize network flattening management, and its modeling formula is as follows:

[0086] H i =ζ(Z i |Y i W 2i ) (2)

[0087] Among them, Y i Represents the input data of the computing node layer; W 2i is the second connection weight, whose value is randomly generated and ranges from (-1,1); H iThe output of the control node and the computing node is merged; the feature merging layer obtains the output features of the control node and the computing node, merges the output features of the control node and the computing node based on the deep and wide network, obtains the merged features, and obtains the third connection weight, merges the two layers of feature outputs, and reduces the training time and iteration time for the next step of cloud host operation status evaluation; finally, the operation status evaluation layer is set, and the merged features are used as input, and the slow feature analysis method is used to distinguish them into corresponding classification status weights, that is, four evaluation states of excellent, medium and poor, so as to achieve the purpose of real-time monitoring of the cloud host operation status. The formula is as follows:

[0088] L(i)=<(H i W 3i ) 2 >

[0089] i=1,2...4 (3)

[0090] Among them, W 3i is the third connection weight, which is also the parameter we need to solve; < > represents the first-order derivative of the objective function, and obtains its slow feature number; L(i) is the classification state after the model is established, that is, the four states of excellent, good, medium and poor, among which L(1) is the excellent state; L(2) is the good state; L(3) is the medium state; L(4) is the poor state;

[0091] In the embodiment of the present invention, optionally, the cloud host operation status includes: excellent status, good status, medium status and poor status;

[0092] After outputting the cloud host running status, the method further includes:

[0093] Processing the cloud host according to the operating status of the cloud host;

[0094] When the cloud host is in an excellent state or a good state, the cloud host is not processed;

[0095] When the cloud host is in a medium state, a dial test script is sent to the cloud host for inspection;

[0096] When the cloud host is in a bad state, a fault reminder is sent to the cloud host.

[0097] In the embodiment of the present invention, by providing different analysis strategies and maintenance methods according to each state, intelligent operation and maintenance is realized, and the problems of low concurrency and slow response speed of large-scale cloud host operation monitoring are solved, providing a new solution for ensuring the smooth operation of cloud hosts under public cloud services.

[0098] In the embodiment of the present invention, optionally, the method further includes:

[0099] Constructing the cloud host operation evaluation model;

[0100] Inputting the cloud host operation sample data into the constructed cloud host operation evaluation model, and setting the initialization iteration number and window number of the cloud host operation evaluation model;

[0101] The cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data and the computing node data to obtain corresponding classification state weights, and obtains the optimal weight based on the set number of initialization iterations and number of windows, and uses the optimal weight as the sample classification state weight to obtain the trained cloud host operation evaluation model.

[0102] In the embodiment of the present invention, please refer to Figure 4 , the training process of running the evaluation model on the cloud host includes:

[0103] Enter all cloud host operation data information, set the number of iterations and windows for initializing the network, and set the status threshold to 0.8;

[0104] Control node input data X i After feature mapping, it is sent to the control node layer, and the weight from the input layer to the control node layer, that is, the first connection weight, is randomly adjusted.

[0105] According to formula (1), the control node layer obtains feature information and then performs nonlinear transformation through the activation function, and then retains the useful feature Z through the SVD method. i , discard U,V T Two useless redundant feature parameters are used to achieve feature dimensionality reduction, thereby achieving the purpose of screening useful information of cloud host operation characteristics and reducing network training redundancy and training time.

[0106] The remaining computing node data of the control node layer that has not been processed by dimensionality reduction are combined through activation function and random weight distribution to obtain computing node layer feature information. The feature information of the computing node layer contains useful operation data information of the computing node cloud host.

[0107] The control node output feature and the computing node output feature are merged into one input layer, namely, a feature merging layer, and the merged feature contains all the information of the control node and the computing node.

[0108] The combined features are used as input data for training using a slow feature analysis method to obtain classification state weights, i.e., weights of four categories: excellent, medium, and poor. The optimal weights are obtained by setting the number of iterations according to a grid search training method. The grid search method is to pre-set an iteration window for the combined features, which can be set to (1-100). The weights randomly generated by formula (1) are used within the iteration window to continuously filter feature information. When the iteration is completed, a set of optimal output features and weights are generated and input into the operation state evaluation layer.

[0109] After the operation status evaluation layer obtains the overall output features, it divides the status into four categories: good, medium, and poor according to formula (3), that is, the real-time operation status of the control node and the computing node at this time is obtained, and the trained cloud host operation evaluation model is obtained.

[0110] Processing the cloud host according to the operating status of the cloud host;

[0111] Specifically, when the cloud host is in an excellent state or a good state, the cloud host is not processed; when the cloud host is in a medium state, a dialing script is sent to the cloud host for inspection; when the cloud host is in a poor state, a fault reminder is sent to the cloud host.

[0112] Specifically, taking 3 K8S-master control nodes and 10 nova-computer computing nodes as an example, a total of 1500 cloud computer operation continuous data are screened, among which the cloud host operation evaluation model is divided into four states after network training: good, medium, and poor. The classification corresponds to four groups of data, such as Figure 5 As shown, (0-374) is the cloud host operation data in a bad state. At this time, the operation status is higher than the threshold. It can be seen that the cloud host operation data in this state has failed, and the cause of the failure needs to be sent to the maintenance personnel to take emergency measures according to the guarantee reason; (374-711) is a medium state, indicating that the cloud host operation status may have an alarm at this time, and a dial-up test script needs to be sent for inspection; (712-1225) indicates that the cloud host is in a good state. The cloud host may be changing operations or the cloud host health check may have critical values ​​and other problems, but this operation does not affect the existing business, and maintenance personnel need to continue to observe; (1226-1500) indicates that the cloud host is in an excellent state. At this time, the cloud host is running normally and no adjustment is required. All stock and existing businesses are running normally.

[0113] In the embodiment of the present invention, a deep and wide learning network is used to establish a network model for a large number of cloud hosts. Since the cloud hosts are established based on the openstack framework, a hierarchical network model is established according to the K8S-master and nova-computer nodes to improve the response speed of a large number of cloud hosts. At the same time, the idea of ​​running status evaluation is embedded, and a slow feature analysis method is added to the existing network model to perform weight training and classification on the feature information of the screened cloud hosts to obtain four states of good, medium and poor. The running status of the cloud hosts is judged according to different states, and corresponding processing actions are taken. Combined with artificial intelligence and data-driven algorithms, the problems of low concurrency and slow response speed of running monitoring of a large number of cloud hosts are solved, and a new solution is provided to ensure the smooth operation of cloud hosts under public cloud services.

[0114] Please refer to Figure 6 , an embodiment of the present invention provides a cloud computing management platform, including:

[0115] The acquisition module 61 is used to acquire the operation data information of all cloud hosts, and the operation data information includes: control node data and computing node data;

[0116] The first processing module 62 is used to input the operation data information into a pre-trained cloud host operation evaluation model constructed based on a deep and wide network. The cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data and the features of the computing node data to obtain corresponding classification status weights, and outputs the cloud host operation status according to the classification status weights.

[0117] In the embodiment of the present invention, optionally, the cloud host operation evaluation model includes a data input layer, a control node layer, a computing node layer, a feature merging layer and an operation status evaluation layer;

[0118] The cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data and the computing node data to obtain corresponding classification status weights. Outputting the cloud host operation status according to the classification status weights includes:

[0119] The data input layer obtains the operation data information of all cloud hosts, and screens the operation data information to obtain the control node input data and the computing node input data, and randomly generates a first connection weight, wherein the first connection weight is the connection weight between the data input layer and the control node layer;

[0120] The control node layer obtains the control node input data and computing node input data of the data input layer, and performs feature processing on the control node input data based on the deep and wide network to obtain control node output features, wherein the control node data input data includes at least one of the following: the number of virtual machines and the operation cycle of each computing node, and the number of control node running services;

[0121] The computing node layer obtains computing node input data that has passed through the control node layer but has not been processed, performs feature processing on the computing node input data based on the deep and wide network to obtain computing node output features, and randomly generates a second connection weight, wherein the computing node data input data includes at least one of the following: cloud host operation cycle, service quantity, hardware information, disk memory, and the second connection weight is a connection weight between the computing node layer and the feature merging layer;

[0122] The feature merging layer obtains the control node output feature and the computing node output feature, merges the control node output feature and the computing node output feature based on the deep and wide network to obtain a merged feature, and obtains a third connection weight, where the third connection weight is a connection weight between the feature merging layer and the operation status evaluation layer;

[0123] The operation status evaluation layer obtains the combined features, analyzes the combined features, obtains corresponding classification status weights, and outputs the cloud host operation status according to the classification status weights.

[0124] In the embodiment of the present invention, optionally, the method further includes:

[0125] A construction module, used to construct the cloud host operation evaluation model;

[0126] A second processing module is used to input the cloud host operation sample data into the constructed cloud host operation evaluation model, and set the initialization iteration number and window number of the cloud host operation evaluation model;

[0127] A training module is used for the cloud host operation evaluation model to extract features from the control node data and the computing node data, and merge the extracted features of the control node data and the computing node data to obtain corresponding classification state weights, and obtain the optimal weight according to the set initialization iteration number and window number, and use the optimal weight as the sample classification state weight to obtain the trained cloud host operation evaluation model.

[0128] In the embodiment of the present invention, optionally, the control node layer obtains the control node input data and the computing node input data of the data input layer, and performs feature processing on the control node input data based on the deep and wide network to obtain control node output features, including:

[0129] The first processing submodule is used for the control node layer to perform nonlinear transformation on the feature information of the control node input data through an activation function to obtain the transformed feature information, and then retain the useful feature information in the transformed feature information through the singular value decomposition method, discard the useless feature information in the transformed feature information, and obtain the sample control node output feature after dimensionality reduction.

[0130] In the embodiment of the present invention, optionally, the running state evaluation layer analyzes the merged features to obtain corresponding classification state weights, and obtains the optimal weights according to the set initialization iteration number and window number, including:

[0131] The second processing submodule is used to set an iteration window for the merged features through a grid search method, filter feature information within the iteration window according to the randomly generated first connection weight, and obtain the optimal weight when the number of iterations ends.

[0132] In the embodiment of the present invention, optionally, the cloud host operation status includes: excellent status, good status, medium status and poor status;

[0133] After outputting the cloud host running status, the method further includes:

[0134] Processing the cloud host according to the operating status of the cloud host;

[0135] When the cloud host is in an excellent state or a good state, the cloud host is not processed;

[0136] When the cloud host is in a medium state, a dial test script is sent to the cloud host for inspection;

[0137] When the cloud host is in a bad state, a fault reminder is sent to the cloud host.

[0138] The cloud computing management platform device provided by the embodiment of the present invention can realize Figure 1 The various processes implemented by the method embodiment and achieving the same technical effect are not described here to avoid repetition.

[0139] An embodiment of the present invention provides an electronic device 70, see Figure 7 As shown, Figure 7This is a principle block diagram of an electronic device 70 according to an embodiment of the present invention, including a processor 71, a memory 72, and a program or instruction stored in the memory 72 and executable on the processor 71. When the program or instruction is executed by the processor, the steps in any one of the cloud host operation management methods of the present invention are implemented.

[0140] An embodiment of the present invention provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the various processes of an embodiment of a cloud host operation management method such as any of the above-mentioned items are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0141] The present application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the above Figure 1 The various processes of the method embodiment shown can achieve the same technical effect, and will not be described again here to avoid repetition.

[0142] Computer readable media include permanent and non-permanent, removable and non-removable media, and can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0143] It should be noted that the collection, collection, updating, analysis, processing, use, transmission, storage and other aspects of user personal information involved in the technical solution of this disclosure are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information and network security.

[0144] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0145] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0146] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a service classification device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0147] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A cloud host operation management method, characterized in that: Applied to a cloud computing management platform, the cloud computing management platform includes at least one computing node for controlling a cloud host and at least one control node for controlling the computing node, the method includes: Obtaining operation data information of all cloud hosts, the operation data information including: control node data and computing node data; The operation data information is input into a pre-trained cloud host operation evaluation model based on a deep and wide network. The cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data and the features of the computing node data to obtain corresponding classification status weights, and outputs the cloud host operation status according to the classification status weights.

2. The cloud host operation management method according to claim 1, characterized in that: The cloud host operation evaluation model includes a data input layer, a control node layer, a computing node layer, a feature merging layer and an operation status evaluation layer; The cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data and the computing node data to obtain corresponding classification status weights. Outputting the cloud host operation status according to the classification status weights includes: The data input layer obtains the operation data information of all cloud hosts, and screens the operation data information to obtain the control node input data and the computing node input data, and randomly generates a first connection weight, wherein the first connection weight is the connection weight between the data input layer and the control node layer; The control node layer obtains the control node input data and computing node input data of the data input layer, and performs feature processing on the control node input data based on the deep and wide network to obtain control node output features, wherein the control node data input data includes at least one of the following: the number of virtual machines and the operation cycle of each computing node, and the number of control node running services; The computing node layer obtains computing node input data that has passed through the control node layer but has not been processed, performs feature processing on the computing node input data based on the deep and wide network to obtain computing node output features, and randomly generates a second connection weight, wherein the computing node data input data includes at least one of the following: cloud host operation cycle, service quantity, hardware information, disk memory, and the second connection weight is a connection weight between the computing node layer and the feature merging layer; The feature merging layer obtains the control node output feature and the computing node output feature, merges the control node output feature and the computing node output feature based on the deep and wide network to obtain a merged feature, and obtains a third connection weight, where the third connection weight is a connection weight between the feature merging layer and the operation status evaluation layer; The operation status evaluation layer obtains the combined features, analyzes the combined features, obtains corresponding classification status weights, and outputs the cloud host operation status according to the classification status weights.

3. The cloud host operation management method according to claim 2, characterized in that: Also includes: Constructing the cloud host operation evaluation model; Inputting the cloud host operation sample data into the constructed cloud host operation evaluation model, and setting the initialization iteration number and window number of the cloud host operation evaluation model; The cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data and the computing node data to obtain corresponding classification state weights, and obtains the optimal weight based on the set number of initialization iterations and number of windows, and uses the optimal weight as the sample classification state weight to obtain the trained cloud host operation evaluation model.

4. The cloud host operation management method according to claim 2, characterized in that: The control node layer obtains the control node input data and the computing node input data of the data input layer, and performs feature processing on the control node input data based on the deep and wide network to obtain control node output features, including: The control node layer performs nonlinear transformation on the feature information of the control node input data through an activation function to obtain the transformed feature information, and then retains the useful feature information in the transformed feature information through the singular value decomposition method, discards the useless feature information in the transformed feature information, and obtains the sample control node output feature after dimensionality reduction.

5. The cloud host operation management method according to claim 3, characterized in that: The step of obtaining the optimal weight according to the set initialization iteration number and window number includes: An iteration window is set for the merged features through a grid search method, and feature information is filtered within the iteration window according to a randomly generated first connection weight. When the number of iterations is completed, the optimal weight is obtained.

6. The cloud host operation management method according to claim 1, characterized in that: The cloud host operation status includes: excellent status, good status, medium status and poor status; After outputting the cloud host running status, the method further includes: Processing the cloud host according to the operating status of the cloud host; When the cloud host is in an excellent state or a good state, the cloud host is not processed; When the cloud host is in a medium state, a dial test script is sent to the cloud host for inspection; When the cloud host is in a bad state, a fault reminder is sent to the cloud host.

7. A cloud computing management platform, characterized in that: include: An acquisition module is used to acquire the operation data information of all cloud hosts, wherein the operation data information includes: control node data and computing node data; The first processing module is used to input the operation data information into a pre-trained cloud host operation evaluation model constructed based on a deep and wide network. The cloud host operation evaluation model extracts features from the control node data and the computing node data, and merges the extracted features of the control node data and the features of the computing node data to obtain corresponding classification state weights, and outputs the cloud host operation state according to the classification state weights.

8. An electronic device, characterized in that: It includes a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps in the cloud host operation management method as described in any one of claims 1 to 6.

9. A readable storage medium, characterized in that: The readable storage medium stores programs or instructions, and when the programs or instructions are executed by the processor, the steps in the cloud host operation management method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that It includes computer instructions, which, when executed by a processor, implement the steps in the cloud host operation management method as described in any one of claims 1 to 6.

Images