Data monitoring method and device of cloud host, electronic equipment and readable medium

Abstract

The embodiment of the invention provides a data monitoring method and device for a cloud host, electronic equipment and a readable medium, and is applied to the cloud host, at least one cloud host forms an available area, at least one available area forms a resource pool, and any cloud host is configured with a preset cloud host monitoring model. Monitoring the running state data of the cloud host by using the cloud host monitoring model; configuring a preset available area monitoring model for any available area, and monitoring running state data of at least one cloud host in the available area by using the available area monitoring model; and configuring a preset resource pool monitoring model for the resource pool, and monitoring operation state data of the cloud host in the at least one available area in the resource pool by using the resource pool monitoring model. A multi-layer federated framework is constructed by utilizing a federated learning mode in a distributed scene, data feature information of different levels is fully utilized, and data privacy in an edge end cloud host is effectively protected.

Description

Technical Field

[0001] The present invention relates to the technical field of cloud hosts, and in particular to a data monitoring method for a cloud host, a data monitoring device for a cloud host, an electronic device, and a computer-readable medium. Background Art

[0002] Related technologies provide a method for monitoring cloud host resources with active monitoring. Specifically, machine learning and deep learning technologies can be applied to continuously collect data information from each cloud host, and use this information to continuously train the machine learning model to build a monitoring prediction model. The model can achieve the goal of predictive monitoring and alarm by analyzing the collected data information. However, this active monitoring mode requires continuous acquisition of resource operation information from the tenant's cloud host, which cannot meet the needs of data privacy and security. Summary of the invention

[0003] The embodiments of the present invention provide a data monitoring method, device, electronic device and computer-readable storage medium for a cloud host to solve the problem that a monitoring prediction model can achieve the goal of predictive monitoring and alarm by analyzing collected data information, but this active monitoring mode requires continuous acquisition of resource operation information from the tenant's cloud host and cannot meet the needs of data privacy and security.

[0004] An embodiment of the present invention discloses a data monitoring method for a cloud host, which is applied to a cloud host, wherein at least one of the cloud hosts forms an availability zone, and at least one of the availability zones forms a resource pool. The method includes:

[0005] Configuring a preset cloud host monitoring model for any of the cloud hosts, and using the cloud host monitoring model to monitor the operating status data of the cloud host;

[0006] Configuring a preset availability zone monitoring model for any of the availability zones, and using the availability zone monitoring model to monitor the operating status data of at least one of the cloud hosts in the availability zone;

[0007] A preset resource pool monitoring model is configured for the resource pool, and the resource pool monitoring model is used to monitor the operating status data of at least one cloud host in the available zone in the resource pool.

[0008] Optionally, the method comprises:

[0009] Configuring a preset cloud host machine learning model for any of the cloud hosts;

[0010] Configuring a preset availability zone machine learning model for the availability zone;

[0011] Configuring a preset resource pool machine learning model for the resource pool;

[0012] Obtaining training operation status data of the cloud host in the resource pool and a label corresponding to the training operation status data; the label is used to indicate whether the cloud host corresponding to the training operation status data has an abnormality;

[0013] Using the training running status data of the cloud host and the label corresponding to the training running status data as the training data of the cloud host;

[0014] The cloud host machine learning model, the availability zone machine learning model and the resource pool machine learning model are trained using the training data of the cloud host to obtain the cloud host monitoring model, the availability zone monitoring model and the resource pool monitoring model.

[0015] Optionally, the cloud host machine learning model has initialization parameters; the using the training data of the cloud host to train the cloud host machine learning model, the available zone machine learning model and the resource pool machine learning model to obtain the cloud host monitoring model, the available zone monitoring model and the resource pool monitoring model includes:

[0016] Training the cloud host machine learning model by using the training data of the cloud host, updating the initialization parameters, and obtaining a first model gradient parameter;

[0017] Controlling the cloud host machine learning model to send the first model gradient parameter to the available zone machine learning model;

[0018] The available zone machine learning model is controlled to update the first model gradient parameters based on the training data of the cloud host in the available zone to obtain the second model gradient parameters.

[0019] Optionally, using the training data of the cloud host to train the cloud host machine learning model, the available zone machine learning model, and the resource pool machine learning model to obtain the cloud host monitoring model, the available zone monitoring model, and the resource pool monitoring model includes:

[0020] Controlling the available zone machine learning model to send the second model gradient parameter to the resource pool machine learning model;

[0021] Controlling the resource pool machine learning model to update the second model gradient parameters based on the training data of the cloud host in the resource pool to obtain target model parameters;

[0022] Controlling the resource pool machine learning model to send the target model parameters to the available zone machine learning model;

[0023] Control the availability zone machine learning model to send the target model parameters to the cloud host machine learning model.

[0024] Optionally, controlling the available zone machine learning model to update the first model gradient parameter based on the training data of the cloud host in the available zone to obtain the second model gradient parameter includes:

[0025] Controlling the available zone machine learning model to update the first model gradient parameter based on the training data of the cloud host in the available zone to obtain the model gradient parameter to be processed;

[0026] Controlling the available zone machine learning model to send the to-be-processed model gradient parameter to the cloud host machine learning model;

[0027] Controlling the cloud host machine learning model to use the training data of the cloud host to update the gradient parameters of the model to be processed to obtain a second model gradient parameter;

[0028] Control the cloud host machine learning model to send the second model gradient parameters to the available zone machine learning model.

[0029] Optionally, controlling the cloud host machine learning model to use the training data of the cloud host to update the gradient parameters of the model to be processed to obtain the second model gradient parameters includes:

[0030] Adding at least one preset proportion of Gaussian noise to the training data of the cloud host to obtain target training data of the cloud host;

[0031] The target training data is used to update the gradient parameters of the model to be processed to obtain the second model gradient parameters.

[0032] Optionally, the method comprises:

[0033] Based on a preset Gaussian noise generation function, obtaining Gaussian noise of the preset proportion;

[0034] The Gaussian noise generating function is:

[0035]

[0036] Wherein, f(x) is the Gaussian operation, σ is the standard deviation of the training data, μ is the mean of the training data, and x is any of the training data.

[0037] The embodiment of the present invention further discloses a data monitoring device for a cloud host, which is applied to a cloud host, wherein at least one of the cloud hosts forms an available zone, and at least one of the available zones forms a resource pool. The device comprises:

[0038] A first configuration module, configured to configure a preset cloud host monitoring model for any of the cloud hosts, and use the cloud host monitoring model to monitor the operating status data of the cloud host;

[0039] A second configuration module is used to configure a preset availability zone monitoring model for any of the availability zones, and use the availability zone monitoring model to monitor the operating status data of at least one of the cloud hosts in the availability zone;

[0040] The third configuration module is used to configure a preset resource pool monitoring model for the resource pool, and use the resource pool monitoring model to monitor the operating status data of at least one cloud host in the available zone in the resource pool.

[0041] The embodiment of the present invention further discloses an electronic device, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0042] The memory is used to store computer programs;

[0043] The processor is used to implement the method described in the embodiment of the present invention when executing the program stored in the memory.

[0044] The embodiment of the present invention further discloses one or more computer-readable media on which instructions are stored. When executed by one or more processors, the processors are enabled to execute the method described in the embodiment of the present invention.

[0045] The embodiments of the present invention include the following advantages:

[0046] In an embodiment of the present invention, at least one cloud host forms an availability zone, at least one availability zone forms a resource pool, a preset cloud host monitoring model is configured for any cloud host, and the cloud host monitoring model is used to monitor the operating status data of the cloud host; a preset availability zone monitoring model is configured for any availability zone, and the availability zone monitoring model is used to monitor the operating status data of at least one cloud host in the availability zone; a preset resource pool monitoring model is configured for the resource pool, and the resource pool monitoring model is used to monitor the operating status data of the cloud host in at least one availability zone in the resource pool. The present invention adopts a federated learning method to construct a multi-layer federated framework in a distributed scenario using a federated learning method, making full use of data feature information at different levels and effectively protecting data privacy in edge cloud hosts. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a flow chart of the steps of a data monitoring method of a cloud host provided in an embodiment of the present invention;

[0048] Figure 2is a schematic diagram of a resource pool provided in an embodiment of the present invention;

[0049] Figure 3 It is a flowchart of the steps of another cloud host data monitoring method provided in an embodiment of the present invention;

[0050] Figure 4 A flowchart of steps for monitoring a cloud host using a cloud host monitoring model provided in an embodiment of the present invention;

[0051] Figure 5 It is a structural block diagram of a data monitoring device for a cloud host provided in an embodiment of the present invention;

[0052] Figure 6 is a block diagram of an electronic device provided in an embodiment of the present invention;

[0053] Figure 7 is a schematic diagram of a computer-readable medium provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0055] To facilitate understanding of the technical solutions and technical effects of the embodiments of the present invention, the prior art of the present invention is briefly described below.

[0056] With the accelerated development of the informatization process, the new generation of information technology has accelerated breakthroughs, presenting a new development trend of organic integration and systematic innovation. Information technology has been deeply integrated into all fields and processes of economic, social and people's livelihood, promoting the in-depth evolution of information civilization and providing important support for the construction of new information productivity. As a new production factor, data has become an important driving force for economic development. Data is a new information carrier after language, symbols and text. It has been systematically integrated into various fields and links such as social production, resource allocation, material circulation, and residents' consumption, providing new impetus for economic development. The realization of data value depends on the scheduling and use of computing power, and cloud computing can serve as a carrier of computing power.

[0057] In the field of cloud computing, in order to ensure the stability and reliability of cloud services, it is necessary to monitor cloud host resources, especially to implement monitoring measures on cloud hosts used by tenants to track and manage the resource usage of these cloud hosts. Among them, tenants refer to customers or users who use resources provided by cloud service providers; each tenant usually has an independent virtualization environment, which is isolated from the resources of other tenants to ensure data security and privacy.

[0058] In the related art, when monitoring cloud host resources, monitoring data is usually collected at a predetermined time interval to obtain device status information. However, when a monitoring alarm occurs, this fixed time interval monitoring method may lack the necessary flexibility to respond, resulting in the monitoring problem not being solved in a timely manner, thereby affecting the reliability of the service.

[0059] In order to solve this problem, the related art provides an active monitoring method for cloud host resource monitoring. Specifically, machine learning and deep learning technologies can be applied to continuously collect data information of each cloud host, and use this information to continuously train the machine learning model to build a monitoring prediction model. The model can achieve the goal of predicting monitoring alarms by analyzing the collected data information. Compared with the passive monitoring method that collects monitoring data at predetermined time intervals, this active monitoring method can improve the efficiency of handling cloud host problems, thereby improving the reliability of services.

[0060] However, this active monitoring mode requires continuous acquisition of resource operation information from the tenant's cloud host, which cannot meet the requirements of data privacy and security.

[0061] Reference Figure 1 , shows a flow chart of the steps of a data monitoring method for a cloud host provided in an embodiment of the present invention, which is applied to a cloud host, at least one of the cloud hosts forms an availability zone, and at least one of the availability zones forms a resource pool, and specifically may include the following steps:

[0062] Step 101, configuring a preset cloud host monitoring model for any of the cloud hosts, and using the cloud host monitoring model to monitor the operating status data of the cloud host;

[0063] In the embodiment of the present invention, the entire resource pool is divided into a three-layer structure. Figure 2 , showing a schematic diagram of a resource pool provided in an embodiment of the present invention. Among them, at least one cloud host forms an availability zone, and at least one availability zone forms a resource pool. The first layer of the three-layer resource pool is to use cloud hosts in different resource pools to generate an edge monitoring prediction model using local data training; the second layer is to use all cloud hosts in each availability zone as edge nodes, and use the method proposed in the present invention to perform model training to obtain an availability zone-level monitoring prediction model. The third layer is to migrate the availability zone-level monitoring prediction model trained in each availability zone to the entire resource pool for the same model training to obtain a resource pool-level monitoring prediction model.

[0064] In an embodiment of the present invention, a preset cloud host monitoring model, i.e., a monitoring prediction model at the edge, is configured for any cloud host. The cloud host monitoring model is used to monitor the operating status data of the cloud host. It should be noted that the cloud host monitoring model is embedded in the monitoring plug-in in the cloud host, and analyzes the resource information of the cloud host in real time for monitoring and early warning.

[0065] Step 102: configuring a preset availability zone monitoring model for any of the availability zones, and using the availability zone monitoring model to monitor the operating status data of at least one of the cloud hosts in the availability zone;

[0066] In the embodiment of the present invention, a preset available zone monitoring model, that is, a monitoring prediction model of the available zone, is configured for any available zone, and the available zone monitoring model is used to monitor the operating status data of at least one cloud host in the available zone.

[0067] Step 103: configure a preset resource pool monitoring model for the resource pool, and use the resource pool monitoring model to monitor the operating status data of at least one cloud host in the available zone in the resource pool.

[0068] In an embodiment of the present invention, a preset resource pool monitoring model, namely a resource pool-level monitoring prediction model, is configured for the resource pool. The resource pool monitoring model is used to monitor the operation status data of the cloud host in at least one available zone in the resource pool.

[0069] In some embodiments of the invention, the method comprises:

[0070] Configuring a preset cloud host machine learning model for any of the cloud hosts;

[0071] Configuring a preset availability zone machine learning model for the availability zone;

[0072] Configuring a preset resource pool machine learning model for the resource pool;

[0073] Obtaining training operation status data of the cloud host in the resource pool and a label corresponding to the training operation status data; the label is used to indicate whether the cloud host corresponding to the training operation status data has an abnormality;

[0074] Using the training running status data of the cloud host and the label corresponding to the training running status data as the training data of the cloud host;

[0075] The cloud host machine learning model, the availability zone machine learning model and the resource pool machine learning model are trained using the training data of the cloud host to obtain the cloud host monitoring model, the availability zone monitoring model and the resource pool monitoring model.

[0076] In an embodiment of the present invention, a preset cloud host machine learning model is configured for any cloud host; a preset available zone machine learning model is configured for an available zone; a preset resource pool machine learning model is configured for a resource pool; the training running status data of the cloud host in the resource pool and the label corresponding to the training running status data are obtained; the label is used to indicate whether the cloud host corresponding to the training running status data has an abnormality. The training running status data of the cloud host and the label corresponding to the training running status data are used as the training data of the cloud host; the cloud host machine learning model, the available zone machine learning model and the resource pool machine learning model are trained using the training data of the cloud host to obtain the cloud host monitoring model, the available zone monitoring model and the resource pool monitoring model.

[0077] Among them, the cloud host machine learning model, the available zone machine learning model, and the resource pool machine learning model all have initialization parameters. The cloud host machine learning model, the available zone machine learning model, and the resource pool machine learning model are all deep learning models based on convolutional neural networks. Specifically, they include an input layer, a pooling layer, a convolution layer, a fully connected layer, and an activation function. After the data enters the input layer, the output is connected to a convolution layer as input, and a ReLu activation function is connected after the convolution layer. The output data is connected to a pooling layer and a fully connected layer.

[0078] In each cloud host, the data resource information generated locally is standardized. The data item indicators that need to be collected and monitored are first formatted and uniformly maintained in a two-dimensional structure. Each row is divided by timestamp, and each column is an indicator item that needs to be monitored. The data with the same timestamp is merged, and the null values ​​generated during the merging process are resampled backwards to supplement.

[0079] In some embodiments of the present invention, the cloud host machine learning model has initialization parameters; the using the training data of the cloud host to train the cloud host machine learning model, the available zone machine learning model and the resource pool machine learning model to obtain the cloud host monitoring model, the available zone monitoring model and the resource pool monitoring model includes:

[0080] Training the cloud host machine learning model by using the training data of the cloud host, updating the initialization parameters, and obtaining a first model gradient parameter;

[0081] Controlling the cloud host machine learning model to send the first model gradient parameter to the available zone machine learning model;

[0082] The available zone machine learning model is controlled to update the first model gradient parameters based on the training data of the cloud host in the available zone to obtain the second model gradient parameters.

[0083] In an embodiment of the present invention, the cloud host in each availability zone uses the training data generated locally to train the cloud host machine learning model, uploads the first model gradient parameters obtained by training to the monitoring prediction model at the availability zone level, i.e., the availability zone machine learning model, and aggregates and continuously iterates to obtain the second model gradient parameters. Aggregation refers to obtaining the average value of the first model gradient parameters of at least one cloud host machine learning model.

[0084] In some embodiments of the present invention, the cloud host machine learning model, the available zone machine learning model, and the resource pool machine learning model are trained using the training data of the cloud host to obtain the cloud host monitoring model, the available zone monitoring model, and the resource pool monitoring model, including:

[0085] Controlling the available zone machine learning model to send the second model gradient parameter to the resource pool machine learning model;

[0086] Controlling the resource pool machine learning model to update the second model gradient parameters based on the training data of the cloud host in the resource pool to obtain target model parameters;

[0087] Controlling the resource pool machine learning model to send the target model parameters to the available zone machine learning model;

[0088] Control the availability zone machine learning model to send the target model parameters to the cloud host machine learning model.

[0089] In an embodiment of the present invention, the monitoring prediction model at the availability zone level, i.e., the availability zone machine learning model, uploads the second model gradient parameters obtained by aggregation and updating to the monitoring prediction model at the resource pool level, i.e., the resource pool machine learning model, and also continuously iterates and updates to obtain the target model parameters. After the resource pool-level monitoring prediction model is aggregated and updated, the target model parameters are sent down to the availability zone-level monitoring prediction model in each availability zone, and then the availability zone-level monitoring prediction model sends the updated target model parameters to the cloud host, and finally each cloud host continuously iterates and trains after updating the target model parameters. Among them, the aggregate update refers to obtaining the average value of the second model gradient parameters of at least one availability zone machine learning model and updating the average value.

[0090] In some embodiments of the present invention, controlling the available zone machine learning model to update the first model gradient parameter based on the training data of the cloud host in the available zone to obtain the second model gradient parameter includes:

[0091] Controlling the available zone machine learning model to update the first model gradient parameter based on the training data of the cloud host in the available zone to obtain the model gradient parameter to be processed;

[0092] Controlling the available zone machine learning model to send the to-be-processed model gradient parameter to the cloud host machine learning model;

[0093] Controlling the cloud host machine learning model to use the training data of the cloud host to update the gradient parameters of the model to be processed to obtain a second model gradient parameter;

[0094] Control the cloud host machine learning model to send the second model gradient parameters to the available zone machine learning model.

[0095] In an embodiment of the present invention, the step of updating the first model gradient parameter based on the training data of the cloud host in the available zone to obtain the second model gradient parameter includes: controlling the available zone machine learning model to update the first model gradient parameter based on the training data of the cloud host in the available zone to obtain the model gradient parameter to be processed. Controlling the available zone machine learning model to send the model gradient parameter to be processed to the cloud host machine learning model; controlling the cloud host machine learning model to update the model gradient parameter to be processed using the training data of the cloud host to obtain the second model gradient parameter; controlling the cloud host machine learning model to send the second model gradient parameter to the available zone machine learning model.

[0096] In some embodiments of the present invention, controlling the cloud host machine learning model to use the training data of the cloud host to update the gradient parameters of the model to be processed to obtain the second model gradient parameters includes:

[0097] Adding at least one preset proportion of Gaussian noise to the training data of the cloud host to obtain target training data of the cloud host;

[0098] The target training data is used to update the gradient parameters of the model to be processed to obtain the second model gradient parameters.

[0099] In an embodiment of the present invention, the cloud host machine learning model is controlled to use the training data of the cloud host to update the gradient parameters of the model to be processed. When the second model gradient parameters are obtained, at least one preset proportion of Gaussian noise is added to the training data of the cloud host to obtain the target training data of the cloud host; the target training data is used to update the gradient parameters of the model to be processed to obtain the second model gradient parameters.

[0100] In some embodiments of the invention, the method comprises:

[0101] Based on a preset Gaussian noise generation function, obtaining Gaussian noise of the preset proportion;

[0102] The Gaussian noise generating function is:

[0103]

[0104] Wherein, f(x) is the Gaussian operation, σ is the standard deviation of the training data, μ is the mean of the training data, and x is any of the training data.

[0105] In an embodiment of the present invention, the cloud host machine learning model is trained using the training data in the cloud host, and interactively updates the model gradient parameters with the available zone machine learning model. The specific process is as follows: The available zone machine learning model is updated using the training data of the cloud host in the entire available zone and the first model gradient parameters sent by the cloud host machine learning model to obtain the model gradient parameters to be processed. Then, the available zone machine learning model sends the model gradient parameters to be processed to the cloud host machine learning model, and the cloud host machine learning model uses the comparison processing training data, and uses the training data after the comparison processing to optimize the model gradient parameters to be processed, and obtains the second model gradient parameters.

[0106] The initialization parameter of the machine learning model in the availability zone is w0. During the t+1th training iteration, the cloud host machine learning model obtains the model gradient parameters to be processed from the machine learning model in the availability zone and performs model training. The local cross entropy loss value at the edge is defined as The comparison loss between the cloud host machine learning model and the availability zone machine learning model is defined as The cloud host machine learning model receives the model parameters w from the availability zone machine learning model t The i-th update of the cloud host machine learning model is defined as Define Z to represent the current iteration of the cloud host machine learning model, Z prev Represents the model of the previous iteration, Z global is the available zone machine learning model, and τ represents the temperature parameter.

[0107] When the cloud host machine learning model updates parameters based on the availability zone machine learning model, contrastive learning is used. The contrastive learning loss value function in the relevant technology can be defined as the following formula:

[0108]

[0109] This method cannot make the cloud host machine learning model as close as possible to the availability zone machine learning model, that is, it cannot effectively improve the similarity distribution between the cloud host machine learning model and the availability zone machine learning model to obtain higher model accuracy.

[0110] In an embodiment of the present invention, a method combining strong enhancement contrast and weak enhancement contrast is adopted, and strong enhancement contrast is used to reduce the distribution divergence of the cloud host machine learning model and the available zone machine learning model, and weak enhancement contrast is used to solve the overfitting problem in the process. The specific implementation of weak enhancement contrast and strong enhancement contrast is to add different proportions of Gaussian noise to the data set by using the Gaussian noise method to achieve data enhancement. The function generated by Gaussian noise is:

[0111]

[0112] Among them, μ represents the mean and σ represents the standard deviation.

[0113] In the embodiment of the present invention, σ in the Gaussian noise of the weak enhancement contrast takes a smaller value. The loss function in the weak enhancement contrast process can be expressed by the following formula:

[0114]

[0115] Among them, L C represents the loss value of weak enhancement, and z′ represents the model parameters of weak enhancement.

[0116] In the embodiment of the present invention, σ in the Gaussian noise of the strong enhancement contrast takes a larger value. The loss function in the strong enhancement contrast process can be expressed by the following formula:

[0117]

[0118] Among them, L D represents the loss value of strong enhancement, and z″ represents the model parameters of strong enhancement.

[0119] In the embodiment of the present invention, the expression of the joint loss value combining the weak enhancement contrast and the strong enhancement contrast can be expressed by the following formula:

[0120]

[0121] Among them, β represents the balanced contrast loss value L of weak enhancement contrast C The distribution divergence minimization loss L for strong enhancement contrast D The coefficient of is more general when β=1. If there are k cloud host machine learning models, the updated model parameters w k t+1

[0122] In this embodiment of the present invention, the cloud host machine learning model updates the model weight information w k t+1 Upload to the available zone machine learning model, the available zone machine learning model samples and collects the received k weight parameters to obtain the set S t .t The weights of the cloud host machine learning model in the weighted average are used to obtain the parameter information of the available zone machine learning model. The parameter update process can be expressed by the following formula:

[0123]

[0124] Where n represents the cloud host machine learning model, n k Represents the data set of the cloud host machine learning model k.

[0125] In an embodiment of the present invention, the specific process of interactively updating the model gradient parameters between the cloud host machine learning model and the available zone machine learning model is similar to the specific process of interactively updating the model gradient parameters between the available zone machine learning model and the resource pool machine learning model, and the present invention will not repeat them here.

[0126] After the training is completed, a resource pool-level monitoring prediction model and an updated monitoring prediction model in each availability zone are obtained, and the monitoring prediction model in the availability zone can obtain performance close to that of the monitoring prediction model at the entire resource pool level. Then the model gradient parameter information in the updated availability zone-level model is re-sent to the cloud host in each availability zone to obtain several updated edge models in the cloud host. By obtaining, iteratively updating, and analyzing the gradient information of the edge model in all cloud hosts in the entire availability zone, the overall feature information of the monitoring data resources in the entire availability zone can be obtained, which can more accurately reflect the characteristics of data anomalies in different scenarios.

[0127] In an embodiment of the present invention, at least one cloud host forms an availability zone, at least one availability zone forms a resource pool, a preset cloud host monitoring model is configured for any cloud host, and the cloud host monitoring model is used to monitor the operating status data of the cloud host; a preset availability zone monitoring model is configured for any availability zone, and the availability zone monitoring model is used to monitor the operating status data of at least one cloud host in the availability zone; a preset resource pool monitoring model is configured for the resource pool, and the resource pool monitoring model is used to monitor the operating status data of the cloud host in at least one availability zone in the resource pool. The present invention adopts a federated learning method to construct a multi-layer federated framework in a distributed scenario using a federated learning method, making full use of data feature information at different levels and effectively protecting data privacy in edge cloud hosts.

[0128] The present invention uses an improved enhanced contrastive learning method on the traditional federated learning aggregation update method to solve the problem of low prediction accuracy of the edge model caused by the non-independent and identically distributed data and the small data scale. The present invention uses a method based on federated learning and enhanced contrastive learning to apply to the tenant monitoring problem in the field of cloud computing. Tenant monitoring is essentially a distributed structure, which usually collects resource information in the edge cloud host to the central end and monitors it in real time. This passive monitoring method has problems such as the delay caused by the reporting of monitoring data, which cannot handle monitoring alarms in time. For traditional active monitoring solutions, it is necessary to obtain data information running in the cloud host and use it for monitoring the training of the prediction model, which has the problem of data privacy security. The present invention constructs a multi-layer federated framework based on the method of federated learning, which makes full use of data feature information at different levels and ensures data privacy security. In federated learning, the central model can often learn the feature information distribution of the data more comprehensively by obtaining the gradient parameters of the edge model for updating, thereby having higher model accuracy and more stable performance. At the same time, the edge model is affected by the non-independent and identically distributed data and the small data scale, resulting in low model accuracy and low model generalization ability. In response to the above problems, the present invention uses an enhanced contrastive learning method to enable the edge model to obtain performance close to that of the central model, improve the accuracy of the edge model, and effectively reduce the impact of non-independent and identically distributed data and small data scale in distributed scenarios. The impact of non-independent and identically distributed data and small data scale at the edge on the edge model in distributed scenarios is avoided by enhanced contrastive learning. The edge model in each cloud host achieves performance close to that of the available zone-level model, with higher model accuracy.

[0129] Reference Figure 3 , shows a flowchart of the steps of another cloud host data monitoring method provided in an embodiment of the present invention, which may specifically include the following steps:

[0130] Step 301, start.

[0131] Step 302: data preprocessing.

[0132] Step 303: model initialization.

[0133] Step 304: edge model training.

[0134] Step 305: zone-level model training may be performed.

[0135] Step 306: training the resource pool-level model, and using the trained parameters to update the available zone-level model and the edge-end model.

[0136] Step 307, determine whether the training is finished. If yes, proceed to step 308. If no, proceed to step 304, perform iterative update of the model.

[0137] Step 308, obtaining the trained edge model.

[0138] Step 309, end.

[0139] Reference Figure 4 , shows a flowchart of steps for monitoring a cloud host using a cloud host monitoring model provided in an embodiment of the present invention, which may specifically include the following steps:

[0140] Step 401, start.

[0141] Step 402: embed the model into the monitoring plug-in.

[0142] Step 403, setting the data analysis interval.

[0143] Step 404: The model analyzes the data in real time and reports a prediction alarm.

[0144] Step 405, end.

[0145] It should be noted that, for the sake of simplicity, the method embodiments are described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present invention are not limited by the order of the actions described, because according to the embodiments of the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.

[0146] Reference Figure 5 , shows a structural block diagram of a data monitoring device for a cloud host provided in an embodiment of the present invention, which is applied to a cloud host, at least one of the cloud hosts forms an availability zone, and at least one of the availability zones forms a resource pool, and specifically may include the following modules:

[0147] The first configuration module 501 is used to configure a preset cloud host monitoring model for any of the cloud hosts, and use the cloud host monitoring model to monitor the operating status data of the cloud host;

[0148] A second configuration module 502 is used to configure a preset availability zone monitoring model for any of the availability zones, and use the availability zone monitoring model to monitor the operating status data of at least one of the cloud hosts in the availability zone;

[0149] The third configuration module 503 is used to configure a preset resource pool monitoring model for the resource pool, and use the resource pool monitoring model to monitor the operating status data of the cloud host in at least one of the available zones in the resource pool.

[0150] In an optional embodiment of the present invention, the device comprises:

[0151] A fourth configuration module, used to configure a preset cloud host machine learning model for any of the cloud hosts;

[0152] A fifth configuration module, configured to configure a preset availability zone machine learning model for the availability zone;

[0153] A sixth configuration module, used to configure a preset resource pool machine learning model for the resource pool;

[0154] A data acquisition module, used to acquire the training running status data of the cloud host in the resource pool and the label corresponding to the training running status data; the label is used to indicate whether the cloud host corresponding to the training running status data has an abnormality;

[0155] The training data is used as a module to use the training running status data of the cloud host and the label corresponding to the training running status data as the training data of the cloud host;

[0156] A training module is used to use the training data of the cloud host to train the cloud host machine learning model, the available zone machine learning model and the resource pool machine learning model to obtain the cloud host monitoring model, the available zone monitoring model and the resource pool monitoring model.

[0157] In an optional embodiment of the present invention, the cloud host machine learning model has initialization parameters; the training module includes:

[0158] An initialization parameter updating submodule, configured to train the cloud host machine learning model by using the training data of the cloud host, update the initialization parameters, and obtain a first model gradient parameter;

[0159] A first model gradient parameter sending submodule, used to control the cloud host machine learning model to send the first model gradient parameter to the available zone machine learning model;

[0160] The first model gradient parameter updating submodule is used to control the available zone machine learning model to update the first model gradient parameters based on the training data of the cloud host in the available zone to obtain the second model gradient parameters.

[0161] In an optional embodiment of the present invention, the training module includes:

[0162] A second model gradient parameter sending submodule, used to control the available zone machine learning model to send the second model gradient parameter to the resource pool machine learning model;

[0163] A second model gradient parameter updating submodule is used to control the resource pool machine learning model to update the second model gradient parameters based on the training data of the cloud host in the resource pool to obtain the target model parameters;

[0164] A first target model parameter sending submodule, used to control the resource pool machine learning model to send the target model parameters to the available zone machine learning model;

[0165] The second target model parameter sending submodule is used to control the available zone machine learning model to send the target model parameters to the cloud host machine learning model.

[0166] In an optional embodiment of the present invention, the first model gradient parameter updating submodule includes:

[0167] A first model gradient parameter updating unit, configured to control the available zone machine learning model to update the first model gradient parameters based on the training data of the cloud host in the available zone, to obtain a model gradient parameter to be processed;

[0168] A to-be-processed model gradient parameter sending unit, used to control the available zone machine learning model to send the to-be-processed model gradient parameter to the cloud host machine learning model;

[0169] A to-be-processed model gradient parameter updating unit, used to control the cloud host machine learning model to use the training data of the cloud host to update the to-be-processed model gradient parameter to obtain a second model gradient parameter;

[0170] The second model gradient parameter sending unit is used to control the cloud host machine learning model to send the second model gradient parameter to the available area machine learning model.

[0171] In an optional embodiment of the present invention, the to-be-processed model gradient parameter updating unit includes:

[0172] An adding subunit, configured to add at least one Gaussian noise of a preset proportion to the training data of the cloud host to obtain target training data of the cloud host;

[0173] The updating subunit is used to update the gradient parameters of the model to be processed using the target training data to obtain the second model gradient parameters.

[0174] In an optional embodiment of the present invention, the device comprises:

[0175] A Gaussian noise acquisition module, used to acquire Gaussian noise of the preset proportion based on a preset Gaussian noise generation function;

[0176] The Gaussian noise generating function is:

[0177]

[0178] Wherein, f(x) is the Gaussian operation, σ is the standard deviation of the training data, μ is the mean of the training data, and x is any of the training data.

[0179] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0180] In addition, an embodiment of the present invention further provides an electronic device, such as Figure 6 As shown, it includes a processor 601, a communication interface 602, a memory 603 and a communication bus 604, wherein the processor 601, the communication interface 602, and the memory 603 communicate with each other through the communication bus 604.

[0181] Memory 603, used for storing computer programs;

[0182] The processor 601 is used to execute the program stored in the memory 603 to implement the following steps:

[0183] Configuring a preset cloud host monitoring model for any of the cloud hosts, and using the cloud host monitoring model to monitor the operating status data of the cloud host;

[0184] Configuring a preset availability zone monitoring model for any of the availability zones, and using the availability zone monitoring model to monitor the operating status data of at least one of the cloud hosts in the availability zone;

[0185] A preset resource pool monitoring model is configured for the resource pool, and the resource pool monitoring model is used to monitor the operating status data of at least one cloud host in the available zone in the resource pool.

[0186] In an optional embodiment of the present invention, the method includes:

[0187] Configuring a preset cloud host machine learning model for any of the cloud hosts;

[0188] Configuring a preset availability zone machine learning model for the availability zone;

[0189] Configuring a preset resource pool machine learning model for the resource pool;

[0190] Obtaining training operation status data of the cloud host in the resource pool and a label corresponding to the training operation status data; the label is used to indicate whether the cloud host corresponding to the training operation status data has an abnormality;

[0191] Using the training running status data of the cloud host and the label corresponding to the training running status data as the training data of the cloud host;

[0192] The cloud host machine learning model, the availability zone machine learning model and the resource pool machine learning model are trained using the training data of the cloud host to obtain the cloud host monitoring model, the availability zone monitoring model and the resource pool monitoring model.

[0193] In an optional embodiment of the present invention, the cloud host machine learning model has initialization parameters; the cloud host machine learning model, the available zone machine learning model and the resource pool machine learning model are trained using the training data of the cloud host to obtain the cloud host monitoring model, the available zone monitoring model and the resource pool monitoring model, including:

[0194] Training the cloud host machine learning model by using the training data of the cloud host, updating the initialization parameters, and obtaining a first model gradient parameter;

[0195] Controlling the cloud host machine learning model to send the first model gradient parameter to the available zone machine learning model;

[0196] The available zone machine learning model is controlled to update the first model gradient parameters based on the training data of the cloud host in the available zone to obtain the second model gradient parameters.

[0197] In an optional embodiment of the present invention, the cloud host machine learning model, the available zone machine learning model and the resource pool machine learning model are trained using the training data of the cloud host to obtain the cloud host monitoring model, the available zone monitoring model and the resource pool monitoring model, including:

[0198] Controlling the available zone machine learning model to send the second model gradient parameter to the resource pool machine learning model;

[0199] Controlling the resource pool machine learning model to update the second model gradient parameters based on the training data of the cloud host in the resource pool to obtain target model parameters;

[0200] Controlling the resource pool machine learning model to send the target model parameters to the available zone machine learning model;

[0201] Control the availability zone machine learning model to send the target model parameters to the cloud host machine learning model.

[0202] In an optional embodiment of the present invention, controlling the available zone machine learning model to update the first model gradient parameter based on the training data of the cloud host in the available zone to obtain the second model gradient parameter includes:

[0203] Controlling the available zone machine learning model to update the first model gradient parameter based on the training data of the cloud host in the available zone to obtain the model gradient parameter to be processed;

[0204] Controlling the available zone machine learning model to send the to-be-processed model gradient parameter to the cloud host machine learning model;

[0205] Controlling the cloud host machine learning model to use the training data of the cloud host to update the gradient parameters of the model to be processed to obtain a second model gradient parameter;

[0206] Control the cloud host machine learning model to send the second model gradient parameters to the available zone machine learning model.

[0207] In an optional embodiment of the present invention, the controlling the cloud host machine learning model to use the training data of the cloud host to update the gradient parameters of the model to be processed to obtain the second model gradient parameters includes:

[0208] Adding at least one preset proportion of Gaussian noise to the training data of the cloud host to obtain target training data of the cloud host;

[0209] The target training data is used to update the gradient parameters of the model to be processed to obtain the second model gradient parameters.

[0210] In an optional embodiment of the present invention, the method includes:

[0211] Based on a preset Gaussian noise generation function, obtaining Gaussian noise of the preset proportion;

[0212] The Gaussian noise generating function is:

[0213]

[0214] Wherein, f(x) is the Gaussian operation, σ is the standard deviation of the training data, μ is the mean of the training data, and x is any of the training data.

[0215] The communication bus mentioned in the above terminal can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0216] The communication interface is used for communication between the above terminal and other devices.

[0217] The memory may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0218] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0219] like Figure 7 As shown, in another embodiment provided by the present invention, a computer-readable storage medium 701 is also provided, in which instructions are stored. When the computer-readable storage medium 701 is run on a computer, the computer executes a cloud host data monitoring method described in the above embodiment.

[0220] In another embodiment provided by the present invention, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute a cloud host data monitoring method described in the above embodiment.

[0221] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk Solid State Disk (SSD)), etc.

[0222] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants 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, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0223] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0224] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A data monitoring method for a cloud host, characterized in that: Applied to a cloud host, at least one of the cloud hosts forms an availability zone, at least one of the availability zones forms a resource pool, and the method includes: Configuring a preset cloud host monitoring model for any of the cloud hosts, and using the cloud host monitoring model to monitor the operating status data of the cloud host; Configuring a preset availability zone monitoring model for any of the availability zones, and using the availability zone monitoring model to monitor the operating status data of at least one of the cloud hosts in the availability zone; A preset resource pool monitoring model is configured for the resource pool, and the resource pool monitoring model is used to monitor the operating status data of at least one cloud host in the available zone in the resource pool.

2. The method according to claim 1, characterized in that The method comprises: Configuring a preset cloud host machine learning model for any of the cloud hosts; Configuring a preset availability zone machine learning model for the availability zone; Configuring a preset resource pool machine learning model for the resource pool; Obtaining training operation status data of the cloud host in the resource pool and a label corresponding to the training operation status data; the label is used to indicate whether the cloud host corresponding to the training operation status data has an abnormality; Using the training running status data of the cloud host and the label corresponding to the training running status data as the training data of the cloud host; The cloud host machine learning model, the available zone machine learning model and the resource pool machine learning model are trained using the training data of the cloud host to obtain the cloud host monitoring model, the available zone monitoring model and the resource pool monitoring model.

3. The method according to claim 2, characterized in that The cloud host machine learning model has initialization parameters; the cloud host machine learning model, the available zone machine learning model and the resource pool machine learning model are trained using the training data of the cloud host to obtain the cloud host monitoring model, the available zone monitoring model and the resource pool monitoring model, including: Training the cloud host machine learning model by using the training data of the cloud host, updating the initialization parameters, and obtaining a first model gradient parameter; Controlling the cloud host machine learning model to send the first model gradient parameter to the available zone machine learning model; The available zone machine learning model is controlled to update the first model gradient parameters based on the training data of the cloud host in the available zone to obtain the second model gradient parameters.

4. The method according to claim 3, characterized in that The method of using the training data of the cloud host to train the cloud host machine learning model, the available zone machine learning model, and the resource pool machine learning model to obtain the cloud host monitoring model, the available zone monitoring model, and the resource pool monitoring model includes: Controlling the available zone machine learning model to send the second model gradient parameter to the resource pool machine learning model; Controlling the resource pool machine learning model to update the second model gradient parameters based on the training data of the cloud host in the resource pool to obtain target model parameters; Controlling the resource pool machine learning model to send the target model parameters to the available zone machine learning model; Control the availability zone machine learning model to send the target model parameters to the cloud host machine learning model.

5. The method according to claim 3, characterized in that: The controlling the available zone machine learning model to update the first model gradient parameter based on the training data of the cloud host in the available zone to obtain the second model gradient parameter includes: Controlling the available zone machine learning model to update the first model gradient parameter based on the training data of the cloud host in the available zone to obtain the model gradient parameter to be processed; Controlling the available zone machine learning model to send the to-be-processed model gradient parameter to the cloud host machine learning model; Controlling the cloud host machine learning model to use the training data of the cloud host to update the gradient parameters of the model to be processed to obtain a second model gradient parameter; Control the cloud host machine learning model to send the second model gradient parameters to the available zone machine learning model.

6. The method according to claim 5, characterized in that The controlling the cloud host machine learning model to use the training data of the cloud host to update the gradient parameter of the model to be processed to obtain the second model gradient parameter includes: Adding at least one preset proportion of Gaussian noise to the training data of the cloud host to obtain target training data of the cloud host; The target training data is used to update the gradient parameters of the model to be processed to obtain the second model gradient parameters.

7. The method according to claim 6, characterized in that The method comprises: Based on a preset Gaussian noise generation function, obtaining Gaussian noise of the preset proportion; The Gaussian noise generating function is: Wherein, f(x) is the Gaussian operation, σ is the standard deviation of the training data, μ is the mean of the training data, and x is any of the training data.

8. A data monitoring device for a cloud host, characterized in that: Applied to a cloud host, at least one of the cloud hosts forms an availability zone, at least one of the availability zones forms a resource pool, and the device comprises: A first configuration module, configured to configure a preset cloud host monitoring model for any of the cloud hosts, and use the cloud host monitoring model to monitor the operating status data of the cloud host; A second configuration module is used to configure a preset availability zone monitoring model for any of the availability zones, and use the availability zone monitoring model to monitor the operating status data of at least one of the cloud hosts in the availability zone; The third configuration module is used to configure a preset resource pool monitoring model for the resource pool, and use the resource pool monitoring model to monitor the operating status data of at least one cloud host in the available zone in the resource pool.

9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; The memory is used to store computer programs; The processor is used to implement the method according to any one of claims 1 to 7 when executing the program stored in the memory.

10. One or more computer-readable media having instructions stored thereon, which when executed by one or more processors cause the processors to perform the method according to any one of claims 1 to 7.

Images