Computing resource monitoring method in multi-cloud environment

Abstract

The invention discloses a computing resource monitoring method in a multi-cloud environment, and particularly relates to the technical field of resource monitoring. The method comprises the following steps: constructing a multi-cloud monitoring data set by collecting operation index data of virtual computing resources and physical host resource distribution data in a multi-cloud environment; according to the multi-cloud monitoring data set, constructing a resource adjacency topological relation graph between the virtual instances, and extracting load interference feature data of adjacent virtual instances; meanwhile, identifying load fluctuation characteristics and abrupt change trends of the target virtual instances, and generating abnormal fluctuation identification data of the target virtual instances; and based on the load interference characteristic data and the abnormal fluctuation identification data, establishing a distinguishing model of a real load of the target virtual instance and an interference load of the adjacent virtual instance, adjusting an abnormal monitoring judgment logic according to a distinguishing result, and outputting monitoring state label data after interference correction. Therefore, real load change and adjacent instance load interference are effectively distinguished, and the accuracy of resource monitoring and the reliability of resource scheduling decision are improved.

Description

Technical Field

[0001] This invention relates to the field of resource monitoring technology, and more specifically, to a method for monitoring computing resources in a multi-cloud environment. Background Technology

[0002] With the development and application of cloud computing technology, enterprises are choosing to deploy their business systems simultaneously in multi-cloud environments built by multiple different cloud service providers to achieve higher resource utilization and service reliability. In a multi-cloud environment, users typically obtain virtual computing resources and execute specific tasks through virtualization technology. However, due to significant differences in the underlying physical host resource configuration strategies adopted by different cloud service providers, there is a potential for mutual interference between virtual computing resources.

[0003] In existing technologies, when monitoring the operational status of virtual computing resources in a multi-cloud environment, the impact of load interference caused by the sharing of physical resources between virtual computing resources is not considered, making it difficult to accurately distinguish between the actual load changes in the operational status data of virtual computing resources and the interference fluctuations caused by adjacent virtual computing resources. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for monitoring computing resources in a multi-cloud environment to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for monitoring computing resources in a multi-cloud environment includes the following steps: S1: Collect operational metrics data and physical host resource distribution data of virtual computing resources deployed in multi-cloud environments, and perform time alignment and normalization processing to generate a standardized multi-cloud monitoring dataset. S2: Based on a standardized multi-cloud monitoring dataset, construct a resource adjacency topology graph between virtual instances; S3: Based on the resource adjacency topology graph, calculate the load similarity index between virtual instances and extract the load interference feature data of adjacent virtual instances; S4: Based on the operational metric change data extracted from the virtual instance in the multi-cloud monitoring dataset, identify the load fluctuation characteristics and sudden change trends of the target virtual instance, and generate abnormal fluctuation identification data of the target virtual instance. S5: Based on load interference characteristic data and abnormal fluctuation identification data, establish a model to distinguish between the real load of the target virtual instance and the interference load of adjacent virtual instances; S6: Based on the distinction between the actual load of the target virtual instance and the interference load of adjacent virtual instances, adjust the abnormal monitoring and judgment logic of multiple cloud computing resources, and output the monitoring status label data after interference correction.

[0006] In a preferred embodiment, S1 specifically refers to: Collect operational metrics data of virtual computing resources deployed in multiple cloud service provider environments; Collect distribution data of physical host resources deployed in multiple different cloud service provider environments; Perform unified time base calibration on the operational metrics data of virtual computing resources and the distribution data of physical host resources; The calibrated virtual computing resource operation index data and physical host resource distribution data are normalized to generate a standardized multi-cloud monitoring dataset.

[0007] In a preferred embodiment, S2 specifically refers to: Based on the standardized multi-cloud monitoring dataset, virtual instances are grouped and a mapping table between virtual instances and physical host resources is generated. Based on the mapping table, a unique node number is assigned to each physical host resource. All virtual instances mapped to the same physical host resource node number are marked as same-level nodes, and all virtual instances mapped to different physical host resource node numbers are marked as different-level nodes. Undirected edges are established between nodes at the same level based on shared processor core identifiers, shared memory channel identifiers, and shared input / output bus identifiers. No edges are established between nodes at different levels, generating a resource adjacency topology graph between virtual instances.

[0008] In a preferred embodiment, S3 specifically refers to: Based on the resource adjacency topology graph, processor utilization data, memory utilization data, and network bandwidth utilization data of each node in each adjacent virtual instance are extracted from the standardized multi-cloud monitoring dataset. Calculate the similarity of processor utilization data, memory utilization data, and network bandwidth utilization data between node pairs of each adjacent virtual instance, and perform a weighted sum to obtain the load similarity index between node pairs of each adjacent virtual instance. Extract adjacent virtual instance node pairs whose load similarity index exceeds a preset similarity threshold, and generate load interference feature data of adjacent virtual instances.

[0009] In a preferred embodiment, S4 specifically refers to: Based on a standardized multi-cloud monitoring dataset, the node identification information of the target virtual instance is determined; Based on the node identification information of the target virtual instance, extract the disk input / output operation rate data, disk latency data, and processor utilization data of the target virtual instance within the continuous monitoring time window from the standardized multi-cloud monitoring dataset; The rate of change of disk input / output operation rate data, disk latency data, and processor utilization data within a continuous monitoring time window is calculated, and the fluctuation characteristics of the corresponding rate of change are identified. For each fluctuation feature, abrupt trend detection is performed to generate abnormal fluctuation identification data for the target virtual instance.

[0010] In a preferred embodiment, S5 specifically refers to: The interference intensity level of adjacent virtual instance node pairs is determined based on the load interference characteristic data of adjacent virtual instances; The abnormal fluctuation level of the target virtual instance is determined based on the abnormal fluctuation identification data of the target virtual instance; Based on the interference intensity level and abnormal fluctuation level, a model is established to distinguish between the actual load of the target virtual instance and the interference load of adjacent virtual instances. Based on the differentiation model, each abnormal fluctuation feature in the abnormal fluctuation identification data of the target virtual instance is classified and determined to determine whether each abnormal fluctuation feature belongs to the real load change of the target virtual instance or the interference load change of the adjacent virtual instance.

[0011] In a preferred embodiment, S6 specifically refers to: When the load fluctuation category of the target virtual instance is determined to be the load change of the adjacent virtual instance, the abnormal fluctuation identification data of the target virtual instance is marked and corrected to a non-abnormal monitoring status label in the standardized multi-cloud monitoring dataset. When the load fluctuation category of the target virtual instance is determined to be the actual load change of the target virtual instance, the abnormal fluctuation identification data of the target virtual instance is marked in the standardized multi-cloud monitoring dataset and identified as an abnormal monitoring status label. The standardized multi-cloud monitoring dataset is updated based on non-abnormal monitoring status labels and abnormal monitoring status labels, and the corrected target virtual instance monitoring status label data is output.

[0012] The technical effects and advantages of the computing resource monitoring method in a multi-cloud environment of the present invention are as follows: By constructing a standardized multi-cloud monitoring dataset, unified management of cross-cloud environment data is achieved, eliminating monitoring biases caused by clock differences and inconsistencies in units. Constructing a resource adjacency topology graph accurately depicts the physical resource sharing relationships between virtual instances. Calculating the load similarity index and extracting load interference characteristic data identifies monitoring anomalies caused by load fluctuations in neighboring virtual instances. Identifying the load fluctuation characteristics and abrupt change trends of target virtual instances effectively separates real load changes from interference fluctuations. Introducing a differentiation model based on interference intensity levels and abnormal fluctuation levels enables the separation and determination of real load changes from neighboring interference loads. Based on the differentiation results, the anomaly monitoring judgment logic for multi-cloud computing resources is dynamically adjusted, outputting interference-corrected monitoring status label data, significantly improving the accuracy and stability of resource monitoring and providing reliable data support for multi-cloud resource scheduling and capacity planning. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of a computing resource monitoring method in a multi-cloud environment according to the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0015] Example

[0016] Figure 1 This invention provides a method for monitoring computing resources in a multi-cloud environment, comprising the following steps: S1: Collect operational metrics data and physical host resource distribution data of virtual computing resources deployed in multi-cloud environments, and perform time alignment and normalization processing to generate a standardized multi-cloud monitoring dataset. S2: Based on a standardized multi-cloud monitoring dataset, construct a resource adjacency topology graph between virtual instances; S3: Based on the resource adjacency topology graph, calculate the load similarity index between virtual instances and extract the load interference feature data of adjacent virtual instances; S4: Based on the operational metric change data extracted from the virtual instance in the multi-cloud monitoring dataset, identify the load fluctuation characteristics and sudden change trends of the target virtual instance, and generate abnormal fluctuation identification data of the target virtual instance. S5: Based on load interference characteristic data and abnormal fluctuation identification data, establish a model to distinguish between the real load of the target virtual instance and the interference load of adjacent virtual instances; S6: Based on the distinction between the actual load of the target virtual instance and the interference load of adjacent virtual instances, adjust the abnormal monitoring and judgment logic of multiple cloud computing resources, and output the monitoring status label data after interference correction.

[0017] S1: Collect operational metrics data and physical host resource distribution data of virtual computing resources deployed in multi-cloud environments, perform time alignment and normalization processing, and generate a standardized multi-cloud monitoring dataset, including: Collect operational metrics data of virtual computing resources deployed in multiple cloud service provider environments; A multi-cloud environment refers to a user simultaneously deploying computing resources across two or more cloud computing environments provided by different cloud service providers, with each environment maintained and managed separately by its respective provider. Virtual computing resources refer to virtual devices such as virtual servers, virtual network storage devices, and virtual containers created within a cloud computing environment using virtualization technology. Each virtual device can independently execute user-assigned computing tasks. Performance metrics data are quantitative measurements of the operational status of virtual computing resources, such as the percentage of processor usage, memory usage, network throughput, and the frequency and latency of disk read / write operations. Each performance metric is continuously acquired in real-time to reflect the current operational status of the virtual server.

[0018] Collect distribution data of physical host resources deployed in multiple different cloud service provider environments; Physical host resource distribution data refers to the resource configuration and allocation of physical servers, physical storage devices, and network interface devices that actually host and run virtual computing resources in a cloud computing environment. For example, this includes the total number of CPU cores on a physical server and the current usage of each core, the capacity and actual used capacity of physical memory modules, the currently available and allocated network bandwidth of each network interface device, and the total capacity, remaining available capacity, and used capacity of storage devices. This physical host resource distribution data also needs to be collected in real time. For instance, continuing with the example of the two cloud service providers (A and B) deployed by the aforementioned enterprise, for each physical server within these two cloud service providers' environments, the total number of processor cores, the number of used processor cores, the total memory capacity, the used memory capacity, the total and remaining disk capacity, and the available and used bandwidth of the network interfaces are collected for each physical server to obtain a complete and accurate physical resource distribution.

[0019] Perform unified time base calibration on the operational metrics data of virtual computing resources and the distribution data of physical host resources; Unified time reference calibration refers to the unified time alignment of virtual computing resource operation index data and physical host resource distribution data with the same standard clock frequency as the reference. Specifically, it involves: first, determining a unified clock standard, such as Coordinated Universal Time (UTC) or a unified time protocol standard clock signal as the time reference; recording the original timestamp corresponding to the acquisition of each piece of virtual computing resource operation index data and each piece of physical host resource distribution data; then, calculating the time difference between the original timestamp of each piece of data and the unified standard clock signal, and using the time difference as an offset to adjust the timestamps of each piece of operation index data and physical host resource distribution data, thereby enabling all data to have a unified time reference reference.

[0020] The calibrated virtual computing resource operation index data and physical host resource distribution data are normalized to generate a standardized multi-cloud monitoring dataset. Normalization refers to the scaling transformation of all time-based calibrated operational metric data and resource distribution data, ensuring that all data are converted to the same value range. Specifically, this involves: first, determining the value range of various data types, such as the maximum and minimum possible values ​​for processor utilization, memory utilization, and disk read / write speeds; then, subtracting the minimum possible value within the corresponding value range from the original value of each data item, and dividing by the difference between the maximum and minimum possible values ​​to obtain a unified standardized data value, ensuring that the standardized data value is between zero and one; finally, performing normalization calculations on all operational metric data and physical host resource distribution data to obtain a standardized multi-cloud monitoring dataset with a unified scale.

[0021] S2: Based on a standardized multi-cloud monitoring dataset, construct a resource adjacency topology graph between virtual instances, including: Based on the standardized multi-cloud monitoring dataset, virtual instances are grouped and a mapping table between virtual instances and physical host resources is generated. Virtual instances are virtual devices, such as virtual servers, virtual network storage devices, or virtual containers, built using virtualization technology, capable of independently completing computing tasks. Physical host resources refer to the physical servers, storage devices, and network interface devices on which the virtual instances are actually deployed. Virtual instances are grouped according to the distribution of their respective physical host resources based on a standardized multi-cloud monitoring dataset. Assuming an enterprise deploys ten virtual servers in a first cloud service provider environment and fifteen virtual servers in a second cloud service provider environment, each virtual server runs on different physical host resources within its respective cloud service provider environment. The physical host resource distribution data in the standardized multi-cloud monitoring dataset determines the location of the physical host resources corresponding to each virtual server, forming a one-to-one grouping relationship, thus associating each virtual server with the physical server hosting it. For example, the ten virtual servers in the first cloud service provider environment are mapped to four different physical servers, and the fifteen virtual servers in the second cloud service provider environment are mapped to five different physical servers, thereby generating a mapping table, for example:

[0022] Based on the mapping table, a unique node number is assigned to each physical host resource. All virtual instances mapped to the same physical host resource node number are marked as same-level nodes, and all virtual instances mapped to different physical host resource node numbers are marked as different-level nodes. Based on the mapping table between virtual instances and physical host resources, each physical host resource is sequentially assigned a unique and non-repeating node number. The node number serves as a marker to identify different physical host resources, allowing for differentiation between them. For example, considering four physical servers in a first cloud service provider environment and five physical servers in a second cloud service provider environment, each physical server sequentially obtains a distinct node number. The first physical server receives the first node number, the second physical server receives the second node number, and so on, until the last physical server receives the last node number. Based on the physical host resource node number mapped to each virtual instance, all virtual instances are categorized into two node types: if multiple virtual instances are mapped to the same node number, they are defined as same-level nodes, meaning a set of virtual instances running on the same physical server; if multiple virtual instances are mapped to different node numbers, they are defined as different-level nodes, meaning a set of virtual instances running on different physical servers.

[0023] Undirected edges are established between nodes at the same level based on shared processor core identifiers, shared memory channel identifiers, and shared input / output bus identifiers. No edges are established between nodes at different levels, generating a resource adjacency topology graph between virtual instances. The physical resource sharing relationships between virtual instances are represented by a topological graph. Nodes in the graph are labeled virtual instances at the same or different levels, with each node representing one virtual instance. Edges represent resource sharing relationships between two virtual instances; undirected edges are non-directional and only indicate a resource-sharing association between nodes. The criteria for establishing undirected edges between nodes are shared processor core identifiers, shared memory channel identifiers, and shared input / output bus identifiers, representing shared resources for the processor, memory, and input / output devices, respectively. If multiple virtual instances share the same processor core, the same memory channel, or the same input / output bus, an undirected edge is established between nodes at the same level representing the virtual instances. For example, considering a company with four physical servers in a first cloud service provider environment and five physical servers in a second cloud service provider environment, assuming a physical server in the first cloud service provider environment contains four processor cores, each with a unique identifier, and two virtual servers simultaneously occupy a processor core (i.e., occupying a processor core with the same identifier), an undirected edge is established between the nodes representing the two virtual servers. Similarly, if the other three virtual servers occupy the same memory channel or input / output bus, corresponding undirected edges are also established between the nodes. Conversely, no edges are established between nodes at different levels mapped to different node numbers, indicating that there is no resource sharing relationship between the virtual instances. Through this method, the true physical resource sharing relationship between virtual instances can be expressed, thereby generating a resource adjacency topology graph between virtual instances.

[0024] S3: Based on the resource adjacency topology graph, calculate the load similarity index between virtual instances and extract load interference feature data of adjacent virtual instances, including: Based on the resource adjacency topology graph, processor utilization data, memory utilization data, and network bandwidth utilization data of each node in each adjacent virtual instance are extracted from the standardized multi-cloud monitoring dataset. The resource adjacency topology graph is a topology graph generated based on the grouping and mapping relationship between virtual instances and physical host resources. Each node represents a virtual instance, and undirected edges represent the shared physical resource relationship between two virtual instances. A node pair refers to two virtual instances that have a shared relationship and are connected by an undirected edge, called an adjacent virtual instance node pair. For example, an enterprise has two physical servers in a first cloud computing service provider environment, and each physical server runs two virtual servers, so there are a total of four virtual servers, forming several node pairs at the same level. For each adjacent virtual instance node pair, operational indicator data is extracted from a standardized multi-cloud monitoring dataset, including processor utilization data, memory utilization data, and network bandwidth utilization data for each virtual instance. Processor utilization data refers to the proportion of processor computing resources currently occupied by the virtual instance, memory utilization data refers to the proportion of memory space resources occupied by the virtual instance, and network bandwidth utilization data refers to the proportion of network resources currently used by the virtual instance.

[0025] Calculate the similarity of processor utilization data, memory utilization data, and network bandwidth utilization data between node pairs of each adjacent virtual instance, and perform a weighted sum to obtain the load similarity index between node pairs of each adjacent virtual instance. Similarity is used to quantify the consistency between the changing trends and usage characteristics of operational metrics data between adjacent virtual instance node pairs. Processor utilization data similarity refers to the similarity between the changing trends of processor utilization data of two virtual instances within the same time period. Memory utilization data similarity is the similarity between the memory resource usage trends of two virtual instances. Network bandwidth utilization data similarity is the similarity between the network bandwidth usage trends of two virtual instances. The calculation method is as follows: First, taking processor utilization data as an example, calculate the squared value of the difference in processor utilization data between two virtual instances at each consecutive monitoring time point. Sum the squared values ​​of all differences to obtain the total squared difference. Then, divide the total squared difference of processor utilization data by the squared value of the total range of change in the corresponding processor utilization data to obtain the difference ratio of processor utilization data. Example value; then subtract the difference ratio value from the value to obtain the similarity of processor utilization data. The closer the similarity is to 1, the more similar the processor usage trends of the two virtual instances are; the closer the similarity is to zero, the less similar they are. Similarly, calculate the similarity of memory utilization data and network bandwidth utilization data. Assign different weight values ​​to the three similarities, the weight values ​​are determined according to the importance of similarity in load analysis, usually assigning a higher weight to the similarity of processor utilization data, and a second lower weight to memory utilization data and network bandwidth utilization data, respectively. Multiply the similarity of processor utilization data by the corresponding weight value to obtain the weighted similarity of processor utilization data; then calculate the weighted similarity of memory utilization data and the weighted similarity of network bandwidth utilization data separately; finally, add the three weighted similarities to obtain the load similarity index.

[0026] Extract adjacent virtual instance node pairs whose load similarity index exceeds a preset similarity threshold, and generate load interference feature data of adjacent virtual instances; The preset similarity threshold is a pre-determined critical value used to judge the magnitude of the load similarity index. The setting of the critical value depends on specific business needs and experience accumulation. If the load similarity index of adjacent virtual instance node pairs exceeds the preset similarity threshold, it indicates that the virtual instances have correlation and similarity in the trend of operating indicators, and there may be mutual influence and interference between loads, thus marking them as having load interference characteristics. All node pairs with load similarity indices exceeding the preset similarity threshold are extracted to form a load interference characteristic dataset.

[0027] S4: Based on the operational metric change data extracted from the multi-cloud monitoring dataset of virtual instances, identify the load fluctuation characteristics and abrupt change trends of the target virtual instance, and generate abnormal fluctuation identification data for the target virtual instance, including: Based on a standardized multi-cloud monitoring dataset, the node identification information of the target virtual instance is determined; A target virtual instance is a specific virtual computing resource currently of interest during monitoring and analysis. Node identification information is the characteristic data or attribute information that uniquely identifies and distinguishes the target virtual instance's location within a standardized multi-cloud monitoring dataset. For example, an enterprise deploys multiple virtual servers in a first cloud service provider's environment. Each virtual server has a unique identifier; the identifier for virtual server one is a specific string encoding combination, while that for virtual server two is a different string encoding combination. Through a standardized multi-cloud monitoring dataset, based on monitoring needs and task requirements, the specific target virtual server to be monitored and analyzed is determined—that is, the unique node identification information of the target virtual instance. For instance, if the current monitoring task requires monitoring a specific virtual server in the enterprise's first cloud service provider's environment, the unique identifier code of the specific virtual server is used to mark the specific virtual server as the target virtual instance, giving the monitoring and analysis a definite object.

[0028] Based on the node identification information of the target virtual instance, extract the disk input / output operation rate data, disk latency data, and processor utilization data of the target virtual instance within the continuous monitoring time window from the standardized multi-cloud monitoring dataset; A continuous monitoring time window refers to a continuous and fixed-length time period during which data related to the running status of the target virtual instance is continuously and in real-time collected. Disk I / O operation rate data refers to the number of disk read / write operations performed by the target virtual instance per unit time within the continuous monitoring time window. Disk latency data refers to the average waiting time for the target virtual instance to complete disk read / write requests within the continuous monitoring time window. Processor utilization data refers to the proportion of processor computing resources used by the target virtual instance within the continuous monitoring time window. Using a standardized multi-cloud monitoring dataset, based on node identification information, the disk I / O operation rate data, disk latency data, and processor utilization data of the target virtual instance within each continuous monitoring time window are extracted sequentially from the standardized multi-cloud monitoring dataset. For example, if the current continuous monitoring time window is a fixed-length continuous time interval, the disk I / O operation rate data records the number of disk operations per unit time, the disk latency data records the average waiting time for disk operation requests, and the processor utilization data records the proportion of processor resources used by the target virtual instance to the total processor resource capacity per unit time.

[0029] The rate of change of disk input / output operation rate data, disk latency data, and processor utilization data within a continuous monitoring time window is calculated, and the fluctuation characteristics of the corresponding rate of change are identified. The rate of change is a quantitative description of the trend and magnitude of changes in various operational metrics of a target virtual instance within a continuous monitoring time window. Taking disk input / output (I / O) rate data as an example, the difference between consecutive adjacent time points within each continuous monitoring time window is calculated. The difference between two adjacent time points is divided by the disk I / O rate at the previous time point to obtain the relative rate of change of the disk I / O rate data, representing the magnitude and trend of changes in the target virtual instance's disk operation rate. The relative rates of change for disk latency and processor utilization are then calculated using the same method. The trend characteristics of the rate of change for each type of data are identified, specifically by analyzing the rate of change of each data point within the continuous monitoring time window. The patterns, peaks, and fluctuations within a time window are analyzed. For example, taking the disk latency data of a target virtual instance as an example, the difference between every two consecutive adjacent time points of the disk latency data within the continuous monitoring time window is first calculated. Then, the ratio of the difference to the disk latency data at the previous time point is calculated to obtain the disk latency data change rate. The fluctuation trend of the disk latency data change rate within the entire continuous monitoring time window is observed, including whether there are fluctuation peaks, change cycles, or abnormal changes, in order to determine the fluctuation characteristics of the disk latency data change rate. Similarly, the processor utilization rate data and disk input / output operation rate data are also calculated and analyzed according to the above method to determine their corresponding fluctuation characteristics, thereby completing the identification of the fluctuation characteristics of different indicator data of the target virtual instance.

[0030] For each fluctuation feature, abrupt trend detection is performed to generate abnormal fluctuation identification data for the target virtual instance; The mutation trend detection method aims to determine whether there are abnormal data segments in the change rates of various operational indicators of the target virtual instance that deviate from the normal trend and change drastically in a short period of time. First, it analyzes the overall change pattern of each fluctuation feature within a continuous monitoring time window. Then, it calculates the average change trend of each data change rate within the continuous monitoring time window. Next, based on the difference between the change trend and the actual fluctuation change rate, it determines whether an abnormal mutation trend has occurred when the difference exceeds the normal range or a certain critical standard. The time period of the mutation trend and the corresponding indicator are recorded and marked. For example, taking the processor utilization rate data of the target virtual instance as an example, it first calculates the average change trend of the processor utilization rate data within the continuous monitoring time window. If the change rate of the processor utilization rate data exceeds the reasonable range of the average trend within a short continuous period, the processor utilization rate data for that short period is marked as mutation trend data and considered abnormal fluctuation. Similarly, mutation trend detection is also performed on disk input / output operation rate data and disk latency data. Finally, all abnormal mutation trend data detected in the various indicators of the target virtual instance are uniformly marked, summarized, and organized to form abnormal fluctuation identification data.

[0031] S5: Based on load disturbance characteristic data and abnormal fluctuation identification data, establish a model to distinguish between the actual load of the target virtual instance and the disturbance load of adjacent virtual instances, including: The interference intensity level of adjacent virtual instance node pairs is determined based on the load interference characteristic data of adjacent virtual instances; Load interference characteristic data refers to data extracted by calculating the load similarity index between each pair of adjacent virtual instance nodes and based on a preset similarity threshold, indicating the existence of mutual interference loads between virtual instances. Adjacent virtual instance node pairs refer to two virtual instances connected to each other in the resource adjacency topology graph by identifiers of shared physical resources. Interference intensity level is a grading standard used to describe the degree of mutual interference between adjacent virtual instance node pairs due to shared physical resources; the interference intensity level is divided into multiple levels, such as mild interference level, moderate interference level, and severe interference level; based on the load similarity index of adjacent virtual instance node pairs and a preset similarity threshold... Determine the interference intensity level; when the difference between the load similarity index of adjacent virtual instance node pairs and the preset similarity threshold is between 0% and 10%, the interference intensity level is determined as mild interference; when the difference between the load similarity index of adjacent virtual instance node pairs and the preset similarity threshold is between 10% and 30%, the interference intensity level is determined as moderate interference; when the difference between the load similarity index of adjacent virtual instance node pairs and the preset similarity threshold exceeds 30%, the interference intensity level is determined as severe interference.

[0032] The abnormal fluctuation level of the target virtual instance is determined based on the abnormal fluctuation identification data of the target virtual instance; The abnormal fluctuation identification data includes data on abnormal and sudden changes in disk input / output operation rate, disk latency, and processor utilization during the operation of the target virtual instance. The method for determining the level of abnormal fluctuations is as follows: Statistical analysis is performed on the abnormal fluctuation identification data of the target virtual instance. The abnormal fluctuation identification data includes the number of sudden changes in the target virtual instance's operating indicators such as processor utilization, disk input / output operation rate, and disk latency within a continuous monitoring time window, as well as the magnitude value corresponding to each sudden change. The abnormal fluctuation frequency is defined as the proportion of the number of times a target virtual instance exhibits an abnormal sudden trend within a certain continuous monitoring time window to the total number of monitoring points within the monitoring time window. The abnormal fluctuation amplitude is defined as the ratio of the change in the operating indicator when an abnormal sudden trend occurs to the change in the average value of the corresponding indicator during normal operation. The classification method for abnormal fluctuation levels is as follows: When the abnormal fluctuation frequency value does not exceed 5% of the total number of monitoring points within the continuous monitoring time window, and the ratio of each abnormal fluctuation amplitude value to the normal operating average value of the corresponding indicator does not exceed 10%, the abnormal fluctuation level of the target virtual instance within this monitoring window is determined as low-level abnormal fluctuation. When the abnormal fluctuation frequency value exceeds 5% but does not exceed 15% of the total number of monitoring points within the continuous monitoring time window, and the abnormal fluctuation amplitude value of each time exceeds 10% but does not exceed 30% of the average value of the corresponding indicator under normal operation, the abnormal fluctuation level of the target virtual instance within this monitoring window is determined to be medium-level abnormal fluctuation. When the abnormal fluctuation frequency exceeds 15% of the total number of monitoring points within the continuous monitoring time window, and the abnormal fluctuation amplitude of each instance exceeds 30% of the average normal operating value of the corresponding indicator, the abnormal fluctuation level of the target virtual instance within this monitoring window is determined to be a high-level abnormal fluctuation.

[0033] Based on the interference intensity level and abnormal fluctuation level, a model is established to distinguish between the actual load of the target virtual instance and the interference load of adjacent virtual instances. Real load refers to load changes directly caused by the target virtual instance's own computational tasks or internal operations. Interference load refers to load interference caused by the target virtual instance sharing the same physical resources with adjacent virtual instances. The distinction model is a rule system for judging and distinguishing the sources of load changes. The establishment process is as follows: Based on the interference intensity level of adjacent virtual instance node pairs, different levels of interference are assigned different impact weights; based on the abnormal fluctuation level of the target virtual instance, different levels of abnormal fluctuation are assigned sensitivity weight values; distinction rules are established based on the two types of weight values. For example: when the abnormal fluctuation is high-level and the interference level is severe, the abnormal fluctuation of the target virtual instance is mainly caused by interference from adjacent instances; if the abnormal fluctuation is high-level and the interference level is not severe, or if the abnormal fluctuation is low-level and the interference level is mild, it is mainly caused by real load; if the abnormal fluctuation is medium-level and the interference level is moderate, real load and interference load coexist. The above rule system is the distinction model.

[0034] Based on the differentiation model, each abnormal fluctuation feature in the abnormal fluctuation identification data of the target virtual instance is classified and determined to determine whether each abnormal fluctuation feature belongs to the real load change of the target virtual instance or the interference load change of adjacent virtual instances. Each abnormal fluctuation feature in the abnormal fluctuation identification data of the target virtual instance is extracted, and the interference intensity level corresponding to the occurrence time of each abnormal fluctuation feature is determined. A discrimination model rule is applied for judgment: if the time corresponding to the abnormal fluctuation feature is at a severe or moderate interference level, the abnormal fluctuation feature is determined to be caused by interference load; if the time corresponding to the abnormal fluctuation feature is at a mild interference level, it is determined to be caused by actual load changes. For example, if a target virtual instance in an enterprise's second cloud computing service provider environment records ten abnormal fluctuation features within a monitoring time window, by comparing the interference intensity levels of adjacent instance node pairs, six of the abnormal fluctuation features correspond to moderate or severe interference levels, and are therefore determined to be interference load; the remaining four abnormal fluctuation features have a low interference level, and are therefore determined to be actual load. Through the above judgment method, the abnormal fluctuation identification data of the target virtual instance is classified.

[0035] S6: Based on the distinction between the actual load of the target virtual instance and the interference load of adjacent virtual instances, adjust the anomaly monitoring and judgment logic of multiple cloud computing resources, and output the monitoring status label data after interference correction, including: When the load fluctuation category of the target virtual instance is determined to be the load change of the adjacent virtual instance, the abnormal fluctuation identification data of the target virtual instance is marked and corrected to a non-abnormal monitoring status label in the standardized multi-cloud monitoring dataset. The status label of the operational indicator data corresponding to the abnormal fluctuation characteristics recorded in the abnormal fluctuation identification data of the target virtual instance is changed from the initial state to a non-abnormal monitoring status label. The initial state is the default labeling state of the abnormal fluctuation data by the monitoring system. The non-abnormal monitoring status label is the labeling state that the abnormal fluctuation characteristics are confirmed by the judgment result of the discrimination model to be caused by the load interference generated by adjacent virtual instances. For example, in a multi-cloud environment, a target virtual instance records five abnormal changes in processor utilization within a continuous monitoring time window. After the discrimination model determines that the five abnormal changes correspond to the severe or moderate load interference intensity levels of adjacent virtual instances, that is, all five abnormal changes are caused by the load interference of adjacent virtual instances. Therefore, changing the status label of the processor utilization data corresponding to the five abnormal changes from the default initial abnormal state to a non-abnormal monitoring status label indicates that the abnormal changes in the monitoring indicators of the target virtual instance come from the interference of adjacent virtual instances sharing physical resources, thus avoiding misjudgment of the operational status of the target virtual instance by the monitoring system and operation and maintenance personnel.

[0036] When the load fluctuation category of the target virtual instance is determined to be the actual load change of the target virtual instance, the abnormal fluctuation identification data of the target virtual instance is marked in the standardized multi-cloud monitoring dataset and identified as an abnormal monitoring status label. The status of the operational indicator data corresponding to the abnormal fluctuation characteristics recorded in the abnormal fluctuation identification data of the target virtual instance is changed from the initial state to an abnormal monitoring status label. The abnormal monitoring status label indicates that the abnormal fluctuation characteristics are real anomalies caused by the actual load changes of the target virtual instance itself. For example, in a multi-cloud environment, a target virtual instance records three abnormal changes in disk input / output operation rate within a continuous monitoring time window. After the differentiation model determines that the interference intensity level of the adjacent virtual instances corresponding to the three abnormal changes is all at the mild interference level, which cannot cause interference to the target virtual instance. Therefore, the three abnormal changes are confirmed to be caused by the actual load changes of the target virtual instance itself. Changing the status of the three abnormal changes recorded in the disk input / output operation rate data of the target virtual instance from the default initial state to an abnormal monitoring status label reflects the true source of the abnormal fluctuation characteristics, which facilitates accurate judgment of the operating status of the target virtual instance and timely monitoring, early warning and resource scheduling measures.

[0037] The standardized multi-cloud monitoring dataset is updated based on non-abnormal monitoring status labels and abnormal monitoring status labels, and the corrected target virtual instance monitoring status label data is output. Operational metric data tagged with non-abnormal monitoring status are uniformly corrected to the non-abnormal monitoring status category in the standardized multi-cloud monitoring dataset. This category indicates that the target virtual instance's monitoring data exhibits an abnormal abrupt change trend, caused by interference from adjacent virtual instances sharing physical resources; therefore, no emergency resource scheduling or load adjustment strategies are required. Operational metric data tagged with abnormal monitoring status are uniformly corrected to the abnormal monitoring status category in the standardized multi-cloud monitoring dataset. This category indicates that the abnormal changes in the target virtual instance's monitoring data are caused by its own actual load changes, potentially requiring resource reallocation or early warning intervention. Through this classification correction, the abnormal metric data of the target virtual instances in the standardized multi-cloud monitoring dataset are accurately assigned to their respective monitoring status categories. This process generates corrected monitoring status label data that accurately reflects the actual operating status of the target virtual instance. For example, for a target virtual instance deployed by an enterprise within a second cloud service provider environment, if ten abnormal trends in operating indicators are recorded within a continuous monitoring time window, after classification by a differentiation model, six of these trends are classified as non-abnormal monitoring states, while the remaining four are classified as abnormal monitoring states. After updating the above classification results in the standardized multi-cloud monitoring dataset, corrected monitoring status label data for the target virtual instance within this monitoring window is output. The monitoring status label data includes classified non-abnormal monitoring status data and abnormal monitoring status data, enabling monitoring managers to distinguish the actual operating status of the virtual instance and make accurate decisions and optimization adjustments.

[0038] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. 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. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. 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 sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0039] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0040] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0041] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0042] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0043] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0044] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0045] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0046] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring computing resources in a multi-cloud environment, characterized in that, Includes the following steps: S1: Collect operational metrics data and physical host resource distribution data of virtual computing resources deployed in multi-cloud environments, and perform time alignment and normalization processing to generate a standardized multi-cloud monitoring dataset. S2: Based on a standardized multi-cloud monitoring dataset, construct a resource adjacency topology graph between virtual instances; S3: Based on the resource adjacency topology graph, calculate the load similarity index between virtual instances and extract the load interference feature data of adjacent virtual instances; S4: Based on the operational metric change data extracted from the virtual instance in the multi-cloud monitoring dataset, identify the load fluctuation characteristics and sudden change trends of the target virtual instance, and generate abnormal fluctuation identification data of the target virtual instance. S5: Based on load interference characteristic data and abnormal fluctuation identification data, establish a model to distinguish between the real load of the target virtual instance and the interference load of adjacent virtual instances; S6: Based on the distinction between the actual load of the target virtual instance and the interference load of adjacent virtual instances, adjust the abnormal monitoring and judgment logic of multiple cloud computing resources, and output the monitoring status label data after interference correction.

2. The method for monitoring computing resources in a multi-cloud environment according to claim 1, characterized in that, S1, specifically: Collect operational metrics data of virtual computing resources deployed in multiple cloud service provider environments; Collect distribution data of physical host resources deployed in multiple different cloud service provider environments; Perform unified time base calibration on the operational metrics data of virtual computing resources and the distribution data of physical host resources; The calibrated virtual computing resource operation index data and physical host resource distribution data are normalized to generate a standardized multi-cloud monitoring dataset.

3. The method for monitoring computing resources in a multi-cloud environment according to claim 2, characterized in that, S2, specifically: Based on the standardized multi-cloud monitoring dataset, virtual instances are grouped and a mapping table between virtual instances and physical host resources is generated. Based on the mapping table, a unique node number is assigned to each physical host resource. All virtual instances mapped to the same physical host resource node number are marked as same-level nodes, and all virtual instances mapped to different physical host resource node numbers are marked as different-level nodes. Undirected edges are established between nodes at the same level based on shared processor core identifiers, shared memory channel identifiers, and shared input / output bus identifiers. No edges are established between nodes at different levels, generating a resource adjacency topology graph between virtual instances.

4. The method for monitoring computing resources in a multi-cloud environment according to claim 3, characterized in that, S3, specifically: Based on the resource adjacency topology graph, processor utilization data, memory utilization data, and network bandwidth utilization data of each node in each adjacent virtual instance are extracted from the standardized multi-cloud monitoring dataset. Calculate the similarity of processor utilization data, memory utilization data, and network bandwidth utilization data between node pairs of each adjacent virtual instance, and perform a weighted sum to obtain the load similarity index between node pairs of each adjacent virtual instance. Extract adjacent virtual instance node pairs whose load similarity index exceeds a preset similarity threshold, and generate load interference feature data of adjacent virtual instances.

5. The method for monitoring computing resources in a multi-cloud environment according to claim 4, characterized in that, S4, specifically: Based on a standardized multi-cloud monitoring dataset, the node identification information of the target virtual instance is determined; Based on the node identification information of the target virtual instance, extract the disk input / output operation rate data, disk latency data, and processor utilization data of the target virtual instance within the continuous monitoring time window from the standardized multi-cloud monitoring dataset; The rate of change of disk input / output operation rate data, disk latency data, and processor utilization data within a continuous monitoring time window is calculated, and the fluctuation characteristics of the corresponding rate of change are identified. For each fluctuation feature, abrupt trend detection is performed to generate abnormal fluctuation identification data for the target virtual instance.

6. The method for monitoring computing resources in a multi-cloud environment according to claim 5, characterized in that, S5, specifically: The interference intensity level of adjacent virtual instance node pairs is determined based on the load interference characteristic data of adjacent virtual instances; The abnormal fluctuation level of the target virtual instance is determined based on the abnormal fluctuation identification data of the target virtual instance; Based on the interference intensity level and abnormal fluctuation level, a model is established to distinguish between the actual load of the target virtual instance and the interference load of adjacent virtual instances. Based on the differentiation model, each abnormal fluctuation feature in the abnormal fluctuation identification data of the target virtual instance is classified and determined to determine whether each abnormal fluctuation feature belongs to the real load change of the target virtual instance or the interference load change of the adjacent virtual instance.

7. The method for monitoring computing resources in a multi-cloud environment according to claim 6, characterized in that, S6, specifically: When the load fluctuation category of the target virtual instance is determined to be the load change of the adjacent virtual instance, the abnormal fluctuation identification data of the target virtual instance is marked and corrected to a non-abnormal monitoring status label in the standardized multi-cloud monitoring dataset. When the load fluctuation category of the target virtual instance is determined to be the actual load change of the target virtual instance, the abnormal fluctuation identification data of the target virtual instance is marked in the standardized multi-cloud monitoring dataset and identified as an abnormal monitoring status label. The standardized multi-cloud monitoring dataset is updated based on non-abnormal monitoring status labels and abnormal monitoring status labels, and the corrected target virtual instance monitoring status label data is output.

Images