A cloud computing equipment operation and maintenance fault analysis and early warning method

Abstract

The present invention discloses a cloud computing equipment operation and maintenance fault analysis and early warning method, which relates to the technical field of big data analysis. The performance parameter data time series of the cloud server is obtained for preprocessing, the preprocessed performance parameter data time series is segmented according to different computing tasks, a performance evaluation model is constructed, the comprehensive load information value of the cloud server when performing different types of computing tasks is calculated, the time subsequences of the comprehensive load information value of the cloud server when performing different types of computing tasks are respectively obtained, the data values ​​of the comprehensive load information value are clustered and analyzed using the k-means clustering algorithm, the cloud server operation status classification results under different computing tasks are obtained, the operation status of the current cloud server when performing the computing task is evaluated according to the obtained operation status classification results, the operation status warning level threshold is set, and when the evaluation result exceeds the warning level threshold, an operation status alarm is issued.

Description

Technical Field

[0001] The present invention relates to the field of big data analysis technology, and in particular to a cloud computing equipment operation and maintenance fault analysis and early warning method. Background Art

[0002] Cloud computing is a type of distributed computing, which means breaking down huge data computing programs into countless small programs through the network "cloud", processing and analyzing these small programs through a system composed of multiple servers to obtain results and return them to users. Through this technology, tens of thousands of data can be processed in a very short time, thus achieving powerful network services. As the core component of cloud computing equipment, regular operation and maintenance detection and analysis of cloud servers is the basis for ensuring the implementation of cloud computing services.

[0003] The accumulation of abnormal data in cloud servers will lead to software aging. When cloud servers encounter problems such as insufficient computing resources, aging-related failures often occur, which will reduce the performance of cloud servers or increase the failure rate, and may even cause cloud servers to fail to operate normally. For the failure analysis caused by cloud server software aging, the commonly used method is to collect periodic data of multiple key indicators during the operation of the server, use data anomaly detection algorithms to accurately locate the time point when the relevant abnormal data appears, and further analyze the causes of the abnormal data. Among them, the key indicators collected mainly include computing resources such as CPU usage, memory usage, network bandwidth usage, and disk I / O usage; however, the existing anomaly analysis method does not take into account the difference in the contribution of different computing resources to the computing tasks of cloud servers. Therefore, when using indicator data to evaluate the aging status of the server, there is a certain deviation between the evaluation results and the actual situation, which leads to inaccurate positioning of the time point of abnormal data. To this end, we propose a cloud computing equipment operation and maintenance failure analysis and early warning method. Summary of the invention

[0004] The main purpose of the present invention is to provide a cloud computing equipment operation and maintenance fault analysis and early warning method, which can effectively solve the problems in the background technology.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] A cloud computing equipment operation and maintenance fault analysis and early warning method, comprising:

[0007] Obtain the performance parameter data time series F of the cloud server within the load operation period λ at sampling frequency f, and calculate the estimated value of the i-th performance parameter in the sequence F at time t , the calculation formula is:

[0008]

[0009] In the formula, is a constant coefficient, and ∈(0,1); It is represented as the sampling value of the i-th performance parameter at time t-1; It is represented as the sampling value of the i-th performance parameter at time tj; t∈λ;

[0010] The performance parameters of cloud servers include four categories: CPU indicators, memory indicators, network indicators, and I / O indicators. The CPU indicators include system CPU usage and user CPU usage; the memory indicators include system available memory and swap area available memory; the network indicators include upload peak rate and download peak rate; the I / O indicators include rates under different read and write modes;

[0011] The sampling frequency f is determined according to the load operation cycle λ, and the determination principle is:

[0012] If the load operation cycle λ≤1h, the sampling frequency f≤1min;

[0013] If the load operation cycle 1d≥λ>1h, then 1min<sampling frequency f≤5min;

[0014] If the load operation cycle is 30d>λ>1d, then 5min<sampling frequency f≤10min;

[0015] And the load operation cycle λ≤30d;

[0016] The prediction allowable error of the estimated value of the i-th performance parameter is set to , according to the formula: , calculate the confidence interval of the performance parameter of the i-th category , according to the obtained confidence interval, determine the sampling value of the i-th performance parameter at time t Whether it is an outlier, and remove the outliers in the performance parameter data time series, sampling value The principle for determining whether it is an outlier is:

[0017] like ∈ , then determine Not an outlier;

[0018] Otherwise, the sample value is determined is an outlier;

[0019] According to the different computing tasks of the cloud server within the load operation cycle λ, the normalized performance parameter data time series is segmented to obtain the u-th time subsequence F when the cloud server only performs computing-intensive tasks. 1u, the vth time subsequence F when only memory-intensive tasks are performed 2v , the w-th time subsequence F when performing both computation-intensive tasks and memory-intensive tasks 3w , the data normalization processing formula in time series data is:

[0020]

[0021] In the formula, is the normalized data of the i-th performance parameter at time t; are the minimum and maximum sampling values ​​of the i-th performance parameter within the load operation cycle λ respectively;

[0022] According to the formula: Calculate the comprehensive load information value C of the cloud server when performing different types of computing tasks q , where It is represented as the kth performance parameter sampling value of the cloud server when performing the qth type of computing task; is the weight of the kth performance parameter of the cloud server when performing the qth type of computing task, and ; q=1,2,3; where C1 is the comprehensive load information value when the cloud server only performs computing-intensive tasks; C2 is the comprehensive load information value when the cloud server only performs memory-intensive tasks; C3 is the comprehensive load information value when the cloud server performs both computing-intensive tasks and memory-intensive tasks. The performance parameter weights when the cloud server only performs computing-intensive tasks and performance parameter weights when the cloud server only performs memory-intensive tasks The determination principle is:

[0023] Obtain performance parameter data and processing time efficiency data of different types of computing tasks during processing;

[0024] The performance parameter data and the processing time data were subjected to correlation analysis, and the correlation coefficient between the processing time and the performance parameter was calculated. The calculation formula is:

[0025]

[0026] In the formula, It is expressed as the correlation coefficient between the processing time of the p-th computing task and the sampling value of the k-th performance parameter; It is represented as the kth performance parameter sampling value when the cloud server performs the pth computing task; It is represented as the sample mean of the kth performance parameter when the cloud server performs the pth computing task; It is represented as the processing time of the cloud server when performing the p-th computing task; It is represented as the average processing time of the cloud server when performing the p-th computing task;

[0027] The calculated result of the obtained correlation coefficient is normalized, and the normalized correlation coefficient is used as the weight value of the corresponding performance parameter.

[0028] Performance parameter weights when a cloud server performs both compute-intensive and memory-intensive tasks The determination principle is:

[0029] Get performance parameter weights and The calculation results of

[0030] Obtain the task amounts of computing-intensive tasks and memory-intensive tasks in a single computing task respectively;

[0031] According to the obtained task information and Perform weighted processing and calculate , the calculation formula is: , where They are the proportion of computing workload of computing-intensive tasks and memory-intensive tasks in a single computing task.

[0032] The time subseries of the comprehensive load information values ​​of the cloud server when performing different types of computing tasks are obtained respectively, and the data values ​​in the time subseries of the comprehensive load information values ​​are clustered and analyzed using the k-means clustering algorithm, so as to output K categories of the time series clustering results as the classification level of the cloud server load degree, and obtain the cloud server operation status classification results under different computing tasks. The process of obtaining the operation status classification results includes the following steps:

[0033] Step S1: Obtaining a time subsequence of comprehensive load information values ​​of a cloud server when performing different types of computing tasks;

[0034] Step S2: The elbow method is used to determine the clustering number K value of the time subsequence of each comprehensive load information value, wherein the judgment basis is the sum of square errors SSE within the cluster sample object class, and the calculation formula is:

[0035]

[0036] In the formula, is any category, p is a sample in any category, is the center point of the class;

[0037] Step S3: Randomly select the initial cluster center and remove the cluster center Other objects outside (1<n<K) are assigned to the class where the center point with the smallest distance to them is located;

[0038] Step S4: Update the center point of each cluster according to the principle of minimizing the average value of the Euclidean distance;

[0039] Step S5: repeating steps S3 and S4 until the algorithm converges, where convergence is defined as the number of clustering algorithm iterations not changing for multiple times;

[0040] Step S6: Output K categories of time series clustering results, use the output K categories as the classification levels of cloud server load levels, and obtain cloud server operation status classification results under different computing tasks;

[0041] According to the obtained operation status classification results, the operation status of the current cloud server when performing computing tasks is evaluated, and the operation status warning level threshold is set. When the evaluation result exceeds the warning level threshold, an operation status alarm is issued.

[0042] The present invention has the following beneficial effects:

[0043] Compared with the prior art, the performance parameter data time series of the cloud server within the load operation period λ is obtained at a sampling frequency f, the obtained data is preprocessed, and the preprocessed performance parameter data time series is segmented according to different computing tasks to obtain the u-th time subsequence F when the cloud server only performs computing-intensive tasks. 1u , the vth time subsequence F when only memory-intensive tasks are performed 2v , the w-th time subsequence F when performing both computation-intensive tasks and memory-intensive tasks 3w , build a performance evaluation model, calculate the comprehensive load information value of the cloud server when performing different types of computing tasks, obtain the time subseries of the comprehensive load information value of the cloud server when performing different types of computing tasks, use the k-means clustering algorithm to perform cluster analysis on the data values ​​in the time subseries of the comprehensive load information value, obtain the cloud server operation status classification results under different computing tasks, evaluate the operation status of the current cloud server when performing computing tasks according to the obtained operation status classification results, set the operation status warning level threshold, and issue an operation status alarm when the evaluation result exceeds the warning level threshold. By extracting and classifying the time subseries of performance parameter data under different computing tasks, the deviation between the server aging status evaluation result and the actual situation caused by the difference in the contribution of different computing resources to the computing tasks of the cloud server can be reduced, thereby improving the accuracy of abnormal data time point positioning, and accurately evaluating and warning the operation status of the cloud server when performing computing tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1The present invention is a flow chart of a cloud computing equipment operation and maintenance fault analysis and early warning method. DETAILED DESCRIPTION

[0045] The present invention will be further described below in conjunction with specific implementation methods, wherein the accompanying drawings are only used for exemplary descriptions and represent only schematic diagrams rather than actual drawings, and should not be understood as limiting the present invention. In order to better illustrate the specific implementation methods of the present invention, some parts of the accompanying drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product.

[0046] The implementation process of the technical solution of the present invention includes the following steps:

[0047] Step 1: Obtain the performance parameter data time series F of the cloud server within the load operation period λ at the sampling frequency f, and calculate the estimated value of the i-th performance parameter in the sequence F at time t , the calculation formula is:

[0048]

[0049] In the formula, is a constant coefficient, and ∈(0,1); It is represented as the sampling value of the i-th performance parameter at time t-1; It is represented as the sampling value of the i-th performance parameter at time tj; t∈λ;

[0050] Among them, the sampling frequency f is determined according to the load operation cycle λ, and the determination principle is:

[0051] If the load operation cycle λ≤1h, the sampling frequency f≤1min;

[0052] If the load operation cycle 1d≥λ>1h, then 1min<sampling frequency f≤5min;

[0053] If the load operation cycle is 30d>λ>1d, then 5min<sampling frequency f≤10min;

[0054] And the load operation cycle λ≤30d;

[0055] The performance parameters of cloud servers include CPU indicators, memory indicators, network indicators, and I / O indicators.

[0056] CPU indicators include system CPU usage and user CPU usage;

[0057] Memory indicators include system available memory and swap area available memory;

[0058] Network indicators include peak upload rate and peak download rate;

[0059] I / O metrics include rates in different read and write modes;

[0060] The data preprocessing process includes the following steps:

[0061] The prediction allowable error of the estimated value of the i-th performance parameter is set to , according to the formula: , calculate the confidence interval of the performance parameter of the i-th category , according to the obtained confidence interval, determine the sampling value of the i-th performance parameter at time t Whether it is an outlier, and remove the outliers in the performance parameter data time series, sampling value The principle for determining whether it is an outlier is:

[0062] like ∈ , then determine Not an outlier;

[0063] Otherwise, the sample value is determined is an outlier;

[0064] Normalize the data in the time series data, where the normalization formula is:

[0065]

[0066] In the formula, is the normalized data of the i-th performance parameter at time t; are the minimum and maximum sampling values ​​of the i-th performance parameter within the load operation cycle λ respectively;

[0067] In this embodiment, the performance parameters of the cloud server include CPU usage, system available memory, download peak rate, and read and write rate as an example for explanation;

[0068] Step 2: Segment the preprocessed performance parameter data time series according to different computing tasks, and obtain the u-th time subsequence F when the cloud server only performs computing-intensive tasks 1u , the vth time subsequence F when only memory-intensive tasks are performed 2v , the w-th time subsequence F when performing both computation-intensive tasks and memory-intensive tasks 3w ;

[0069] It should be noted that the collected cloud server performance parameter data time series F is a continuous data time series within the load operation cycle λ. During the operation of the cloud server, the computing tasks processed in each stage are different. The specific computing tasks can be generally classified into three categories, including computing-intensive tasks; memory-intensive tasks; and computing-intensive tasks and memory-intensive tasks at the same time. Different computing tasks have different requirements for performance parameters. For example, for computing-intensive tasks, the requirements for CPU indicators are more stringent, while for memory-intensive tasks, the requirements for memory indicators are more stringent.

[0070] Step 3: According to the formula: Calculate the comprehensive load information value C of the cloud server when performing different types of computing tasks q , where It is represented as the kth performance parameter sampling value of the cloud server when performing the qth type of computing task; is the weight of the kth performance parameter of the cloud server when performing the qth type of computing task, and ; q=1,2,3; where C1 is the comprehensive load information value when the cloud server only performs computing-intensive tasks; C2 is the comprehensive load information value when the cloud server only performs memory-intensive tasks; C3 is the comprehensive load information value when the cloud server performs both computing-intensive tasks and memory-intensive tasks;

[0071] Comprehensive load information value;

[0072] The performance parameter data time series F is divided into three types of time subsequences by computing tasks, corresponding to different computing tasks in the operation of the cloud server. When the cloud server only performs computing-intensive tasks or only performs memory-intensive tasks, due to the different requirements of different performance indicators, the corresponding performance indicators have different contributions to the load state of the cloud server, resulting in differences in the time required to process tasks. Based on this, the performance parameter weights are and The determination principle is:

[0073] Obtain performance parameter data and processing time efficiency data of different types of computing tasks during processing;

[0074] The performance parameter data and the processing time data were subjected to correlation analysis, and the correlation coefficient between the processing time and the performance parameter was calculated. The calculation formula is:

[0075]

[0076] In the formula, It is expressed as the correlation coefficient between the processing time of the p-th computing task and the sampling value of the k-th performance parameter; It is represented as the kth performance parameter sampling value when the cloud server performs the pth computing task; It is represented as the sample mean of the kth performance parameter when the cloud server performs the pth computing task; It is represented as the processing time of the cloud server when performing the p-th computing task; It is represented as the average processing time of the cloud server when performing the p-th computing task;

[0077] The calculation results of the obtained correlation coefficients are normalized, and the normalized correlation coefficients are used as the weight values ​​of the corresponding performance parameters. That is, when processing the corresponding computing tasks, the performance parameters with higher correlation with the processing time efficiency have higher contribution in processing such tasks, and the corresponding weight values ​​are larger.

[0078] Similarly, when a cloud server performs both computing-intensive tasks and memory-intensive tasks, it can be regarded as a superposition of the processing flows of the two tasks. Therefore, the weight value in this state can be comprehensively evaluated based on the proportion of the task volume. The specific determination principle is:

[0079] Get performance parameter weights and The calculation results of

[0080] Obtain the task amounts of computing-intensive tasks and memory-intensive tasks in a single computing task respectively;

[0081] According to the obtained task information and Perform weighted processing and calculate , the calculation formula is: , where They are the proportion of computing workload of computing-intensive tasks and memory-intensive tasks in a single computing task.

[0082] Step 4: Obtain the time subsequences of the comprehensive load information values ​​of the cloud server when performing different types of computing tasks, and use the k-means clustering algorithm to perform cluster analysis on the data values ​​in the time subsequences of the comprehensive load information values ​​to obtain the classification results of the cloud server operating status under different computing tasks. The acquisition process includes the following steps:

[0083] Step S1: Obtaining a time subsequence of comprehensive load information values ​​of a cloud server when performing different types of computing tasks;

[0084] Step S2: The elbow method is used to determine the clustering number K value of the time subsequence of each comprehensive load information value, wherein the judgment basis is the sum of square errors SSE within the cluster sample object class, and the calculation formula is:

[0085]

[0086] In the formula, is any category, p is a sample in any category, is the center point of the class;

[0087] Step S3: Randomly select the initial cluster center and remove the cluster center Other objects outside (1<n<K) are assigned to the class where the center point with the smallest distance to them is located;

[0088] Step S4: Update the center point of each cluster according to the principle of minimizing the average value of the Euclidean distance;

[0089] Step S5: repeating steps S3 and S4 until the algorithm converges, where convergence is defined as the number of clustering algorithm iterations not changing for multiple times;

[0090] Step S6: Output K categories of time series clustering results, use the output K categories as the classification levels of cloud server load levels, and obtain cloud server operation status classification results under different computing tasks;

[0091] It should be noted that the K output categories correspond to the operating status of the cloud server when processing different computing tasks. When only computing-intensive tasks are processed, if the number of output categories K is 4, the corresponding time subsequence F 1u The comprehensive load information value C1 is divided into four states, which can be defined as light load state, medium load state, heavy load state, and overload state according to the load degree. The same is true when other types of computing tasks are processed, which will not be repeated here;

[0092] Step 5: Evaluate the running status of the current cloud server when performing computing tasks based on the obtained running status classification results, set the running status warning level threshold, and issue a running status alarm when the evaluation result exceeds the warning level threshold. The specific process is as follows:

[0093] Get the type of computing tasks and current performance parameters of the current cloud server;

[0094] Determine the comprehensive load information value calculation formula according to the computing task type, and bring the current performance parameters into the formula to obtain the comprehensive load information value of the current cloud server;

[0095] According to the classification result of the operation status of the cloud server when processing the computing task of this type, the obtained comprehensive load information value is classified to determine the current operation status of the cloud server;

[0096] Among them, the light load state is the stage where the cloud server gradually increases its load after the initial stage, and the utilization rate of various resources of the cloud server gradually increases in this stage; the medium load state is the most stable state of the cloud server, characterized by the utilization rate of various resources tending to be stable and not changing significantly; in the heavy load state, the cloud server is under a large load, and the characteristics of the time series data are manifested as the utilization rate of various resources breaking the stable state; the overload state is the state where the cloud server is under excessive load. At this time, the cloud server is prone to abnormal errors, and even downtime and unable to provide normal services. Various performance indicators are in the state of highest utilization. In this state, the server generally runs slowly and responds unsatisfactorily, and it is appropriate to restart the server.

[0097] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A cloud computing equipment operation and maintenance fault analysis and early warning method, characterized in that: include: Obtain the performance parameter data time series F of the cloud server within the load operation period λ at sampling frequency f, and calculate the estimated value of the i-th performance parameter in the sequence F at time t , the calculation formula is: ； In the formula, is a constant coefficient, and ∈(0,1); It is represented as the sampling value of the i-th performance parameter at time t-1; It is represented by the sampling value of the i-th performance parameter at time tj; t∈λ; The prediction allowable error of the estimated value of the i-th performance parameter is set to , according to the formula: , calculate the confidence interval of the performance parameter of the i-th category , according to the obtained confidence interval, determine the sampling value of the i-th performance parameter at time t Whether it is an outlier, and remove the outliers in the performance parameter data time series; According to the different computing tasks of the cloud server within the load operation cycle λ, the normalized performance parameter data time series is segmented to obtain the u-th time subsequence F when the cloud server only performs computing-intensive tasks. 1u , the vth time subsequence F when only memory-intensive tasks are performed 2v , the w-th time subsequence F when performing both computation-intensive tasks and memory-intensive tasks 3w ; According to the formula: Calculate the comprehensive load information value C of the cloud server when performing different types of computing tasks q , where It is represented as the kth performance parameter sampling value of the cloud server when performing the qth type of computing task; is the weight of the kth performance parameter of the cloud server when performing the qth type of computing task, and ; q=1,2,3; where C1 is the comprehensive load information value when the cloud server only performs computing-intensive tasks; C2 is the comprehensive load information value when the cloud server only performs memory-intensive tasks; C3 is the comprehensive load information value when the cloud server performs both computing-intensive tasks and memory-intensive tasks; The time subseries of the comprehensive load information values ​​of the cloud server when performing different types of computing tasks are obtained respectively, and the data values ​​in the time subseries of the comprehensive load information values ​​are clustered using the k-means clustering algorithm, and the K categories of the time series clustering results are output as the classification levels of the cloud server load degree, and the classification results of the cloud server operation status under different computing tasks are obtained; According to the obtained operation status classification results, the operation status of the current cloud server when performing computing tasks is evaluated, and the operation status warning level threshold is set. When the evaluation result exceeds the warning level threshold, an operation status alarm is issued.

2. A cloud computing equipment operation and maintenance fault analysis and early warning method according to claim 1, characterized in that: Sample value The principle for determining whether it is an outlier is: like ∈ , then determine Not an outlier; Otherwise, the sample value is determined For outliers.

3. A cloud computing equipment operation and maintenance fault analysis and early warning method according to claim 1, characterized in that: The data normalization formula in time series data is: ； In the formula, is the normalized data of the i-th performance parameter at time t; are respectively the minimum and maximum sampling values ​​of the i-th performance parameter within the load operation cycle λ.

4. A cloud computing equipment operation and maintenance fault analysis and early warning method according to claim 1, characterized in that: Performance parameter weights when the cloud server only performs computationally intensive tasks and performance parameter weights when the cloud server only performs memory-intensive tasks The determination principle is: Obtain performance parameter data and processing time efficiency data of different types of computing tasks during processing; The performance parameter data and the processing time data were subjected to correlation analysis, and the correlation coefficient between the processing time and the performance parameter was calculated. The calculation formula is: ； In the formula, It is expressed as the correlation coefficient between the processing time of the p-th computing task and the sampling value of the k-th performance parameter; It is represented as the kth performance parameter sampling value when the cloud server performs the pth computing task; It is represented as the sample mean of the kth performance parameter when the cloud server performs the pth computing task; It is represented as the processing time of the cloud server when performing the p-th computing task; It is represented as the average processing time of the cloud server when performing the p-th computing task; The calculated result of the obtained correlation coefficient is normalized, and the normalized correlation coefficient is used as the weight value of the corresponding performance parameter.

5. A cloud computing equipment operation and maintenance fault analysis and early warning method according to claim 1, characterized in that: Performance parameter weights when a cloud server performs both compute-intensive and memory-intensive tasks The determination principle is: Get performance parameter weights and The calculation results of Obtain the task amounts of computing-intensive tasks and memory-intensive tasks in a single computing task respectively; According to the obtained task information and Perform weighted processing and calculate , the calculation formula is: , where They are the proportion of computing workload of computation-intensive tasks and memory-intensive tasks in a single computing task.

6. A cloud computing equipment operation and maintenance fault analysis and early warning method according to claim 1, characterized in that: The performance parameters of cloud servers include CPU indicators, memory indicators, network indicators, and I / O indicators. The CPU indicators include system CPU usage and user CPU usage; The memory indicators include system available memory and swap area available memory; The network indicators include upload peak rate and download peak rate; The I / O indicators include rates in different read and write modes.

7. A cloud computing equipment operation and maintenance fault analysis and early warning method according to claim 1, characterized in that: The sampling frequency f is determined according to the load operation cycle λ, and the determination principle is: If the load operation cycle λ≤1h, the sampling frequency f≤1min; If the load operation cycle 1d≥λ>1h, then 1min<sampling frequency f≤5min; If the load operation cycle is 30d>λ>1d, then 5min<sampling frequency f≤10min; And the load operation cycle λ≤30d.

8. A cloud computing equipment operation and maintenance fault analysis and early warning method according to claim 1, characterized in that: The process of obtaining the operation status classification results includes the following steps: Step S1: Obtaining a time subsequence of comprehensive load information values ​​of a cloud server when performing different types of computing tasks; Step S2: The elbow method is used to determine the clustering number K value of the time subsequence of each comprehensive load information value, wherein the judgment basis is the sum of square errors SSE within the cluster sample object class, and the calculation formula is: ； In the formula, is any category, p is a sample in any category, is the center point of the class; Step S3: Randomly select the initial cluster center and remove the cluster center , 1<n<K, other objects are assigned to the class where the center point with the smallest distance to them is located; Step S4: Update the center point of each cluster according to the principle of minimizing the average value of the Euclidean distance; Step S5: repeating steps S3 and S4 until the algorithm converges, where convergence is defined as the number of clustering algorithm iterations not changing for multiple times; Step S6: Output the time series clustering results of K categories, use the output K categories as the classification level of the cloud server load degree, and obtain the classification results of the cloud server operation status under different computing tasks.

Images