A network storage server efficiency evaluation method and system

Abstract

The present invention relates to the technical field of performance evaluation, and specifically to a method and system for evaluating the efficiency of a network storage server, comprising the following steps: collecting continuous input and output performance data from the network storage server, applying a time window to the collected data, filtering data within a target time period, and digitally encoding the data to obtain preliminary encoded data. In the present invention, comprehensive monitoring of the operation of the network storage server is achieved through the precise collection and analysis of continuous performance data, thereby enabling detailed identification and prediction of system performance issues. By converting the data into frequency domain information through the application of discrete Fourier transforms, the key frequency components of hard disk performance and system response time are analyzed in detail. A support vector machine model is used to further correlate frequency characteristics with performance changes, optimizing the prediction accuracy of performance trends, making performance management more proactive and responding to potential risks more rapidly, thereby significantly improving the efficiency and stability of data center operations.

Description

Technical Field

[0001] The present invention relates to the technical field of performance evaluation, and in particular to a method and system for evaluating the efficiency of a network storage server. Background Art

[0002] Performance evaluation is a key technical area in computer science that focuses on measuring and analyzing the behavior and output of a system or component under specific conditions. Key tasks in this area include determining a system's response time when handling a large number of requests, throughput, and stability and scalability under various load conditions. Performance evaluation can help engineers and system administrators identify bottlenecks, optimize resource allocation, and improve system architecture to enhance overall efficiency and user satisfaction. By comprehensively analyzing software and hardware configurations, operating systems, and applications, performance evaluation provides valuable information about system design and operational efficiency, playing a key role in product development and maintenance.

[0003] The network storage server efficiency assessment method evaluates the performance of network storage servers by analyzing and measuring their data processing and responsiveness in actual operating environments. This method is often used to determine how server configuration affects data access speed and system stability, helping administrators and technical teams optimize server performance and enhance data center operational efficiency. Through systematic performance assessments, enterprises can ensure that their data storage solutions meet current business needs while also addressing the challenges of future data growth.

[0004] Traditional performance evaluations typically focus on short-term or single-condition system performance, such as response time and throughput, and fail to fully reflect system performance in long-term or complex environments. The lack of in-depth analysis of continuously changing data limits comprehensive system performance evaluations, making it difficult to effectively predict and respond to performance fluctuations, especially when handling high loads or complex tasks. Furthermore, commonly used evaluation methods are inadequate in dynamically adapting to emerging business needs, limiting the ability of data storage solutions to adapt to future data growth and potentially leading to long-term operational efficiency and stability issues. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a network storage server efficiency evaluation method and system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a method for evaluating the efficiency of a network storage server, comprising the following steps:

[0007] S1: Collect continuous input and output performance data from the network storage server, apply a time window to the collected data, filter the data within the target time period, perform digital encoding, and obtain preliminary encoded data;

[0008] S2: Based on the preliminary encoded data, signal processing is performed, a window function is applied to control the signal boundary, a discrete Fourier transform is performed, the time series is converted into frequency domain data in complex form including hard disk performance and system response time information, key frequency components are identified, and a frequency domain conversion result is generated;

[0009] S3: Using the frequency domain conversion result, perform spectrum estimation, calculate the average of the square of the amplitude of each frequency point to estimate the power spectrum density, determine the main energy distribution frequency in the data, detect the cache management efficiency, and output the power spectrum analysis result;

[0010] S4: Based on the power spectrum analysis results, identify peaks in the spectrum by comparing frequencies and amplitudes, analyze peak sizes, and select key frequency peaks that characterize server load and data transmission speed to obtain key frequency features;

[0011] S5: Based on the key frequency characteristics, configure the support vector machine model parameters, train the support vector machine model to learn the relationship between the frequency characteristics and performance changes, and verify the model accuracy to generate performance trend prediction results.

[0012] The preliminary coding data includes time period data and digital coding. The frequency domain conversion results include hard disk performance frequency domain, system response time frequency domain, and key frequencies. The power spectrum analysis results include power spectrum density, energy distribution frequency, and cache management efficiency. The key frequency characteristics include server load frequency and data transmission speed frequency. The performance trend prediction results include model parameters, frequency-performance relationship, and model accuracy.

[0013] As a further solution of the present invention, the step of obtaining the preliminary coded data is specifically as follows:

[0014] S111: extracting continuous input and output performance data from the network storage server, filtering data segments that fall within a specified time window, and generating a filtered performance data set;

[0015] S112: applying data formatting technology to the screened performance data set to convert it into a standard format suitable for encoding processing to obtain a standardized data set;

[0016] S113: Encoding the standardized data set using the encoding formula:

[0017]

[0018] Calculate the coding efficiency of multiple data points and generate preliminary coding data C;

[0019] Among them, w irepresents the weight parameter of the i-th data point, which is used to adjust the relative criticality of the data point in the encoding process, g i is the gain coefficient, adjusting the data point deviation p i -q i The impact strength, p i represents the value of the actual data point, q i The preset baseline value for the data points.

[0020] As a further solution of the present invention, the step of obtaining the frequency domain conversion result is specifically:

[0021] S211: Based on the preliminary coded data, adjusting the signal boundary by using a window function to reduce signal leakage during the conversion process, and generating signal data with adjusted boundary;

[0022] S212: Performing discrete Fourier transform on the signal data of the adjusted boundary to convert the time domain signal into a frequency domain signal to obtain frequency domain data;

[0023] S213: Analyze the frequency domain data, identify key frequency components, and calculate the energy contribution of the key frequency components using the formula:

[0024]

[0025] Quantify hard drive performance and system response time at multiple frequency points and generate frequency domain conversion results;

[0026] Where F(f) represents the power density at frequency f, X(k) is the complex spectrum value at the kth frequency point, N is the total number of data points, and f is the target frequency.

[0027] As a further solution of the present invention, the steps of obtaining the power spectrum analysis results are specifically as follows:

[0028] S311: Using the frequency domain conversion result, calculate the square of the complex amplitude of each frequency point, measure the energy contribution of multiple frequency points in the total signal, and generate frequency point energy data;

[0029] S312: Calculate the average power spectrum density of each frequency point from the frequency point energy data, and obtain average power spectrum data by averaging multiple frequency point data;

[0030] S313: Analyze the average power spectrum data, identify key energy distribution frequencies, and detect energy peaks using the formula:

[0031]

[0032] Calculate the weighted power spectrum density of the frequency point f and generate the power spectrum analysis result;

[0033] Where E(f) represents the weighted power spectral density of frequency f, which is used to measure the energy distribution of the signal at the frequency. k | 2 is the energy of frequency point k, representing the square of the amplitude of the signal at the frequency point, used to quantify the energy contribution of multiple frequency points, N is the total number of data points, indicating the total number of sampling points in the signal, It is the weighting coefficient of frequency point k, which is adjusted by the cosine function to enhance the weight of the target frequency component.

[0034] As a further solution of the present invention, the step of obtaining the key frequency characteristics is specifically as follows:

[0035] S411: Based on the power spectrum analysis result, determine the energy distribution in the overall signal by performing frequency analysis on the power spectrum data, identify significant peaks in the power spectrum, and obtain a peak data list;

[0036] S412: Sort the peak data list by size and compare the frequencies to screen out the frequencies with significant energy contributions using the formula:

[0037]

[0038] By weighting the energy of each frequency and adjusting it using the square root of the frequency and amplitude, the key frequency data is obtained;

[0039] Among them, P key Represents the weighted energy of the key frequency, f i Represents the frequency value after screening, indicating that the change of high frequency has a greater impact on the total energy, reflecting the key role of high frequency components in power spectrum analysis. i The amplitude representing the frequency, by taking the square root of the absolute value, reduces the impact of outliers on the overall calculation and enhances the model's compatibility and sensitivity to amplitude changes;

[0040] S413: Based on the key frequency data, iteratively analyze the role of frequency in server load and data transmission speed, identify frequencies closely related to server performance, and generate key frequency features.

[0041] As a further solution of the present invention, the steps of obtaining the performance trend prediction results are specifically as follows:

[0042] S511: Configure support vector machine model parameters, adjust the support vector configuration to match input data based on the key frequency features, and generate model configuration parameters;

[0043] S512: Using the model configuration parameters, training the support vector machine model, refining the relationship between learning frequency characteristics and performance changes through iterative optimization, and obtaining a trained model;

[0044] S513: Execute accuracy verification of the trained model, use the verification data set to evaluate the consistency between the model prediction result and the actual data, and generate the model verification result by calculating the error rate;

[0045] S514: Based on the model verification results, the model prediction results are analyzed using the formula:

[0046]

[0047] Generate performance trend prediction results;

[0048] Among them, P trend Represents the performance trend prediction result, α i is the model weight, which adjusts the influence of each data point in the performance trend prediction, y i is the actual performance change data, which is the basic input for model prediction. β is the error adjustment coefficient, which is used to optimize the accuracy of the prediction model.

[0049] A network storage server efficiency evaluation system is provided, wherein the network storage server efficiency evaluation system is used to execute the above-mentioned network storage server efficiency evaluation method, and the system comprises:

[0050] The data acquisition module collects continuous input and output performance data from the network storage server, performs time window screening on the data, determines the data within the target time period, performs digital coding, and obtains preliminary coded data;

[0051] The signal analysis module controls the signal boundary based on the preliminary encoded data, performs a discrete Fourier transform, converts the time series data into frequency domain data in complex form, identifies the key frequency components of the hard disk performance and system response time, and generates a frequency domain conversion result;

[0052] The spectrum estimation module uses the frequency domain conversion results to average the squared amplitudes of the frequency points, estimate the power spectrum density, analyze the main energy distribution frequencies, detect the server cache management efficiency, and output the power spectrum analysis results;

[0053] The performance prediction module identifies and analyzes the peaks in the spectrum based on the power spectrum analysis results, extracts key frequency features that characterize server load and data transmission speed, configures and trains a support vector machine model, learns the performance change relationship, verifies the model accuracy, and obtains performance trend prediction results.

[0054] Compared with the prior art, the advantages and positive effects of the present invention are:

[0055] This invention achieves comprehensive monitoring of network storage server operations through the precise collection and analysis of continuous performance data, enabling detailed identification and prediction of system performance issues. By applying discrete Fourier transforms to frequency domain information, key frequency components of hard drive performance and system response time are analyzed in detail. A support vector machine model is used to further correlate frequency characteristics with performance variations, optimizing the accuracy of performance trend predictions. This enables more proactive performance management and faster response to potential risks, significantly improving the efficiency and stability of data center operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic diagram of the workflow of the present invention;

[0057] Figure 2 Flowchart of the steps for obtaining preliminary coded data of the present invention;

[0058] Figure 3 Flowchart of the steps for obtaining the frequency domain conversion result of the present invention;

[0059] Figure 4 Flowchart of the steps for obtaining the power spectrum analysis results of the present invention;

[0060] Figure 5 Flowchart of the steps for obtaining key frequency characteristics of the present invention;

[0061] Figure 6 This is a flow chart of the steps for obtaining the performance trend prediction results of the present invention. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0063] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0064] Example 1

[0065] See also Figure 1The present invention provides a technical solution: a method for evaluating the efficiency of a network storage server, comprising the following steps:

[0066] S1: Collect continuous input and output performance data from the network storage server, apply a time window to the collected data, filter the data within the target time period, perform digital encoding, and obtain preliminary encoded data;

[0067] S2: Based on the preliminary encoded data, signal processing is performed. A window function is applied to control the signal boundary and a discrete Fourier transform is performed to convert the time series into complex frequency domain data containing information about the hard drive performance and system response time. Key frequency components are identified to generate a frequency domain conversion result.

[0068] S3: Use the frequency domain conversion results to perform spectrum estimation. Calculate the average of the squared amplitude of each frequency point to estimate the power spectrum density, determine the dominant energy distribution frequency in the data, detect cache management efficiency, and generate power spectrum analysis results.

[0069] S4: Based on the power spectrum analysis results, identify the peaks in the spectrum by comparing the frequency and amplitude, analyze the peak size, and screen the key frequency peaks that represent the server load and data transmission speed to obtain the key frequency characteristics;

[0070] S5: Based on key frequency features, configure support vector machine model parameters, train the support vector machine model to learn the relationship between frequency features and performance changes, verify the model accuracy, and generate performance trend prediction results.

[0071] The preliminary coding data includes time period data and digital coding. The frequency domain conversion results include hard disk performance frequency domain, system response time frequency domain, and key frequencies. The power spectrum analysis results include power spectrum density, energy distribution frequency, and cache management efficiency. The key frequency characteristics include server load frequency and data transmission speed frequency. The performance trend prediction results include model parameters, frequency-performance relationship, and model accuracy.

[0072] See also Figure 2 ,The specific steps for obtaining preliminary coding data are:

[0073] S111: extracting continuous input and output performance data from the network storage server, filtering data segments that fall within a specified time window, and generating a filtered performance data set;

[0074] Continuous input and output performance data is extracted from the network storage server, and the data timestamp and data integrity are checked to ensure the continuity and accuracy of data collection. The validity of the data is confirmed by verifying the degree of overlap between the data timestamp and the preset time window, ensuring that the collected data segment fully meets the time window requirements, thereby obtaining a filtered performance data set.

[0075] S112: applying data formatting technology to the screened performance data set to convert it into a standard format suitable for encoding processing to obtain a standardized data set;

[0076] The filtered performance data set is converted into a standard format suitable for encoding processing using data formatting technology. This conversion process involves the conversion of data types and the reorganization of structures to facilitate the application of subsequent encoding algorithms. The conversion method includes the normalization of data values ​​and the standardization of timestamps. The purpose is to reduce abnormal errors in the data processing process, improve the efficiency and accuracy of data processing, and obtain a standardized data set.

[0077] S113: Encode the standardized data set using the encoding formula:

[0078]

[0079] Calculate the coding efficiency of multiple data points and generate preliminary coding data C;

[0080] Among them, w i represents the weight parameter of the i-th data point, which is used to adjust the relative criticality of the data point in the encoding process, g i is the gain coefficient, adjusting the data point deviation p i -q i The impact strength, p i represents the value of the actual data point, q i The preset baseline value for the data points.

[0081] formula:

[0082]

[0083] The benefit of the formula is that by adjusting the weights and gain coefficients between data points, the compression ratio and error rate of the encoding can be effectively controlled, which is crucial in data transmission and storage, especially in network communication and batch storage environments where compression is required.

[0084] Detailed explanation of the formula and the process of formula calculation and derivation:

[0085] Suppose there is a dataset where n=3, and w1=0.5, g1=3, p1=100, q1=95, w2=0.3, g2=4, p2=102, q2=100, w3=0.2, g3=5, p3=99, q3=97.

[0086] Compute the encoding contribution of the first data point:

[0087]

[0088] Calculate the encoding contribution of the second data point:

[0089]

[0090] Calculate the encoding contribution of the third data point:

[0091]

[0092] Total coding contribution:

[0093] 1.2925+0.2946+0.1986=1.7857

[0094] This means that for these three data points, the total coding efficiency is 1.7857.

[0095] The results show that after applying the coding method, a relatively high data compression rate can be achieved while maintaining a low error rate, which is very beneficial for data transmission and storage in practical applications.

[0096] See also Figure 3 , the steps to obtain the frequency domain conversion results are as follows:

[0097] S211: Based on the preliminary coded data, adjusting the signal boundary by using a window function to reduce signal leakage during the conversion process, and generating signal data with adjusted boundaries;

[0098] Based on the preliminary coded data, the window function is used to adjust the signal boundary, reduce the signal leakage caused by the edge effect in the conversion process, and ensure the integrity of the data and the transmission of the signal. In this way, the correct application of the window function is directly related to the effect of the subsequent discrete Fourier transform. Therefore, the selection and adjustment of the window function is crucial, which not only affects the frequency domain performance of the signal, but also affects the accuracy of the analysis.

[0099] S212: Performing discrete Fourier transform on the signal data with adjusted boundaries to convert the time domain signal into a frequency domain signal to obtain frequency domain data;

[0100] After adjusting the signal data at the boundary, a discrete Fourier transform is performed to convert the time domain signal into a frequency domain signal. The conversion process involves mathematical calculations. Accurate acquisition of frequency domain data is key to analyzing hard drive performance and system response time. Through the discrete Fourier transform, key frequency components in the frequency domain are identified, which has a direct impact on subsequent performance evaluation and system optimization.

[0101] S213: Analyze frequency domain data, identify key frequency components, and calculate the energy contribution of key frequency components using the formula:

[0102]

[0103] Quantify hard drive performance and system response time at multiple frequency points and generate frequency domain conversion results;

[0104] Where F(f) represents the power density at frequency f, X(k) is the complex spectrum value at the kth frequency point, N is the total number of data points, and f is the target frequency.

[0105] formula:

[0106]

[0107] The benefit of this formula is that it can measure the contribution of differentiated frequency components to the total signal through a weighted average method combined with a sine weighting function, helping to identify key frequency components in hard drive performance and system response time.

[0108] Detailed explanation of the formula and the process of formula calculation and derivation:

[0109] Set N = 100, f = 50 Hz, and X(k) represents the frequency domain complex value of the discrete-time signal at point k. Select k = 10, X(10) = 5 + 3i, and calculate |X(10)| 2 =34. Substituting into the formula we get:

[0110]

[0111] The results show that the energy at 50 Hz is zero, indicating that there is no hard drive activity or system response at this frequency point. This ties in with the results of this step and illustrates the fact that hard drive performance and system response time are not affected by frequency.

[0112] See also Figure 4 , the specific steps for obtaining the power spectrum analysis results are:

[0113] S311: Using the frequency domain conversion result, calculate the square of the complex amplitude of each frequency point, measure the energy contribution of multiple frequency points in the total signal, and generate frequency point energy data;

[0114] Using the frequency domain conversion results, the square of the complex amplitude of each frequency point is calculated. This step is actually a standard process in signal processing. First, the complex value of each frequency point is obtained through the actual monitored frequency domain conversion results. These values ​​are obtained by the discrete Fourier transform in the signal conversion process. Then, the square of the amplitude of each complex value is calculated. The calculation formula for the square of the amplitude is |X(f)| 2 =Re(X(f)) 2 +Im(X(f)) 2, where Re(X(f)) and Im(X(f)) represent the real and imaginary parts of the complex number, respectively. By calculation, the energy contribution of each frequency point in the signal is obtained, which is the basis for the next step of power spectrum density estimation. In addition, this process also assists in analyzing the energy distribution of multiple frequency points in the frequency domain.

[0115] S312: Calculate the average power spectrum density of each frequency point from the frequency point energy data, and obtain average power spectrum data by averaging multiple frequency point data;

[0116] Starting from the frequency energy data, the goal of this step is to calculate the overall power spectrum density. Power spectrum density is a key parameter for measuring the total energy of a signal in signal analysis. By averaging the energy data of each frequency point, a smooth power spectrum is obtained, which not only helps to better analyze the energy distribution of the signal, but also forms the basis for iterative analysis of signal characteristics. The calculation formula of the average power spectrum is: Where N is the total number of frequency points, |X k | 2 is the energy of the kth frequency point. Through this step, we can obtain an average power spectrum that reflects the overall energy distribution of the signal, providing key data support for energy distribution analysis.

[0117] S313: Analyze the average power spectrum data, identify the key energy distribution frequency, and detect the energy peak using the formula:

[0118]

[0119] Calculate the weighted power spectrum density of the frequency point f and generate the power spectrum analysis result;

[0120] Among them, E(f) represents the weighted power spectral density of frequency f, which is used to measure the energy distribution of the signal under the frequency, and X k | 2 is the energy of frequency point k, representing the square of the amplitude of the signal at the frequency point, used to quantify the energy contribution of multiple frequency points, N is the total number of data points, indicating the total number of sampling points in the signal, It is the weighting coefficient of frequency point k, which is adjusted by the cosine function to enhance the weight of the target frequency component.

[0121] formula:

[0122]

[0123] The benefit of this formula is that by associating the energy of each frequency point with the cosine value of its frequency position, the model's sensitivity to the frequency distribution characteristics is enhanced, so that the contribution of high-energy frequency points is effectively amplified, which is particularly useful for analyzing signals with periodic fluctuations.

[0124] Detailed explanation of the formula and the process of formula calculation and derivation:

[0125] Set the frequency energy to |X k | 2 =200, total number of frequency points N = 1024, frequency f = 512, first calculate the cosine part When k = 0, the cosine part is 1, so the weight coefficient is 2, and the weighted energy of each frequency point is 400, so the total weighted power spectrum density is:

[0126]

[0127] The results show that for the signal at frequency f = 512, its weighted power spectral density is 400. This value reflects that the energy distribution of the signal is particularly concentrated at this frequency. If this frequency corresponds to a key signal feature, including the natural frequency of mechanical vibration, this information is very critical for fault diagnosis applications.

[0128] See also Figure 5 , the steps for obtaining key frequency features are as follows:

[0129] S411: Based on the power spectrum analysis results, determine the energy distribution in the overall signal by performing frequency analysis on the power spectrum data, identify significant peaks in the power spectrum, and obtain a peak data list;

[0130] By performing frequency analysis on power spectrum data, data is first collected and preprocessed, including signal amplification, filtering, and Fourier transform to convert it to the frequency domain. This process ensures data accuracy and operability. Through a series of operations, frequency domain data can be obtained to perform power spectrum analysis. This analysis includes determining the energy distribution in the overall signal and identifying significant peaks in the power spectrum. This step is the basis of the entire signal processing process. By identifying significant peaks, we can initially understand the main frequency points of the signal. These peak data will be recorded and used for the next step of data analysis and processing.

[0131] S412: Sort the peak data list by size and compare the frequencies to screen out the frequencies with significant energy contribution using the formula:

[0132]

[0133] By weighting the energy of each frequency and adjusting it using the square root of the frequency and amplitude, the key frequency data is obtained;

[0134] Among them, P key Represents the weighted energy of the key frequency, f i Represents the frequency value after screening, indicating that the change of high frequency has a greater impact on the total energy, reflecting the key role of high frequency components in power spectrum analysis. iThe amplitude representing the frequency, by taking the square root of the absolute value, reduces the impact of outliers on the overall calculation and enhances the model's compatibility and sensitivity to amplitude changes;

[0135] formula:

[0136]

[0137] The benefit of the formula is that the combined use of square and root signs not only improves the response sensitivity to the peak frequency, but also reduces the impact of outliers on the overall results through the square root. This design makes the formula more robust when processing data with extreme amplitudes.

[0138] Detailed explanation of the formula and the process of formula calculation and derivation:

[0139] Set the frequency value f detected from the dataset i For 100Hz, 200Hz, 300Hz, the corresponding amplitude is A i 30, 50, 20. First, calculate the square of each frequency to get 10000, 40000, 90000, the square root of each amplitude is 5.48, 7.07, 4.47, and then calculate the numerator:

[0140] 10000×5.48+40000×7.07+90000×4.47=1029190

[0141] The denominator is:

[0142] 30+50+20=100

[0143] get:

[0144]

[0145] The results show that the calculated weighted energy of the critical frequency is 10291.9, which indicates that the 300Hz frequency has the highest energy contribution in the dataset and is crucial for analyzing the performance impact of server load and data transmission speed.

[0146] S413: Based on the key frequency data, iteratively analyze the role of frequency in server load and data transmission speed, identify frequencies closely related to server performance, and generate key frequency features.

[0147] Based on the key frequency data, the data is analyzed, focusing on the role of key frequencies in server load and data transmission speed. By comprehensively utilizing the identified frequency data, it can be determined which frequencies are closely related to server performance. This analysis relies on the relationship between frequency and server operation. High frequency is associated with changes in data transmission speed, while low frequency is associated with stable server load. Through this process, the key frequency characteristics that affect server performance are identified. These characteristics are the key basis for optimizing server operating efficiency and data processing capabilities, thereby providing data support for server configuration and network optimization.

[0148] See also Figure 6 ,The specific steps for obtaining the performance trend prediction results are:

[0149] S511: Configure support vector machine model parameters, adjust the support vector configuration to match input data based on key frequency features, and generate model configuration parameters;

[0150] Configure the support vector machine model parameters. Based on key frequency features, adjust the support vector configuration to match the input frequency data. This step involves setting and adjusting parameters, including the kernel function type, penalty coefficient, and slack variable. These parameters are set based on the distribution characteristics of the data and the results of previous statistical analysis to ensure that the model can effectively identify and learn the relationship between frequency and performance. The goal is to generate a model configuration parameter with strong matching and excellent generalization ability.

[0151] S512: Using the model configuration parameters, the support vector machine model is trained. Through iterative optimization, the relationship between the learning frequency characteristics and the performance change is refined to obtain a trained model.

[0152] Use model configuration parameters to train the support vector machine model. By inputting batch experimental data, the model weights are cyclically adjusted and the loss function in the algorithm is optimized to reduce prediction errors. Throughout the entire process, the adjustment of model parameters is based on gradient descent or other optimization algorithms to ensure that the model can capture the relationship between frequency characteristics and performance changes. Through this process, a well-trained model with strong predictive capabilities is obtained.

[0153] S513: Execute accuracy verification of the trained model, use the verification data set to evaluate the consistency between the model prediction results and the actual data, and generate the model verification results by calculating the error rate;

[0154] The key to performing accuracy verification on a trained model is to use an independent validation dataset that was not used in model training. By comparing the model's predictions with actual performance data, the model's accuracy and error are calculated. This process focuses on the model's responsiveness and generalization to new data. The evaluation results will directly reflect the model's reliability and effectiveness in actual applications. The generated model verification results include accuracy, recall, and F1 score, providing decision support for model deployment.

[0155] S514: Based on the model validation results, analyze the model prediction results using the formula:

[0156]

[0157] Generate performance trend forecast results;

[0158] Among them, P trend Represents the performance trend prediction result, α i is the model weight, which adjusts the influence of each data point in the performance trend prediction, y i is the actual performance change data, which is the basic input for model prediction. β is the error adjustment coefficient, which is used to optimize the accuracy of the prediction model.

[0159] formula:

[0160]

[0161] The benefit of the formula is that by introducing the error adjustment coefficient β, the prediction stability and accuracy of the model are enhanced when the data fluctuates greatly. This formula is particularly suitable for real-time data analysis with dynamic changes.

[0162] Detailed explanation of the formula and the process of formula calculation and derivation:

[0163] Set the parameter value to α i =0.5,y i =100, β=10, and referring to three data points, the performance trend results predicted by the model are calculated as follows: First, calculate each α i ×y i =50, the sum is 150, and the denominator is:

[0164]

[0165] The prediction result is:

[0166]

[0167] The results show that the performance trends predicted by the model are very close to the actual data, indicating that the model has high prediction accuracy and can effectively predict performance trends, which helps to identify potential performance problems or optimization opportunities in advance.

[0168] A network storage server efficiency evaluation system is provided, which is used to execute the above-mentioned network storage server efficiency evaluation method. The system includes:

[0169] The data acquisition module collects continuous input and output performance data from the network storage server, performs time window screening on the data, determines the data within the target time period, performs digital coding, and obtains preliminary coded data;

[0170] The signal analysis module controls the signal boundaries based on the preliminary encoded data and performs a discrete Fourier transform to convert the time series data into complex frequency domain data. It then identifies the key frequency components of the hard drive performance and system response time and generates the frequency domain conversion results.

[0171] The spectrum estimation module uses the frequency domain conversion results to average the squared amplitudes of the frequency points, estimate the power spectrum density, analyze the main energy distribution frequencies, detect the server cache management efficiency, and produce power spectrum analysis results;

[0172] Based on the power spectrum analysis results, the performance prediction module identifies and analyzes the peaks in the spectrum, extracts key frequency features that characterize server load and data transmission speed, configures and trains the support vector machine model, learns the performance change relationship, verifies the model accuracy, and obtains performance trend prediction results.

[0173] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for evaluating the efficiency of a network storage server, characterized in that: The following steps are involved: Collect continuous input and output performance data from the network storage server, apply a time window to the collected data, filter the data within the target time period, perform digital encoding, and obtain preliminary encoded data; Based on the preliminary encoded data, signal processing is performed, a window function is applied to control the signal boundary, a discrete Fourier transform is performed, the time series is converted into frequency domain data in complex form including hard disk performance and system response time information, key frequency components are identified, and a frequency domain conversion result is generated; Using the frequency domain conversion result, perform spectrum estimation, calculate the average of the square of the amplitude of each frequency point to estimate the power spectrum density, determine the main energy distribution frequency in the data, detect cache management efficiency, and output power spectrum analysis results; Based on the power spectrum analysis results, the peaks in the spectrum are identified by comparing the frequency and amplitude, the peak sizes are analyzed, and the key frequency peaks that characterize the server load and data transmission speed are screened to obtain the key frequency characteristics; Based on the key frequency characteristics, support vector machine model parameters are configured, the support vector machine model is trained to learn the relationship between frequency characteristics and performance changes, and the model accuracy is verified to generate performance trend prediction results.

2. The network storage server efficiency evaluation method according to claim 1, characterized in that: The preliminary coding data includes time period data and digital coding. The frequency domain conversion results include hard disk performance frequency domain, system response time frequency domain, and key frequencies. The power spectrum analysis results include power spectrum density, energy distribution frequency, and cache management efficiency. The key frequency characteristics include server load frequency and data transmission speed frequency. The performance trend prediction results include model parameters, frequency-performance relationship, and model accuracy.

3. The network storage server efficiency evaluation method according to claim 2, characterized in that: The steps for obtaining the preliminary coded data are specifically as follows: Extract continuous input and output performance data from the network storage server, filter the data segments that fall within the specified time window, and generate a filtered performance data set; Applying data formatting technology to convert the screened performance data set into a standard format suitable for encoding processing to obtain a standardized data set; The standardized data set is encoded using the encoding formula: Calculate the coding efficiency of multiple data points and generate preliminary coding data C; Among them, w i represents the weight parameter of the i-th data point, which is used to adjust the relative criticality of the data point in the encoding process, g i is the gain coefficient, adjusting the data point deviation p i -q i The impact strength, p i represents the value of the actual data point, q i The preset baseline value for the data points.

4. The network storage server efficiency evaluation method according to claim 3, characterized in that: The steps for obtaining the frequency domain conversion result are specifically as follows: Based on the preliminary coded data, adjusting the signal boundary by using a window function to reduce signal leakage during the conversion process, and generating signal data with adjusted boundary; Performing discrete Fourier transform on the signal data of the adjusted boundary to convert the time domain signal into a frequency domain signal to obtain frequency domain data; The frequency domain data is analyzed to identify key frequency components and calculate the energy contribution of the key frequency components using the formula: Quantify hard drive performance and system response time at multiple frequency points and generate frequency domain conversion results; Where F(f) represents the power density at frequency f, X(k) is the complex spectrum value at the kth frequency point, N is the total number of data points, and f is the target frequency.

5. The network storage server efficiency evaluation method according to claim 4, characterized in that: The steps for obtaining the power spectrum analysis results are specifically as follows: Using the frequency domain conversion result, calculating the square of the complex amplitude of each frequency point, testing the energy contribution of multiple frequency points in the total signal, and generating frequency point energy data; Calculating the average power spectrum density of each frequency point from the frequency point energy data, and obtaining average power spectrum data by averaging multiple frequency point data; Analyze the average power spectrum data, identify key energy distribution frequencies, and detect energy peaks using the formula: Calculate the weighted power spectrum density of the frequency point f and generate the power spectrum analysis result; Where E(f) represents the weighted power spectral density of frequency f, which is used to measure the energy distribution of the signal at the frequency. k | 2 is the energy of frequency point k, representing the square of the amplitude of the signal at the frequency point, used to quantify the energy contribution of multiple frequency points, N is the total number of data points, indicating the total number of sampling points in the signal, It is the weighting coefficient of frequency point k, which is adjusted by the cosine function to enhance the weight of the target frequency component.

6. The network storage server efficiency evaluation method according to claim 5, characterized in that: The steps for obtaining the key frequency features are specifically as follows: According to the power spectrum analysis results, the power spectrum data is subjected to frequency analysis to determine the energy distribution in the overall signal, identify significant peaks in the power spectrum, and obtain a peak data list; The peak data list is sorted by size and frequency, and the frequencies with significant energy contribution are screened using the formula: By weighting the energy of each frequency and adjusting it using the square root of the frequency and amplitude, the key frequency data is obtained; Among them, P key Represents the weighted energy of the key frequency, f i Represents the frequency value after screening, indicating that the change of high frequency has a greater impact on the total energy, reflecting the key role of high frequency components in power spectrum analysis. i The amplitude representing the frequency, by taking the square root of the absolute value, reduces the impact of outliers on the overall calculation and enhances the model's compatibility and sensitivity to amplitude changes; Based on the key frequency data, the role of frequency in server load and data transmission speed is iteratively analyzed, frequencies closely related to server performance are identified, and key frequency features are generated.

7. The network storage server efficiency evaluation method according to claim 6, characterized in that: The steps for obtaining the performance trend prediction results are specifically as follows: Configuring support vector machine model parameters, adjusting the support vector configuration to match input data based on the key frequency features, and generating model configuration parameters; The support vector machine model is trained using the model configuration parameters, and the relationship between the learning frequency characteristics and the performance changes is refined through iterative optimization to obtain a trained model; Perform accuracy verification of the trained model, use the verification data set to evaluate the consistency between the model prediction results and the actual data, and generate the model verification results by calculating the error rate; Based on the model validation results, the model prediction results were analyzed using the formula: Generate performance trend forecast results; Among them, P trend Represents the performance trend prediction result, α i is the model weight, which adjusts the influence of each data point in the performance trend prediction, y i is the actual performance change data, which is the basic input for model prediction. β is the error adjustment coefficient, which is used to optimize the accuracy of the prediction model.

8. A network storage server efficiency evaluation system, characterized in that: The network storage server efficiency evaluation method according to any one of claims 1 to 7, wherein the system comprises: The data acquisition module collects continuous input and output performance data from the network storage server, performs time window screening on the data, determines the data within the target time period, performs digital coding, and obtains preliminary coded data; The signal analysis module controls the signal boundary based on the preliminary encoded data, performs a discrete Fourier transform, converts the time series data into frequency domain data in complex form, identifies the key frequency components of the hard disk performance and system response time, and generates a frequency domain conversion result; The spectrum estimation module uses the frequency domain conversion results to average the squared amplitudes of the frequency points, estimate the power spectrum density, analyze the main energy distribution frequencies, detect the server cache management efficiency, and output the power spectrum analysis results; The performance prediction module identifies and analyzes the peaks in the spectrum based on the power spectrum analysis results, extracts key frequency features that characterize server load and data transmission speed, configures and trains a support vector machine model, learns the performance change relationship, verifies the model accuracy, and obtains performance trend prediction results.

Images