Storage device performance testing method, device, equipment, medium and product

Abstract

The embodiment of the present invention provides a storage device performance test method, apparatus, equipment, medium and product, including: obtaining characteristic parameters corresponding to storage device performance; determining the correlation between characteristic parameters and IOPS of storage devices; performing preset dimension combination processing on different characteristic parameters based on the correlation and preset SISSO rules to obtain a target characteristic parameter combination; training a preset neural network model through the target characteristic parameter combination to obtain a target neural network model; predicting the performance of storage devices in the server in the current state through the target neural network model, and outputting the predicted value of storage device performance. The present invention obtains key characteristic parameter factors that affect the performance of RAID card chips by utilizing a neural network model and a method for screening SISSO characteristic parameters, and can further test the performance of RAID card chips, optimize the performance of RAID card chips to the greatest extent, and further reduce labor costs.

Description

Technical Field

[0001] The present invention relates to the technical field of storage devices, and particularly to a method, device, equipment, medium and product for testing the performance of a storage device. Background Art

[0002] A RAID (Redundant Array of Independent Disks) card organizes multiple disks together, which is an independent and larger disk for users. Therefore, applying RAID technology to handle massive data and server storage has good application prospects. The access speed of an SSD disk (Solid State Disk) is too slow compared to the CPU (Central Processing Unit). Through RAID technology, the overall access speed of the disks can be doubled by concurrently accessing multiple disks. In addition, RAID technology can also provide fault tolerance functions through data verification, backup, etc., making the system more stable and secure.

[0003] Therefore, the performance of the RAID card chip directly affects the overall performance of the information system, and it is necessary to perform performance testing on it in a timely manner to maintain abnormal disks and ensure the normal operation of the information system.

[0004] However, there are still some drawbacks in the existing methods for testing the performance of RAID card chips. In the prior art, the disk performance is tested regularly at fixed times. Mainly, testers use the FIO (Flexible I / O Tester) tool to test the performance of the RAID card chip in specific use case scenarios, which cannot cover many scenarios, and a lot of manpower, test platform and tool resources are consumed within the fixed time, resulting in a reduction in the accuracy and efficiency of the storage device performance test. Summary of the Invention

[0005] The purpose of the embodiments of the present invention is to provide a method, device, equipment, medium and product for testing the performance of a storage device. The specific technical solutions are as follows:

[0006] In the first aspect of the implementation of the present invention, first, a method for testing the performance of a storage device is provided, which acquires characteristic parameters corresponding to the performance of the storage device;

[0007] Determine the correlation between the characteristic parameters and the IOPS of the storage device;

[0008] Based on the correlation and a preset SISSO rule, perform preset dimension combination processing on different characteristic parameters to obtain a target characteristic parameter combination;

[0009] Training a preset neural network model with the target feature parameter combination to obtain a target neural network model;

[0010] Predicting the performance of the storage device in the server in the current state through the target neural network model, and outputting a storage device performance prediction value.

[0011] Optionally, the preset dimensionality combination processing of different feature parameters based on the correlation degree and a preset SISSO rule to obtain a target feature parameter combination includes:

[0012] Filtering out descriptors related to the IOPS of the storage device from the feature parameters based on a determined independent screening rule;

[0013] Composing a feature subspace based on the descriptors;

[0014] Obtaining target descriptors from the feature subspace based on a sparse operator rule and a preset specification algorithm;

[0015] Performing preset dimensionality combination processing based on the target descriptors to obtain a target feature parameter combination.

[0016] Optionally, the preset dimensionality combination processing of different feature parameters based on the correlation degree and a preset SISSO rule to obtain a target feature parameter combination includes:

[0017] When it is detected that the correlation degree is less than a first preset threshold, obtaining a second set of feature parameters;

[0018] Performing preset dimensionality combination processing on the second set of feature parameters through a preset SISSO rule to obtain a second target feature parameter combination;

[0019] Determining the correlation degree between the second target feature parameter combination and the IOPS;

[0020] When it is detected that the correlation degree is greater than or equal to the preset threshold, outputting the second target feature parameter combination.

[0021] Optionally, after the step of determining the correlation degree between the second target feature parameter combination and the IOPS, the method includes:

[0022] When it is detected that the correlation degree is less than the first preset threshold, removing the second target feature parameter combination.

[0023] Optionally, the preset dimensionality combination processing of different feature parameters based on the correlation degree and a preset SISSO rule to obtain a target feature parameter combination includes:

[0024] When it is detected that the relevance is greater than or equal to the first preset threshold, obtain the first set of feature parameters;

[0025] Perform preset dimensional combination processing on the first set of feature parameters through a preset SISSO rule to obtain a first target combination of feature parameters.

[0026] Optionally, the storage device includes a RAID card chip; the predicting the performance of the storage device in the server in the current state through the target neural network model and outputting a storage device performance prediction value includes:

[0027] Input the data related to the performance of the RAID card chip in the server in the current state into the target neural network model, and output a storage device performance prediction value.

[0028] Optionally, before the step of obtaining the feature parameters corresponding to the storage device performance, the method includes:

[0029] Obtain an IOPS test data set corresponding to the storage device;

[0030] The obtaining the feature parameters corresponding to the storage device performance includes:

[0031] Obtain the feature parameters corresponding to the storage device performance based on the previously obtained IOPS test data set corresponding to the storage device.

[0032] Optionally, the feature parameters corresponding to the storage device performance include at least one of RAID level, number of disks, stripe size, number of logical units, number of storage output links, disk type, RAID controller information, disk capacity, RAID group status, cache configuration, RAID reconstruction time, RAID stripe depth, RAID write penalty, number of disk failures in the RAID group, and load balancing policy of the RAID group.

[0033] Optionally, the obtaining the IOPS test data set corresponding to the storage device includes:

[0034] In response to a first input on the human-computer interaction interface, obtain test parameters and IOPS for the storage device in the server;

[0035] Generate the IOPS test data set corresponding to the storage device based on the test parameters and the IOPS.

[0036] Optionally, before the step of, in response to a first input on the human-computer interaction interface, obtaining test parameters and IOPS for the storage device in the server, the method includes:

[0037] Create a test environment corresponding to the storage device.

[0038] Optionally, training the preset neural network model with the target feature parameter combination to obtain the target neural network model includes:

[0039] Debugging the hyperparameters corresponding to the preset neural network model to obtain target hyperparameters;

[0040] Based on the target hyperparameters, the target feature parameter combination, and a preset training dataset, iteratively training the preset neural network model until the preset neural network module reaches a preset number of training times to obtain the target neural network model.

[0041] Optionally, after the step of predicting the performance of the storage device in the server in the current state by the target neural network model and outputting a storage device performance prediction value, the method includes:

[0042] Judging whether the IOPS performance of the storage device meets a preset condition based on the storage device performance prediction value;

[0043] If so, continue to predict the performance of the storage device by the target neural network model.

[0044] Optionally, after the step of judging whether the IOPS performance of the storage device meets a preset condition based on the storage device performance prediction value, the method includes:

[0045] If not, adjusting the initial monitoring time based on the storage device performance prediction value and re-obtaining the IOPS test dataset corresponding to the storage device.

[0046] Optionally, the adjusting the initial monitoring time based on the storage device performance prediction value includes:

[0047] When it is monitored that the storage device performance prediction value is less than a second preset threshold, shortening the initial monitoring time according to a preset time rule to obtain a first monitoring time, adding the obtained test data to the test dataset, retraining the neural network model, and predicting the performance of the storage device by using the new neural network model;

[0048] When it is monitored that the storage device performance prediction value is greater than or equal to the second preset threshold, increasing the monitoring time according to the preset time rule, adding the test dataset to the original sample, continuing to train the model, and predicting the performance of the storage device by using the new neural network model.

[0049] Optionally, after the step of predicting the performance of the storage device in the server in the current state by the target neural network model and outputting a storage device performance prediction value, the method includes:

[0050] Obtain the actual value of the storage device performance;

[0051] Perform mean absolute error processing based on the actual value of the storage device performance and the predicted value of the storage device performance to obtain a first target difference;

[0052] Optimize the target neural network model based on the first target difference.

[0053] Optionally, after the step of predicting the performance of the storage device in the server in the current state through the target neural network model and outputting the predicted value of the storage device performance, the method includes:

[0054] Obtain the actual value of the storage device performance;

[0055] Perform root mean square error processing based on the actual value of the storage device performance and the predicted value of the storage device performance to obtain a second target difference;

[0056] Optimize the target neural network model based on the second target difference.

[0057] In the second aspect of the implementation of the present invention, there is also provided a storage device performance testing device, and the device includes:

[0058] An acquisition module, configured to acquire characteristic parameters corresponding to the storage device performance;

[0059] A determination module, configured to determine the correlation between the characteristic parameters and the IOPS of the storage device;

[0060] A combination module, configured to perform preset dimension combination processing on different characteristic parameters based on the correlation and a preset SISSO rule to obtain a target characteristic parameter combination;

[0061] A training module, configured to train a preset neural network model through the target characteristic parameter combination to obtain a target neural network model;

[0062] A prediction module, configured to predict the performance of the storage device in the server in the current state through the target neural network model and output the predicted value of the storage device performance.

[0063] In the third aspect of the implementation of the present invention, there is also provided a communication device, including: a transceiver, a memory, a processor, and a program stored on the memory and executable on the processor;

[0064] The processor is configured to read the program in the memory to implement the storage device performance testing method according to any one of the first aspect.

[0065] In the fourth aspect of the implementation of the present invention, a computer-readable storage medium is further provided. Instructions are stored in the computer-readable storage medium, and when it runs on a computer, it enables the computer to implement the storage device performance testing method described in any one of the first aspects.

[0066] In the fifth aspect of the implementation of the present invention, a computer program product is further provided, including a computer program / instructions. When the computer program / instructions are executed by a processor, they implement the storage device performance testing method described in any one of the first aspects.

[0067] The storage device performance testing method provided by the embodiments of the present invention includes: obtaining characteristic parameters corresponding to the performance of the storage device; determining the correlation between the characteristic parameters and the IOPS of the storage device; performing preset dimension combination processing on different characteristic parameters based on the correlation and a preset SISSO rule to obtain a target characteristic parameter combination; training a preset neural network model through the target characteristic parameter combination to obtain a target neural network model; predicting the performance of the storage device in the server under the current state through the target neural network model, and outputting a storage device performance prediction value. In the embodiments of the present invention, by using a neural network model and a method for screening SISSO characteristic parameters, key characteristic parameter factors affecting the performance of the RAID card chip are obtained, and then the performance of the RAID card chip can be further tested, the performance of the RAID card chip can be optimized to the greatest extent, the labor cost can be further reduced, and furthermore, by predicting the performance of the RAID card chip based on relatively highly correlated characteristic parameters, the maximum performance of the RAID card chip under a certain capacity, a certain RAID scenario, and the surrounding environment can be obtained. Description of the Drawings

[0068] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art.

[0069] Figure 1 It is a flowchart of the steps of a storage device performance testing method provided by an embodiment of the present invention;

[0070] Figure 2 It is a block diagram of a storage device performance testing device provided by an embodiment of the present invention;

[0071] Figure 3 It is a schematic diagram of a communication device provided by an embodiment of the present invention;

[0072] Figure 4 It is a schematic diagram of an exemplary IO path for testing a RAID card chip from a host provided by an embodiment of the present invention;

[0073] Figure 5It is a schematic diagram of an exemplary storage device structure provided by an embodiment of the present invention;

[0074] Figure 6 It is a schematic diagram of the working process of an exemplary SISSO operation provided by an embodiment of the present invention;

[0075] Figure 7 It is a schematic diagram of an exemplary neural network model provided by an embodiment of the present invention;

[0076] Figure 8 It is another schematic diagram of an exemplary neural network model provided by an embodiment of the present invention;

[0077] Figure 9 It is a schematic diagram of an exemplary storage device performance test process provided by an embodiment of the present invention. Detailed implementation manners

[0078] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will elaborate on each implementation manner of the present invention with reference to the accompanying drawings. However, those of ordinary skill in the art can understand that in each implementation manner of the present invention, many technical details are provided to help readers better understand the present invention. However, even without these technical details and various changes and modifications based on the following implementation manners, the technical solutions claimed by the present invention can still be implemented. The division of the following embodiments is for convenience of description and should not constitute any limitation on the specific implementation manner of the present invention. The various embodiments can be combined and cross-referenced with each other on the premise of not being contradictory.

[0079] Referring to Figure 1 , a flowchart of the steps of a storage device performance test method provided by an embodiment of the present invention is shown. The method may include:

[0080] Step 101, obtaining characteristic parameters corresponding to the performance of the storage device;

[0081] Step 102, determining the correlation degree between the characteristic parameters and the IOPS of the storage device;

[0082] Optionally, before the step of obtaining characteristic parameters corresponding to the performance of the storage device, the method includes:

[0083] Obtaining an IOPS test data set corresponding to the storage device;

[0084] The obtaining of the characteristic parameters corresponding to the performance of the storage device includes:

[0085] Obtaining characteristic parameters corresponding to the performance of the storage device based on the previously obtained IOPS test data set corresponding to the storage device.

[0086] Optionally, the characteristic parameters corresponding to the storage device performance include at least one of RAID level, number of disks, stripe size, number of logical units, number of storage output links, disk type, RAID controller information, disk capacity, RAID group status, cache configuration, RAID reconstruction time, RAID stripe depth, RAID write penalty, number of disk failures in the RAID group, and load balancing policy of the RAID group.

[0087] It should be noted that in the embodiments of the present invention, a performance test model of the RAID card chip is constructed based on different RAID level information, disk size, link information, etc. through a neural network model. Through this invention, the performance of the RAID card chip can be understood in real time, the hard disk stocking cost can be effectively reduced, the risk of data loss can be greatly reduced, and reliable data basis for hard disk maintenance is provided. The specific implementation process can refer to Figure 9 As shown, specifically, collect the IOPS test data set of the RAID card chip, obtain multiple characteristic parameters related to performance (such as RAID level, disk size, link information, etc.), and by judging the correlation between these characteristic parameters and the IOPS test data set, obtain a parameter set with higher correlation and a parameter set with lower correlation respectively. For different parameter sets, use SISSO to combine the characteristic parameters in different dimensions and judge their correlation with the IOPS test data set again. Use the filtered characteristic parameters (the combined characteristic parameters in the parameter set with higher correlation and the characteristic parameters with higher correlation after combination in the parameter set with lower correlation) to train the neural network model and construct a prediction model that can closely reflect the performance of the RAID card chip. After obtaining the prediction model, use the model to quickly predict the performance of the RAID card chip in the current state and judge whether the IOPS performance is excellent; if the performance is excellent, further judge whether it is close to the set monitoring time. If it is close to the set monitoring time, continue to use the model for prediction, otherwise end the process; if the performance is not excellent, shorten the monitoring time and obtain a new IOPS test data set.

[0088] Based on the theoretical deduction and test analysis of the performance of the RAID card chip, several characteristic parameters with the highest correlation with the performance of the RAID card chip are selected. For example, the characteristic parameter of the disk size mainly refers to the storage capacity, which determines the amount of data that can be stored. RAID (Redundant Array of Independent Disks) information (such as manufacturer, model, SN, status, etc.), Name Space (logical space accessible by the host) information, SSD (Solid State Disk) information (such as manufacturer, model, SN, status, capacity, etc.), different levels of RAID information: RAID-0, RAID-1, RAID-2, RAID-3, RAID-4, RAID-5, RAID-10, RAID-50, RAID-TP, etc. The stripe of RAID is the basic unit for RAID to process data, and the stripe also has a number starting from 0. The information on the number of disks included in RAID, that is, the size of the disks, the number of disks, and the data in the hard disk usage environment of each solid-state disk manufacturer are used to obtain the values of other characteristic parameters. The information on the number of output links of the storage: refers to the number of external connections or output channels that the storage device (such as a hard disk, storage array, etc.) can support. The number of LUNs (Logical Unit Number, a logical unit on a physical storage device) created for each RAID: actually depends on multiple factors, including the configuration of the RAID, the capacity of the storage device, the performance requirements of the storage system, and the data management strategy, etc. In a RAID environment, a LUN is usually a virtual storage unit created on a RAID group or storage pool, used to provide storage space for the host. Each LUN can be regarded as an independent storage volume and can be recognized and managed by the operating system or application program. The configuration of the RAID (such as RAID 0, RAID 1, RAID 5, etc.) will affect the creation and performance of the LUN. For example, RAID 0 provides high performance but no data redundancy, while RAID 5 may sacrifice some performance while providing data redundancy. Therefore, when selecting the RAID level, the requirements for performance, reliability, and data protection need to be considered.

[0089] The structure of the solid-state disk SSD is as Figure 5 shown, demonstrating that the solid-state disk can be used for garbage collection, bad block handling, address translation, and wear leveling, as well as the flash interface and multiple flash chips.

[0090] Optionally, the obtaining of the IOPS test data set corresponding to the storage device includes:

[0091] In response to the first input on the human-computer interaction interface, obtain the test parameters and IOPS for the storage device in the server;

[0092] Generate an IOPS test data set corresponding to the storage device based on the test parameters and the IOPS.

[0093] It should be noted that in the embodiments of the present invention, test parameters and IOPS (Input / Output Operations Per Second, the number of input / output operations that can be completed per second) input on the human-computer interaction interface for the RAID card chip to be tested in the server are collected. As Figure 4 shown, it is an IO (Input / Output) path diagram for testing the RAID card chip from the host. The host accesses multiple logical spaces through the memory and input / output requests. The logical spaces then organize data through the redundant array of independent disks 5 and redundant array of independent disks 0 technologies, and multiple solid-state drives are used as the storage medium. For different RAID levels, random write, random read, sequential write, sequential read, and mixed read / write operations for block sizes such as 4K, 8K, 16K, 32K, 64K, 128K, 256K, 1M, 4M, 16M, 1G, 10G, etc. for the disk can be performed. In the present invention, the collected IOPS of the RAID card chip is mainly a series of test data sets obtained based on the FIO test tool.

[0094] Optionally, before the step of obtaining the test parameters and IOPS for the storage device in the server in response to the first input on the human-computer interaction interface, the method includes:

[0095] Create a test environment corresponding to the storage device.

[0096] Step 103, perform a preset dimension combination process on different feature parameters based on the relevance and a preset SISSO rule to obtain a target feature parameter combination;

[0097] Optionally, the performing a preset dimension combination process on different feature parameters based on the relevance and a preset SISSO rule to obtain a target feature parameter combination includes:

[0098] Based on a determined independent screening rule, screen out descriptors related to the IOPS of the storage device from the feature parameters;

[0099] Form a feature subspace based on the descriptors;

[0100] Obtain target descriptors from the feature subspace based on a sparse operator rule and a preset specification algorithm;

[0101] Perform a preset dimension combination process on the target descriptors to obtain a target feature parameter combination.

[0102] Optionally, the preset dimension combination processing of different feature parameters based on the relevance and the preset SISSO rule to obtain a target feature parameter combination includes:

[0103] When it is detected that the relevance is less than a first preset threshold, obtain a second set of feature parameters;

[0104] Perform preset dimension combination processing on the second set of feature parameters through the preset SISSO rule to obtain a second target feature parameter combination;

[0105] Determine the relevance between the second target feature parameter combination and the IOPS;

[0106] When it is detected that the relevance is greater than or equal to the preset threshold, output the second target feature parameter combination.

[0107] Optionally, after the step of determining the relevance between the second target feature parameter combination and the IOPS, the method includes:

[0108] When it is detected that the relevance is less than the first preset threshold, remove the second target feature parameter combination.

[0109] Optionally, the preset dimension combination processing of different feature parameters based on the relevance and the preset SISSO rule to obtain a target feature parameter combination includes:

[0110] When it is detected that the relevance is greater than or equal to a first preset threshold, obtain a first set of feature parameters;

[0111] Perform preset dimension combination processing on the first set of feature parameters through the preset SISSO rule to obtain a first target feature parameter combination.

[0112] It should be noted that in the embodiments of the present invention, in order to obtain multiple feature parameters related to the performance of the RAID card chip, the SISSO method based on the compressed sensing theory is used to combine the feature parameters.

[0113] Specifically, on the one hand, after screening out several feature parameters highly relevant to the performance of the RAID card chip, using the SISSO method based on the compressed sensing theory, through the descriptor , perform random non-linear combinations, so that a large number of descriptors can be obtained, and the set they form is called the feature space. After combining the initial feature parameters, as the number of iterations increases, the feature space will show a recursive growth, generally reaching the order of 1010.

[0114] On the other hand, after screening out several feature parameters with relatively low correlation with the performance of the RAID card chip, the SISSO method based on the theory of compressive sensing is used to perform random non-linear combinations on them, a large number of descriptors can be obtained, and several feature parameters highly correlated with the performance of the RAID card chip are screened out, and the descriptors with relatively low correlation are discarded. In order to reduce the dimension of the feature space, SISSO screens out the descriptors related to the IOPS performance of the RAID card chip by determining the Sure Independence Screening (SIS) method, and then forms a feature subspace. Subsequently, SISSO adopts a Sparsifying Operator (SO) to select Canonical minimization or the Canonical minimization algorithm to further find the optimal dimensional descriptors from the feature subspace.

[0115] Among them, Figure 6 shows the working process of the SISSO (Sure Independence Screening and Sparsifying Operator, which finds the most important few features describing the target property or function from a huge feature space) operation. It can handle a huge feature space, and the optimal descriptors determined by SISSO are explicit functions of the input feature parameters. Specifically, by screening key subspaces (such as S 1D 、S 2D 、S nD ) from the huge feature space Φ, and generating one-dimensional, two-dimensional to n-dimensional descriptors. SISSO combines Sure Independence Screening (SIS, a statistical method for feature selection of high-dimensional data) and Sparsifying Operator (SO, a sparsifying operator for introducing sparsity into the model), and finds the most important few features describing the target property or function from a huge feature space, and can be widely applied to data-driven problems such as regression, classification, and multi-task learning.

[0116] Step 104, training a preset neural network model with the target feature parameter combination to obtain a target neural network model;

[0117] Optionally, the training of the preset neural network model with the target feature parameter combination to obtain a target neural network model includes:

[0118] Debugging the hyperparameters corresponding to the preset neural network model to obtain target hyperparameters;

[0119] Iteratively train the preset neural network model based on the target hyperparameters, the target feature parameter combination, and a preset training dataset until the preset neural network module reaches a preset number of training times, obtaining a target neural network model.

[0120] Construct a neural network model for testing the performance of a RAID card chip. To obtain an optimal neural network model, hyperparameters (including the number of hidden layers and their neurons, the size of the batchsize, and the choice of optimizer, etc.) were appropriately debugged, and the training dataset was used to iteratively train the neural network model until the number of training times reached that of the optimal neural network model. Among them, the neural network model includes a series of interconnected neurons. A single neuron mainly has the following five components (input, weights and threshold, summation, activation function, and output), as Figure 7 shown, receiving input data, weighting the input data through weights, calculating the sum of the weighted inputs through summation, introducing a non-linear transformation using the activation function, and finally generating output data. First is the input. Assuming the input of the neuron is: , it can be denoted as an n-dimensional column vector X: 、 The neural network mainly includes two parts.

[0121] The first part is the forward propagation of the signal. Specifically, for the input layer, the input of the j-th node is (j = 1, 2, …, M); the connection weights between the hidden layer and the output layer can be represented by , is the threshold in the hidden layer, is the transfer function; for the output layer, (i = 1, 2, …, q) represents the connection weights from the output layer to the hidden layer, (k = 1, 2, …, L) is the threshold of the k-th node, is the transfer function, is the output of the k-th node. The input of the i-th node in the hidden layer can be denoted as neti: , and the signal output of the i-th node in the hidden layer can be represented by : , and the signal input of the k-th node in the output layer can be represented by : , and the signal output of the k-th node in the output layer can be represented by : .

[0122] The second part is the backpropagation of the error. Specifically, as Figure 8 shown, for the input layer, the input of the j-th node is , the signal output of the k-th node in the output layer is , if the target value of the k-th node in the output layer can be represented by Tk, then the error function of the p-th sample is: 、For the output layer, the weight correction amount can be represented, and the threshold correction amount can be represented; for the hidden layer, represents the weight correction amount, represents the threshold correction amount, is the learning rate. First, the calculation error of the neural network is fed back from the output layer to each hidden layer, and the weights and thresholds in each layer are adjusted using the gradient descent algorithm.

[0123] Step 105, predict the performance of the storage device in the server in the current state through the target neural network model, and output the storage device performance prediction value.

[0124] Further, the storage device includes a RAID card chip; the predicting the performance of the storage device in the server in the current state through the target neural network model and outputting the storage device performance prediction value includes:

[0125] Input the data related to the performance of the RAID card chip in the server in the current state into the target neural network model, and output the storage device performance prediction value.

[0126] Input the data related to the performance of the RAID card chip in the server in the current state into the neural network model for predicting the performance of the RAID card chip, so as to quickly predict the performance of the RAID card chip in the server in the current state.

[0127] Optionally, after the step of predicting the performance of the storage device in the server in the current state through the target neural network model and outputting the storage device performance prediction value, the method includes:

[0128] Judge whether the IOPS performance of the storage device meets the preset condition based on the storage device performance prediction value;

[0129] If so, continue to predict the performance of the storage device through the target neural network model.

[0130] Optionally, after the step of judging whether the IOPS performance of the storage device meets the preset condition based on the storage device performance prediction value, the method includes:

[0131] If not, adjust the initial monitoring time based on the storage device performance prediction value, and re-acquire the IOPS test data set corresponding to the storage device.

[0132] Optionally, the adjustment process of the initial monitoring time based on the predicted value of the storage device performance includes:

[0133] When it is monitored that the predicted value of the storage device performance is less than the second preset threshold, the initial monitoring time is shortened according to the preset time rule to obtain the first monitoring time, and the obtained test data is added to the test data set, the neural network model is retrained, and the performance of the storage device is predicted by using the new neural network model;

[0134] When it is monitored that the predicted value of the storage device performance is greater than or equal to the second preset threshold, the monitoring time is increased according to the preset time rule, the test data set is added to the original sample, the model is continuously trained, and the performance of the storage device is predicted by using the new neural network model.

[0135] It should be noted that in the embodiment of the present invention, when the device system performs a performance test on the RAID card chip, the monitoring time is dynamically adjusted according to the test result (i.e., the size of the IOPS). For example, when it is monitored that the IOPS is small, the monitoring time is shortened, and the obtained test data is added to the test data set, the neural network model is retrained, and the performance of the RAID card chip is predicted by using the new neural network model. When the IOPS is large, the monitoring time is increased, and the test data set is also added to the original sample, the model is continuously trained, and the performance of the RAID card chip is predicted by using the new neural network model. Thereby, the performance of the RAID card chip can be understood in real time, the hard disk stocking cost can be effectively reduced, the risk of data loss can be greatly reduced, and reliable data basis for hard disk maintenance can be provided.

[0136] Optionally, after the step of predicting the performance of the storage device in the server in the current state by using the target neural network model and outputting the predicted value of the storage device performance, the method includes:

[0137] Obtain the actual value of the storage device performance;

[0138] Perform mean absolute error processing based on the actual value of the storage device performance and the predicted value of the storage device performance to obtain the first target difference;

[0139] Optimize the target neural network model based on the first target difference.

[0140] Optionally, after the step of predicting the performance of the storage device in the server in the current state by using the target neural network model and outputting the predicted value of the storage device performance, the method includes:

[0141] Obtain the actual value of the storage device performance;

[0142] Perform root mean square error processing on the actual value of the storage device performance and the predicted value of the storage device performance to obtain a second target difference;

[0143] Optimize the target neural network model based on the second target difference.

[0144] It should be noted that in the embodiments of the present invention, the mean absolute error MAE (Mean Absolute Error) is a statistic that can be used to measure the distance between the predicted value and the actual value. Generally, the smaller the MAE value, the higher the accuracy of the ANN model (Artificial Neural Network) in predicting the performance of the actual RAID card chip. The root mean square error RMSE (Root Mean Square Error) is the square root of the ratio of the sum of the squares of the deviations between the predicted value and the true value to the number of observations n. RMSE measures the deviation between the predicted value and the true value and is sensitive to outliers in the data. When the above-mentioned test device for the performance of the RAID card chip performs performance testing, only the above-mentioned division of each program module is used as an example. In actual applications, the above-mentioned processing can be allocated to different program modules as needed, that is, the internal structure of the device is divided into different program modules to complete all or part of the above-described processing.

[0145] The storage device performance testing method provided by the embodiments of the present invention includes: obtaining characteristic parameters corresponding to the storage device performance; determining the correlation between the characteristic parameters and the IOPS of the storage device; performing preset dimension combination processing on different characteristic parameters based on the correlation and a preset SISSO rule to obtain a target characteristic parameter combination; training a preset neural network model through the target characteristic parameter combination to obtain a target neural network model; predicting the performance of the storage device in the server in the current state through the target neural network model and outputting a predicted value of the storage device performance. In the embodiments of the present invention, by using a neural network model and a method for screening SISSO characteristic parameters, key characteristic parameter factors affecting the performance of the RAID card chip are obtained, and then the performance of the RAID card chip can be further tested, the performance of the RAID card chip can be optimized to the greatest extent, the labor cost can be further reduced, and furthermore, by predicting the performance of the RAID card chip based on the characteristic parameters with higher correlation, the maximum performance of the RAID card chip under a certain capacity, a certain RAID scenario, and the surrounding environment can be obtained.

[0146] Refer to Figure 2, showing a schematic structural diagram of a power backup device provided by an embodiment of the present invention, which is applied to a baseboard management controller in an interconnect switch board of a memory resource all-in-one machine. The memory resource all-in-one machine includes the interconnect switch board, a power module, and a power backup module. The device includes:

[0147] An acquisition module 201, configured to acquire characteristic parameters corresponding to the performance of the storage device;

[0148] A determination module 202, configured to determine the relevance between the characteristic parameters and the IOPS of the storage device;

[0149] A combination module 203, configured to perform preset dimension combination processing on different characteristic parameters based on the relevance and a preset SISSO rule to obtain a target characteristic parameter combination;

[0150] A training module 204, configured to train a preset neural network model through the target characteristic parameter combination to obtain a target neural network model;

[0151] A prediction module 205, configured to predict the performance of the storage device in the server in the current state through the target neural network model and output a storage device performance prediction value.

[0152] The storage device performance testing device provided by the embodiment of the present invention acquires characteristic parameters corresponding to the performance of the storage device; determines the relevance between the characteristic parameters and the IOPS of the storage device; performs preset dimension combination processing on different characteristic parameters based on the relevance and a preset SISSO rule to obtain a target characteristic parameter combination; trains a preset neural network model through the target characteristic parameter combination to obtain a target neural network model; predicts the performance of the storage device in the server in the current state through the target neural network model and outputs a storage device performance prediction value. In the embodiment of the present invention, by using a neural network model and a method for screening SISSO characteristic parameters, key characteristic parameter factors affecting the performance of the RAID card chip are obtained, and further, the performance of the RAID card chip can be further tested, the performance of the RAID card chip is optimized to the greatest extent, the labor cost is further reduced, and moreover, by predicting the performance of the RAID card chip based on relatively highly relevant characteristic parameters, the maximum performance of the RAID card chip under a certain capacity, a certain RAID scenario, and the surrounding environment is obtained.

[0153] The embodiment of the present invention further provides a communication device, as Figure 3 shown, including a processor 301, a communication interface 302, a memory 303, and a communication bus 304. Among them, the processor 301, the communication interface 302, and the memory 303 complete mutual communication through the communication bus 304,

[0154] A memory 303 for storing computer programs;

[0155] A processor 301, when executing the program stored on the memory 303, can implement the following steps:

[0156] Obtain characteristic parameters corresponding to the performance of the storage device;

[0157] Determine the correlation between the characteristic parameters and the IOPS of the storage device;

[0158] Based on the correlation and a preset SISSO rule, perform a preset dimensional combination process on different characteristic parameters to obtain a target characteristic parameter combination;

[0159] Train a preset neural network model through the target characteristic parameter combination to obtain a target neural network model;

[0160] Predict the performance of the storage device in the server in the current state through the target neural network model, and output a storage device performance test value.

[0161] Wherein, the memory and the processor are connected by a bus. The bus can include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors and memories together. The bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art. Therefore, they will not be further described herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a component or multiple components, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on the transmission medium. The data processed by the processor can be transmitted through a wired medium or transmitted on a wireless medium through an antenna. Further, the antenna also receives data and transmits the data to the processor. The processor is responsible for managing the bus and normal processing, and can also provide various functions, including timing, peripheral interface, voltage regulation, power management, and other control functions. The memory can be used to store data used by the processor when executing operations.

[0162] The communication bus mentioned in the above terminal can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, only a thick line is shown in the figure, but it does not mean that there is only one bus or one type of bus.

[0163] The communication interface is used for communication between the above terminal and other devices.

[0164] The memory may include a Random Access Memory (RAM), or may also include a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0165] The aforementioned processor may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0166] In another embodiment provided by the present invention, there is also provided a computer-readable storage medium storing instructions, which when run on a computer, cause the computer to execute the storage device performance testing method described in any one of the above embodiments.

[0167] In another embodiment provided by the present invention, there is also provided a computer program product containing instructions, which when run on a computer, cause the computer to execute the storage device performance testing method described in any one of the above embodiments.

[0168] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. 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 by wire (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that includes one or more integrated available media. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid state disk (SSD)).

[0169] It should be noted that in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including", or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or device that includes a series of elements includes not only those elements but also other elements that are not explicitly listed, or also includes elements that are inherent to such process, method, article, or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article, or device that includes the element.

[0170] Each embodiment in this specification is described in a related manner. The same or similar parts among the embodiments can be referred to each other, and the differences between each embodiment and other embodiments are emphasized. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiment.

[0171] The above are only the preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention are all included in the protection scope of the present invention.

Claims

1. A storage device performance testing method, characterized in that: The method comprises: Acquire characteristic parameters corresponding to the performance of the storage device; Determining the correlation between the characteristic parameter and the IOPS of the storage device; Based on the correlation and the preset SISSO rule, the different feature parameters are processed in a preset dimension combination to obtain a target feature parameter combination; The preset neural network model is trained by the target feature parameter combination to obtain a target neural network model; Predicting the performance of the storage device in the server in the current state through the target neural network model, and outputting the predicted value of the storage device performance; Determine whether the IOPS performance of the storage device meets a preset condition based on the predicted value of the storage device performance; When it is monitored that the predicted value of the storage device performance is less than a second preset threshold, the initial monitoring time is shortened according to the preset time rule to obtain a first monitoring time, and the obtained test data is added to the test data set, the neural network model is retrained, and the new neural network model is used to predict the performance of the storage device; When it is monitored that the predicted value of the storage device performance is greater than or equal to a second preset threshold, the initial monitoring time is increased according to the preset time rule, and the test data set is added to the original sample, the model is continued to be trained, and the new neural network model is used to predict the performance of the storage device.

2. The method according to claim 1, characterized in that The preset dimension combination processing is performed on different feature parameters based on the correlation and the preset SISSO rule to obtain the target feature parameter combination, which includes: Filtering out descriptors related to the IOPS of the storage device from the characteristic parameters based on determining independent filtering rules; forming a feature subspace based on the descriptors; Obtaining a target descriptor from the feature subspace based on a sparse operator rule and a preset canonical algorithm; A preset dimension combination process is performed based on the target descriptor to obtain a target feature parameter combination.

3. The method according to claim 1, characterized in that The preset dimension combination processing is performed on different feature parameters based on the correlation and the preset SISSO rule to obtain the target feature parameter combination, which includes: When it is detected that the correlation is less than a first preset threshold, obtaining a second characteristic parameter set; Performing preset dimension combination processing on the second feature parameter set by a preset SISSO rule to obtain a second target feature parameter combination; Determine the correlation between the second target characteristic parameter combination and the IOPS; When it is detected that the correlation is greater than or equal to the preset threshold, the second target feature parameter combination is output.

4. The method according to claim 3, characterized in that After the step of determining the correlation between the second target characteristic parameter combination and the IOPS, the method includes: When it is detected that the correlation is less than the first preset threshold, the second target feature parameter combination is removed.

5. The method according to claim 1, characterized in that The preset dimension combination processing is performed on different feature parameters based on the correlation and the preset SISSO rule to obtain the target feature parameter combination, which includes: When it is detected that the correlation is greater than or equal to a first preset threshold, obtaining a first feature parameter set; The first feature parameter set is processed by preset dimension combination according to a preset SISSO rule to obtain a first target feature parameter combination.

6. The method according to claim 1, characterized in that The storage device includes a RAID card chip; the target neural network model is used to predict the performance of the storage device in the server in the current state, and the output storage device performance prediction value includes: The performance data related to the RAID card chip in the server in the current state is input into the target neural network model, and the predicted value of the storage device performance is output.

7. The method according to claim 6, characterized in that Before the step of obtaining characteristic parameters corresponding to the performance of the storage device, the method includes: Obtain the IOPS test data set corresponding to the storage device; The acquiring of characteristic parameters corresponding to the performance of the storage device comprises: A characteristic parameter corresponding to the performance of the storage device is obtained based on a pre-acquired IOPS test data set corresponding to the storage device.

8. The method according to claim 7, characterized in that The characteristic parameters corresponding to the storage device performance include at least one of RAID level, number of disks, stripe size, number of logical units, number of storage output links, disk type, RAID controller information, disk capacity, RAID group status, cache configuration, RAID rebuild time, RAID stripe depth, RAID write penalty, number of disk failures in the RAID group, and load balancing strategy of the RAID group.

9. The method according to claim 7, characterized in that: The step of obtaining an IOPS test data set corresponding to the storage device includes: In response to a first input on the human-computer interaction interface, obtaining a test parameter and IOPS for a storage device in the server; An IOPS test data set corresponding to the storage device is generated based on the test parameters and the IOPS.

10. The method according to claim 9, characterized in that Before the step of obtaining test parameters and IOPS for the storage device in the server in response to the first input on the human-computer interaction interface, the method includes: Create a test environment corresponding to the storage device.

11. The method according to claim 1, characterized in that: The training of the preset neural network model by the target feature parameter combination to obtain the target neural network model comprises: Debug the hyperparameters corresponding to the preset neural network model to obtain the target hyperparameters; The preset neural network model is iteratively trained based on the target hyperparameter, the target feature parameter combination and a preset training data set until the preset neural network module reaches a preset number of training times to obtain a target neural network model.

12. The method according to claim 1, characterized in that After the step of judging whether the IOPS performance of the storage device meets a preset condition based on the predicted value of the storage device performance, the method includes: If so, continue to predict the performance of the storage device through the target neural network model.

13. The method according to claim 1, characterized in that After the step of predicting the performance of the storage device in the server in the current state by the target neural network model and outputting the predicted value of the storage device performance, the method includes: Obtaining actual performance value of the storage device; Performing mean absolute error processing based on the actual value of the storage device performance and the predicted value of the storage device performance to obtain a first target difference; The target neural network model is optimized based on the first target difference.

14. The method according to claim 1, characterized in that After the step of predicting the performance of the storage device in the server in the current state by the target neural network model and outputting the predicted value of the storage device performance, the method includes: Obtaining actual performance value of the storage device; Performing root mean square error processing based on the actual value of the storage device performance and the predicted value of the storage device performance to obtain a second target difference value; The target neural network model is optimized based on the second target difference.

15. A storage device performance testing device, characterized in that: The device comprises: An acquisition module, used to acquire characteristic parameters corresponding to the performance of the storage device; A determination module, used to determine the correlation between the characteristic parameter and the IOPS of the storage device; A combination module, used for performing preset dimension combination processing on different feature parameters based on the correlation and the preset SISSO rule to obtain a target feature parameter combination; A training module, used to train a preset neural network model through the target feature parameter combination to obtain a target neural network model; A prediction module is used to predict the performance of the storage device in the server in the current state through the target neural network model and output the predicted value of the storage device performance; based on the predicted value of the storage device performance, determine whether the IOPS performance of the storage device meets the preset conditions; when it is monitored that the predicted value of the storage device performance is less than a second preset threshold, shorten the initial monitoring time according to the preset time rule to obtain the first monitoring time, and add the obtained test data to the test data set, retrain the neural network model, and use the new neural network model to predict the performance of the storage device; when it is monitored that the predicted value of the storage device performance is greater than or equal to the second preset threshold, increase the initial monitoring time according to the preset time rule, add the test data set to the original sample, continue to train the model, and use the new neural network model to predict the performance of the storage device.

16. A communication device, characterized in that: include: A transceiver, a memory, a processor, and a program stored on the memory and executable on the processor; The processor is used to read the program in the memory to implement the storage device performance testing method as described in any one of claims 1-14.

17. A readable storage medium for storing a program, characterized in that: When the program is executed by a processor, the storage device performance testing method as described in any one of claims 1 to 14 is implemented.

18. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, a storage device performance testing method as described in any one of claims 1-14 is implemented.

Images