A storage device control method and system based on pulse neural network

Abstract

The present application provides a storage device control method and system based on a pulse neural network, which relates to the field of intelligent control technology, including: performing multi-dimensional feature perception and collection; matching a first operating state; reading a first predetermined state feature index, and traversing the matching in the first operating state; when the fitness of the first state does not reach the fitness limit, issuing a first control instruction; obtaining a first real-time control signal of a first storage unit, and encoding it; comparing the real-time pulse sequence with the predetermined pulse sequence to obtain a comparison result, analyzing to obtain the optimal control signal, and performing storage control. This application can solve the technical problems in the prior art that the storage control relies on preset algorithms and rules, cannot dynamically adjust the control strategy according to real-time changes, and lacks flexibility and adaptability. Through multi-layer perceptrons and pulse neural networks, the personalized and adaptive adjustment of the control strategy is realized, and the adaptability and flexibility of the control strategy are enhanced.

Description

Technical Field

[0001] The present application relates to the field of intelligent control technology, and in particular to a storage device control method and system based on a pulse neural network. Background Art

[0002] With the advancement of data recording technology, the demand for data storage is increasing, and solid-state drives (SSDs) have become mainstream storage devices due to their high-speed read and write performance and stability. However, with the increase in storage density and workload, traditional SSD control methods face a series of challenges. Traditional SSD control methods usually rely on fixed algorithms and preset control strategies, and cannot be dynamically adjusted according to different data storage requirements and changing working environments encountered in actual operation. When dealing with complex and dynamic data storage requirements, they often lack flexibility and adaptability. Because the operating status of each storage unit of the hard disk cannot be fully considered, the control strategy is not sophisticated enough.

[0003] In summary, the prior art has the technical problem that the storage control relies on preset algorithms and rules, cannot dynamically adjust the control strategy according to real-time changes, and lacks flexibility and adaptability. Summary of the invention

[0004] The purpose of this application is to provide a storage device control method and system based on a pulse neural network to solve the technical problems in the prior art that storage control relies on preset algorithms and rules, cannot dynamically adjust the control strategy according to real-time changes, and lacks flexibility and adaptability.

[0005] In view of the above problems, the present application provides a storage device control method and system based on pulse neural network.

[0006] In a first aspect, the present application provides a storage device control method based on a pulse neural network, and the storage device control method based on a pulse neural network is implemented by a storage device control system based on a pulse neural network, wherein the storage device control method based on a pulse neural network includes: performing multi-dimensional feature perception and collection on a target solid-state hard disk through a multilayer perceptron to obtain a target hard disk state, wherein the target hard disk state includes multiple operating states of multiple storage units; matching a first operating state of a first storage unit in the multiple operating states, wherein the first storage unit is any one of the multiple storage units, and the first operating state refers to information when the first storage unit writes a first data set; reading a first predetermined state feature indicator, and traversing and matching in the first operating state based on the first predetermined state feature indicator to obtain first state feature information; when the first state fitness obtained by analyzing the first state feature information does not reach the fitness limit, issuing a first control instruction; obtaining a first real-time control signal of the first storage unit based on the first control instruction, and encoding the first real-time control signal to obtain a first real-time pulse sequence; comparing the first real-time pulse sequence with the predetermined pulse sequence to obtain a first comparison result, and analyzing the first comparison result to obtain a first optimal control signal; according to the first optimal control signal, controlling the target solid-state hard disk to perform storage control on the first data set.

[0007] In a second aspect, the present application further provides a storage device control system based on a pulse neural network, which is used to execute a storage device control method based on a pulse neural network as described in the first aspect, wherein the storage device control system based on a pulse neural network includes: a feature perception module, the feature perception module is used to perform multi-dimensional feature perception and collection on a target solid-state hard disk through a multi-layer perceptron to obtain a target hard disk state, wherein the target hard disk state includes multiple operating states of multiple storage units; a state matching module, the state matching module is used to match a first operating state of a first storage unit among the multiple operating states, the first storage unit is any one of the multiple storage units, and the first operating state refers to information when the first storage unit writes a first data set; a feature information acquisition module, the feature information acquisition module is used to read a first predetermined state feature indicator, and based on the first predetermined state feature indicator, The state characteristic index is traversed and matched in the first operating state to obtain the first state characteristic information; a control instruction generation module, the control instruction generation module is used to issue a first control instruction when the first state fitness obtained by analyzing the first state characteristic information does not reach the fitness limit; a control signal processing module, the control signal processing module is used to obtain the first real-time control signal of the first storage unit based on the first control instruction, and encode the first real-time control signal to obtain a first real-time pulse sequence; an optimal control signal determination module, the optimal control signal determination module is used to compare the first real-time pulse sequence with a predetermined pulse sequence to obtain a first comparison result, and analyze the first comparison result to obtain a first optimal control signal; a storage control execution module, the storage control execution module is used to control the target solid-state hard disk to perform storage control on the first data set according to the first optimal control signal.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] The target solid state drive is subjected to multi-dimensional feature perception and collection through a multi-layer perceptron to obtain the target hard disk state, wherein the target hard disk state includes multiple operating states of multiple storage units; the first operating state of the first storage unit is matched in the multiple operating states, wherein the first storage unit is any one of the multiple storage units, and the first operating state refers to the information when the first storage unit writes the first data set; the first predetermined state feature index is read, and the first state feature information is obtained by traversing and matching in the first operating state based on the first predetermined state feature index; when the first state fitness obtained by analyzing the first state feature information does not reach the fitness limit, a first control instruction is issued; based on the first control instruction, a first real-time control signal of the first storage unit is obtained, and the first real-time control signal is encoded to obtain a first real-time pulse sequence; the first real-time pulse sequence is compared with the predetermined pulse sequence to obtain a first comparison result, and the first comparison result is analyzed to obtain a first optimal control signal; according to the first optimal control signal, the target solid state drive is controlled to perform storage control on the first data set. That is, through the multi-layer perceptron and the pulse neural network, the personalization and adaptive adjustment of the control strategy are realized, and the adaptability and flexibility of the control strategy are enhanced.

[0010] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0012] Figure 1 A schematic diagram of a flow chart of a storage device control method based on a pulse neural network in the present application;

[0013] Figure 2 This is a schematic diagram of the structure of a storage device control system based on a pulse neural network in the present application.

[0014] Explanation of the reference numerals: feature perception module 11 , state matching module 12 , feature information acquisition module 13 , control instruction generation module 14 , control signal processing module 15 , optimal control signal determination module 16 , storage control execution module 17 . DETAILED DESCRIPTION

[0015] The present application provides a storage device control method and system based on a pulse neural network, which solves the technical problem in the prior art that the storage control relies on preset algorithms and rules, and the control strategy cannot be dynamically adjusted according to real-time changes, and lacks flexibility and adaptability. Through multi-layer perceptrons and pulse neural networks, the control strategy is personalized and adaptively adjusted, and the adaptability and flexibility of the control strategy are enhanced.

[0016] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0017] For example, please refer to the attached Figure 1 The present application provides a storage device control method based on a pulse neural network, wherein the storage device control method based on a pulse neural network is applied to a storage device control system based on a pulse neural network, and the storage device control method based on a pulse neural network specifically includes the following steps:

[0018] Step 1: Perform multi-dimensional feature perception and collection on the target solid-state hard disk through a multi-layer perceptron to obtain a target hard disk state, wherein the target hard disk state includes multiple operating states of multiple storage units.

[0019] Specifically, a multi-layer perceptron is used to comprehensively monitor and analyze the operating status of the solid-state drive, and to collect multiple operating statuses of each storage unit. A multi-layer perceptron is a deep learning model that consists of multiple layers, including an input layer, one or more hidden layers, and an output layer. Each layer contains multiple neurons and can process complex, nonlinear data relationships, thereby more accurately capturing the overall status of the hard drive. A solid-state drive is a storage device that uses flash memory chips for data storage. Compared with traditional mechanical hard drives, it has faster read and write speeds and higher reliability. Multi-dimensional features refer to data that describes the state of the hard drive from different angles or aspects, such as read and write speeds, temperature, data access patterns, service life, error rates, etc. Through a multi-layer perceptron, the multi-dimensional features of the hard drive are accurately captured, thereby more comprehensively understanding the state of the hard drive.

[0020] Step 2: Match a first operating state of a first storage unit among the multiple operating states, where the first storage unit is any one of the multiple storage units, and the first operating state refers to information when the first storage unit writes a first data set.

[0021] Specifically, from the collected operating states of multiple storage units, identify and locate a specific storage unit and its operating state when writing a specific data set. A specific storage unit is selected from the solid-state drive as the first storage unit, which can be randomly selected or selected according to a certain standard (such as the most frequently used, expected failure, etc.). When the first storage unit performs a write operation, its operating state is recorded, which usually includes the properties when writing the first data set (such as data size, data type), the performance indicators of the write operation (such as write speed, error rate, energy consumption), the state of the unit (such as current health status, temperature), etc. The first data set is a set of related data, which is usually used in data analysis or machine learning. Here, it is a set of data written to the target storage unit. By accurately matching the operating state of a specific storage unit, performance problems or failures can be quickly located, which helps to optimize its performance in a targeted manner.

[0022] Step three: read the first predetermined state characteristic indicator, and traverse and match in the first running state based on the first predetermined state characteristic indicator to obtain first state characteristic information.

[0023] Specifically, the operation state of a specific storage unit is analyzed and understood using predetermined state characteristic indicators. These indicators are searched for in the actual operation state of the storage unit to obtain detailed information about the performance of the storage unit. The first predetermined state characteristic indicator is a predefined key parameter for evaluating the performance of the storage unit, which usually includes response time, I / O throughput, and load. The first response time refers to the response speed of the storage unit to read and write requests; the first I / O throughput refers to the number of input / output operations that the storage unit can process per unit time; the first load refers to the workload being processed by the storage unit, which can be the number of read and write operations or the amount of data. Information matching the predetermined state characteristic indicators is searched in the first operation state, that is, the actual operation state is compared with these indicators to identify the state that meets these characteristics. Through the matching process, the first state characteristic information is obtained, which reveals the performance of the storage unit under specific indicators. By using the predetermined state characteristic indicators, the performance of the storage unit is analyzed, the predetermined indicators are matched, and performance problems or potential faults are quickly diagnosed, which helps to carry out targeted performance adjustment and optimization.

[0024] Step 4: When the first state fitness obtained by analyzing the first state characteristic information does not reach the fitness limit, a first control instruction is issued.

[0025] Specifically, it is determined whether a control instruction needs to be issued to adjust the operating state of the storage unit based on the evaluation result of the state fitness. Based on the performance standard of the solid-state drive, historical data analysis or the needs of a specific application, a fitness limit is pre-set to determine whether the performance of the storage unit meets specific requirements. If the state fitness is lower than this limit, it is considered that the performance of the storage unit is not ideal, and a control instruction needs to be issued to adjust it to improve its performance. If the first state fitness reaches or exceeds the fitness limit, it is considered that the performance of the storage unit meets the requirements and no control instruction needs to be issued. By dynamically adjusting the operation of the storage unit according to its actual performance state, it is ensured that the solid-state drive can operate stably and efficiently. When the analysis of the first state characteristic information shows that the performance of the storage unit does not meet the standard, a control instruction is automatically issued to promote performance improvement.

[0026] Step 5: Acquire a first real-time control signal of the first storage unit based on the first control instruction, and encode the first real-time control signal to obtain a first real-time pulse sequence.

[0027] Specifically, according to the first control instruction, a first real-time control signal is obtained from the control system of the storage unit. This signal is an instruction to perform a specific operation on the storage unit, such as writing, reading or erasing data. The first real-time control signal is encoded using a pulse neural network, and a coding scheme is determined to map different states of the control signal to different modes of the pulse sequence. A series of pulse modes, such as single pulse, double pulse, pulse interval, etc., are defined to represent different control instructions. According to the determined coding scheme, the first real-time control signal is converted into a pulse sequence. Each state in the control signal is mapped to a corresponding pulse mode, and a series of pulses are generated according to the coding rules to form a first real-time pulse sequence. The encoded pulse sequence is recorded as the first real-time pulse sequence. Spiking neural network (SNN) is an artificial intelligence algorithm that imitates the transmission and operation of neural pulse signals. It uses pulses (or neuron discharges) to transmit information and has the characteristics of strong learning ability and low energy consumption. Neurons transmit information through pulses (spikes), and these pulses represent the moment of neuronal activity. By encoding the control signal as a pulse sequence, these control instructions can be executed more efficiently in the SNN, thereby optimizing the performance and efficiency of the storage system. This encoding method is particularly useful in spiking neural networks because SNNs use pulse signals to transmit information instead of continuous signals in traditional computers. The encoding process converts the control signal into a pulse form suitable for SNN processing. For example, the parameters of the control signal are converted into a pulse sequence through a specific encoding algorithm. The generated first real-time pulse sequence is used to guide the operation of the storage unit. In SNNs, these pulse sequences are used to control the behavior of the storage unit, such as read and write operations, data allocation, etc. By encoding the control signal into a pulse sequence, precise control of the storage unit is achieved.

[0028] Step six: Compare the first real-time pulse sequence with the predetermined pulse sequence to obtain a first comparison result, and analyze the first comparison result to obtain a first optimal control signal.

[0029] Specifically, the first real-time pulse sequence obtained by encoding is compared and analyzed with the predetermined pulse sequence to identify the difference between the two and obtain the first comparison result. An ideal pulse sequence, i.e., a predetermined pulse sequence, is set based on a certain optimization goal or performance indicator, including predetermined power consumption, predetermined read and write speed, etc. According to the results of the comparative analysis, it is determined whether the pulse sequences are consistent. If they are consistent, the current control signal is used as the first optimal control signal. If they are inconsistent, it indicates that the current control signal is not optimal, and an optimization process is required to adjust the control signal and generate a new real-time pulse sequence until it is consistent with the predetermined pulse sequence, and the current control signal is used as the first optimal control signal. By comparing the first real-time pulse sequence with the predetermined pulse sequence and analyzing the comparison results, the first optimal control signal is obtained, which helps to optimize the control strategy and ensure that the control signal is always in the optimal state.

[0030] Step seven: According to the first optimal control signal, control the target solid state drive to perform storage control on the first data set.

[0031] Specifically, the data access frequency is obtained according to the data access record in the first data set, and the data in the data set is sorted according to the access frequency, with the data with high access frequency being placed in front. According to the determined first optimal control signal, the target solid-state drive is controlled to store the data set according to the sorted data sequence, and data with high access frequency is stored preferentially. By determining the priority of the data according to its historical access frequency, the storage and access strategy of the solid-state drive is optimized, which helps to improve the efficiency of data access, reduce latency, and ensure that the most important data is processed first.

[0032] Further, step three of this application includes:

[0033] The first predetermined state characteristic indicators include a first response time, a first I / O throughput and a first load.

[0034] Specifically, the first predetermined state characteristic indicator is a key parameter predefined for evaluating the performance of the storage unit, including response time, I / O throughput and load. The first response time refers to the time required from receiving an I / O request to starting to process the request. Response time is an important indicator for measuring the performance of a storage device. A shorter first response time means that the device can respond to I / O requests faster, thereby improving the overall system performance. The first I / O throughput refers to the ability of a storage device to process I / O requests per unit time, usually measured by the number of read and write operations or the amount of data per second. I / O throughput is a key indicator for evaluating the processing capacity of a storage device. A higher first I / O throughput means that the device can process more I / O requests, thereby improving the efficiency of data processing. The first load refers to the degree of use or workload of a storage device within a specific time, usually measured by the frequency of read and write operations or the amount of data. Load is an important indicator for evaluating the working status of a storage device. Understanding the first load helps predict the performance of the device and make adjustments when necessary to avoid overload and performance degradation.

[0035] Further, step 4 of this application includes:

[0036] The first state characteristic information is standardized and normalized to obtain first preprocessing state characteristic information; the first preprocessing state characteristic information is weighted and analyzed using the coefficient of variation principle to obtain a first weight distribution; the first state fitness is obtained by combining the first weight distribution with the first preprocessing state characteristic information.

[0037] Specifically, the first state feature information is standardized and normalized to eliminate the influence of different dimensions. Standardization is a process that makes the data conform to the standard normal distribution, calculates the mean and standard deviation of the data, and then uses statistics to convert the data into a standard form. Next, the standardized data is normalized so that the data falls within the range of [0,1]. Normalization is a process of scaling data to a specific range, and the most common method is to scale the data between 0 and 1. The standardized and normalized data is recorded as the first preprocessing state feature information.

[0038] The coefficient of variation is calculated for the preprocessed feature index to measure its relative degree of dispersion. The importance or stability of each state feature information is evaluated, so as to assign corresponding weights to each feature. The coefficient of variation is the ratio of the standard deviation to the mean value, usually expressed as a percentage. The larger the coefficient of variation, the higher the degree of dispersion of the data and the greater the impact on the overall performance, so a higher weight can be assigned. The preprocessed feature index is multiplied by its corresponding weight, and then these weighted feature information are added to obtain the weighted feature information. The weighted feature information is summarized to obtain a comprehensive first state fitness. State fitness is a quantitative indicator used to evaluate the overall performance and efficiency of the storage unit under a specific operating state. Through standardization and normalization, the consistency and comparability of the data are ensured. The weight calculation is performed using the principle of the coefficient of variation, the importance of each feature is accurately evaluated, and the state fitness is calculated to obtain a quantitative indicator to comprehensively evaluate the performance of the storage unit.

[0039] Furthermore, the present application also includes the following steps:

[0040] Retrieve a first historical read-write record of the first storage unit; analyze the first historical read-write record to obtain a first historical read-write frequency; extract a first historical record in the first historical read-write record, the first historical record including a first historical read-write data volume; obtain a first average historical read-write data volume based on the first historical read-write data volume; weight the first historical read-write frequency and the first average historical read-write data volume to obtain a first wear level; and adjust the first state fitness based on the first wear level.

[0041] Specifically, the historical read and write records of the first storage unit are retrieved from the management system of the storage unit, that is, all read and write operation records in the past period of time, including the number of read and write operations, data volume, operation time and other detailed information. The historical read and write records are analyzed in depth to calculate the frequency of the read and write operations, which is usually obtained by calculating the ratio of the number of read and write operations and the time period. For example, if 1,000 read and write operations are recorded in a month, the first historical read and write frequency may be about 1.4 read and write operations per hour. The read and write frequency is an important indicator to measure the activity level and performance of the storage unit, reflecting the frequency of the read and write operations of the storage unit in a specific time. Specific records are extracted from the first historical read and write records, including the amount of data transferred. All read and write data volumes are added up and then divided by the total number of read and write operations to calculate the average read and write data volume, which measures the average read and write data volume of the storage unit in a specific time period. For example, if the storage unit has performed a total of 1000 read and write operations in the past month, and the accumulated data volume is 500 GB, then the first average historical read and write data volume is 500 GB / 1000=0.5 GB / time.

[0042] According to the impact of the first historical read and write frequency and the first average historical read and write data volume on the wear of the storage unit, appropriate weights are assigned. The weighted read and write frequency and the read and write data volume are combined to obtain the first wear level. According to the calculated first wear level, the first state fitness obtained previously is adjusted. Storage units with high wear levels need to reduce their fitness to reflect their potential performance degradation or shortened life. State fitness was originally used to evaluate the performance of storage units. The wear level indicator is introduced to comprehensively consider the current status of the storage unit, including its health status and performance level. By comprehensively considering the usage history and wear status of the storage unit, its current performance can be accurately evaluated, and understanding the wear degree and performance status of the storage unit will help to formulate performance optimization and maintenance strategies.

[0043] Furthermore, the present application also includes the following steps:

[0044] Determine whether the first wear level is within the wear level limit; if not, issue a second control instruction; perform a comparative analysis of the wear performance of the multiple storage units based on the second control instruction to obtain a second storage unit; and use the second storage unit instead of the first storage unit to store the first data set.

[0045] Specifically, based on the design life, reliability standard or maintenance strategy of the solid-state drive, a wear level limit is pre-set to determine whether the storage unit is over-worn. The calculated first wear level is compared with the wear level limit. If the first wear level exceeds the limit, it indicates that the storage unit has been over-worn; if it does not exceed the limit, it is considered that the storage unit is still within the acceptable range of use and has not yet reached the level where action is required. At this time, a second control instruction is issued, such as adjusting the working mode of the storage unit or performing preventive maintenance. According to the second control instruction, a comparative analysis of the wear performance of multiple storage units in the storage system is performed, and then a storage unit (second storage unit) with a lower wear level and better performance is selected to replace the original first storage unit for storing a specific data set. The first data set is stored on the second storage unit to take advantage of its better performance and longer service life. By optimizing the configuration of the storage unit, it helps to improve the overall performance and efficiency of the storage system.

[0046] Furthermore, the present application also includes the following steps:

[0047] Read a second predetermined state characteristic indicator, and traverse and match in the first operating state based on the second predetermined state characteristic indicator to obtain second state characteristic information; weight the second state characteristic information to obtain a first environmental index; adjust the first state fitness according to the first environmental index; wherein the second predetermined state characteristic indicator includes temperature and vibration.

[0048] Specifically, the second predetermined state characteristic indicators, including temperature and vibration, are read to evaluate the performance of the storage unit. Based on these predetermined state characteristic indicators, traversal matching is performed in the first operating state. The actual operating state is compared with these indicators to determine whether the actual value of each indicator meets the predetermined standard or threshold, and the state that meets these characteristics is identified. Through the matching process, the matching results of temperature and vibration are recorded as the second state characteristic information. According to the degree of influence of temperature and vibration on the performance of the storage unit, weights are assigned to them, and the temperature and vibration data are multiplied by their corresponding weights to obtain weighted state characteristic information. The weighted temperature and vibration data are combined to calculate the first environmental index.

[0049] The first environmental index is used to adjust the first state fitness, and the health of the environment in which the storage unit is located (reflected by the first environmental index) is taken into consideration to adjust the previously calculated first state fitness. The state fitness was originally used to evaluate the performance of the storage unit. By introducing the environmental index as an indicator, the condition of the environment in which the storage unit is located can be more comprehensively considered, including its impact on performance. The environmental index is used as an adjustment factor, multiplied or added to the first state fitness, or the fitness is adjusted through a conversion function. In a specific example, it is assumed that the second predetermined state characteristic index: the predetermined temperature is ≤45°C, the predetermined vibration is ≤0.1; the actual temperature is 42°C, and the actual vibration is 0.08; the temperature and vibration obtained by traversal matching meet the predetermined indicators. Assuming that the storage unit is more sensitive to temperature, the weights assigned are 0.8 for temperature and 0.2 for vibration, and the environmental index calculated accordingly is 33.616. By considering the environmental index, the performance of the storage unit is more comprehensively evaluated, not only based on the current state, but also considering the condition of the environment in which it is located, which helps to improve the reliability and efficiency of the storage system, especially when the environmental conditions change greatly.

[0050] Furthermore, step six of this application includes:

[0051] A: Randomly obtain a first control signal; B: Use a write amplification prediction model to predict and analyze the first control power consumption and the first read / write speed in the first control signal to obtain a first predicted write amplification index; C1: When the first predicted write amplification index meets the write amplification limit, use the first control signal as the first optimal control signal; C2: When the first predicted write amplification index does not meet the write amplification limit, repeat the iteration until the first optimal control signal is obtained.

[0052] Specifically, a signal for controlling the operation of a storage device is randomly selected as a first control signal, including parameters such as read / write speed and power consumption. Historical data is collected, including write amplification index, control signal parameters (such as read / write speed, power consumption, etc.) and corresponding performance and life data. Features related to the write amplification index are extracted from the collected data, including read / write speed, power consumption, type of storage unit, operation frequency, etc. The extracted data set is divided into a training set and a validation set. A prediction model, such as a machine learning model, is established to analyze the relationship between historical data and control signal parameters. The model is trained using the training set, and the model continuously learns how to predict the write amplification index from the input features. The parameters of the model are adjusted to optimize the prediction performance. After the training is completed, the model is verified and tested using the data of the validation set to evaluate the accuracy and generalization ability of the model to ensure that the model can accurately predict unseen data. According to the results of the verification and testing, the model is optimized and adjusted, including adjusting the structure, parameters or feature extraction method of the model. Using the established model, a first predicted write amplification index is obtained by inputting the first control power consumption and the first read / write speed for prediction analysis.

[0053] A threshold value, i.e., the maximum allowable value of the write amplification index, is set in advance to determine whether the write amplification index meets specific performance requirements. If the first predicted write amplification index meets the write amplification limit, it indicates that the configuration of the control signal is valid, and the first control signal is used as the first optimal control signal. If the first predicted write amplification index does not meet the write amplification limit, i.e., it is higher than the write amplification limit, it indicates that the configuration of the control signal may not be ideal, and the control signal needs to be further optimized. Repeat steps A and B until a control signal whose first predicted write amplification index meets the write amplification limit is obtained, and it is used as the first optimal control signal. By randomly acquiring control signals, analyzing them using a write amplification prediction model, and iteratively optimizing control signals, precise control of solid-state drives can be achieved, which helps to improve the performance and life of the target solid-state drives, optimize the performance of the storage system, and improve overall storage efficiency.

[0054] Further, step seven of this application includes:

[0055] Retrieve a first historical access record of the first data in the first data set; analyze the first historical access record to obtain a first historical access frequency; prioritize the first data based on the first historical access frequency to obtain a first priority data sequence of the first data set; and control the target solid-state drive to perform priority storage control on the first priority data sequence according to the first optimal control signal.

[0056] Specifically, retrieve the historical access records of the first data set from the storage management system, including the access time, access method, etc. Analyze the first historical access records, calculate the historical access frequency of the first data, and measure the number of times the first data is accessed within a specific time period. Prioritize the data according to the historical access frequency of the first data to determine which data should be given a higher priority in storage and management. The higher the data is ranked, the higher its access frequency and the higher its priority. Generate a first priority data sequence based on the priority sorting result. Perform storage control on the target solid-state drive based on the determined first optimal control signal, and prioritize data with high access frequency based on the first priority data sequence. By preferentially storing data with high access frequency, storage resources are effectively utilized and storage efficiency is improved.

[0057] In summary, the storage device control method based on pulse neural network provided by the present application has the following technical effects:

[0058] The target solid state drive is subjected to multi-dimensional feature perception and collection through a multi-layer perceptron to obtain the target hard disk state, wherein the target hard disk state includes multiple operating states of multiple storage units; the first operating state of the first storage unit is matched in the multiple operating states, wherein the first storage unit is any one of the multiple storage units, and the first operating state refers to the information when the first storage unit writes the first data set; the first predetermined state feature index is read, and the first state feature information is obtained by traversing and matching in the first operating state based on the first predetermined state feature index; when the first state fitness obtained by analyzing the first state feature information does not reach the fitness limit, a first control instruction is issued; based on the first control instruction, a first real-time control signal of the first storage unit is obtained, and the first real-time control signal is encoded to obtain a first real-time pulse sequence; the first real-time pulse sequence is compared with the predetermined pulse sequence to obtain a first comparison result, and the first comparison result is analyzed to obtain a first optimal control signal; according to the first optimal control signal, the target solid state drive is controlled to perform storage control on the first data set. That is, through the multi-layer perceptron and the pulse neural network, the personalization and adaptive adjustment of the control strategy are realized, and the adaptability and flexibility of the control strategy are enhanced.

[0059] Embodiment 2: Based on the same inventive concept as the storage device control method based on pulse neural network in the aforementioned embodiment 1, the present application also provides a storage device control system based on pulse neural network, please refer to the attached Figure 2 , the storage device control system based on pulse neural network includes:

[0060] The feature perception module 11 is used to perform multi-dimensional feature perception and collection on the target solid state drive through a multi-layer perceptron to obtain a target hard disk state, wherein the target hard disk state includes multiple operating states of multiple storage units.

[0061] The state matching module 12 is used to match the first operating state of the first storage unit among the multiple operating states, the first storage unit is any one of the multiple storage units, and the first operating state refers to the information when the first storage unit writes the first data set.

[0062] The characteristic information acquisition module 13 is used to read the first predetermined state characteristic indicator, and traverse and match in the first operating state based on the first predetermined state characteristic indicator to obtain the first state characteristic information.

[0063] The control instruction generating module 14 is used to issue a first control instruction when the first state fitness obtained by analyzing the first state characteristic information does not reach the fitness limit.

[0064] The control signal processing module 15 is used to obtain a first real-time control signal of the first storage unit based on the first control instruction, and encode the first real-time control signal to obtain a first real-time pulse sequence.

[0065] The optimal control signal determination module 16 is used to compare the first real-time pulse sequence with the predetermined pulse sequence to obtain a first comparison result, and analyze the first comparison result to obtain a first optimal control signal.

[0066] The storage control execution module 17 is used to control the target solid state drive to perform storage control on the first data set according to the first optimal control signal.

[0067] Furthermore, the characteristic information acquisition module 13 in the storage device control system based on the pulse neural network is also used for:

[0068] The first predetermined state characteristic indicators include a first response time, a first I / O throughput and a first load.

[0069] Furthermore, the control instruction generating module 14 in the storage device control system based on pulse neural network is also used for:

[0070] The first state characteristic information is standardized and normalized to obtain first preprocessing state characteristic information; the first preprocessing state characteristic information is weighted and analyzed using the coefficient of variation principle to obtain a first weight distribution; the first state fitness is obtained by combining the first weight distribution with the first preprocessing state characteristic information.

[0071] Furthermore, the storage device control system based on pulse neural network also includes a fitness adjustment module, which is used to:

[0072] Retrieve a first historical read-write record of the first storage unit; analyze the first historical read-write record to obtain a first historical read-write frequency; extract a first historical record in the first historical read-write record, the first historical record including a first historical read-write data volume; obtain a first average historical read-write data volume based on the first historical read-write data volume; weight the first historical read-write frequency and the first average historical read-write data volume to obtain a first wear level; and adjust the first state fitness based on the first wear level.

[0073] Furthermore, the storage device control system based on pulse neural network also includes a second instruction generation module, which is used to:

[0074] Determine whether the first wear level is within the wear level limit; if not, issue a second control instruction; perform a comparative analysis of the wear performance of the multiple storage units based on the second control instruction to obtain a second storage unit; and use the second storage unit instead of the first storage unit to store the first data set.

[0075] Furthermore, the storage device control system based on pulse neural network also includes a second feature acquisition module, which is used to:

[0076] Read a second predetermined state characteristic indicator, and traverse and match in the first operating state based on the second predetermined state characteristic indicator to obtain second state characteristic information; weight the second state characteristic information to obtain a first environmental index; adjust the first state fitness according to the first environmental index; wherein the second predetermined state characteristic indicator includes temperature and vibration.

[0077] Furthermore, the optimal control signal determination module 16 in the storage device control system based on the pulse neural network is also used for:

[0078] A: Randomly obtain a first control signal; B: Use a write amplification prediction model to predict and analyze the first control power consumption and the first read / write speed in the first control signal to obtain a first predicted write amplification index; C1: When the first predicted write amplification index meets the write amplification limit, use the first control signal as the first optimal control signal; C2: When the first predicted write amplification index does not meet the write amplification limit, repeat the iteration until the first optimal control signal is obtained.

[0079] Furthermore, the storage control execution module 17 in the storage device control system based on the pulse neural network is also used for:

[0080] Retrieve a first historical access record of the first data in the first data set; analyze the first historical access record to obtain a first historical access frequency; prioritize the first data based on the first historical access frequency to obtain a first priority data sequence of the first data set; and control the target solid-state drive to perform priority storage control on the first priority data sequence according to the first optimal control signal.

[0081] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. Figure 1 The storage device control method based on pulse neural network and the specific examples in the first embodiment are also applicable to the storage device control system based on pulse neural network in the present embodiment. Through the above detailed description of the storage device control method based on pulse neural network, those skilled in the art can clearly know the storage device control system based on pulse neural network in the present embodiment, so for the sake of brevity of the specification, it will not be described in detail here. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0082] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0083] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A storage device control method based on pulse neural network, characterized in that: include: Performing multi-dimensional feature perception and collection on a target solid-state hard disk through a multi-layer perceptron to obtain a target hard disk state, wherein the target hard disk state includes multiple operating states of multiple storage units; Matching a first operating state of a first storage unit in the multiple operating states, the first storage unit being any one of the multiple storage units, the first operating state referring to information when the first storage unit writes a first data set; Reading a first predetermined state characteristic indicator, and traversing and matching in the first running state based on the first predetermined state characteristic indicator to obtain first state characteristic information; When the first state fitness obtained by analyzing the first state characteristic information does not reach the fitness limit, issuing a first control instruction; Acquire a first real-time control signal of the first storage unit based on the first control instruction, and encode the first real-time control signal to obtain a first real-time pulse sequence; Comparing the first real-time pulse sequence with the predetermined pulse sequence to obtain a first comparison result, and analyzing the first comparison result to obtain a first optimal control signal; Controlling the target solid state drive to perform storage control on the first data set according to the first optimal control signal; When the first state fitness obtained by analyzing the first state characteristic information does not reach the fitness limit, issuing a first control instruction includes: Standardizing and normalizing the first state characteristic information to obtain first preprocessed state characteristic information; Using the coefficient of variation principle to perform weight calculation and analysis on the first preprocessing state feature information to obtain a first weight distribution; Combining the first weight distribution with the first preprocessing state characteristic information to obtain the first state fitness; Comparing the first real-time pulse sequence with the predetermined pulse sequence to obtain a first comparison result, and analyzing the first comparison result to obtain a first optimal control signal, including: A: Randomly obtain the first control signal; B: predicting and analyzing the first control power consumption and the first read / write speed in the first control signal by using a write amplification prediction model to obtain a first predicted write amplification index; C1: when the first predicted write amplification index meets the write amplification limit, using the first control signal as the first optimal control signal; C2: when the first predicted write amplification index does not meet the write amplification limit, repeating the iteration until the first optimal control signal is obtained; Controlling the target solid state drive to perform storage control on the first data set according to the first optimal control signal includes: Retrieving a first historical access record of the first data in the first data set; Analyzing the first historical access record to obtain a first historical access frequency; Prioritizing the first data based on the first historical access frequency as a constraint to obtain a first priority data sequence of the first data set; According to the first optimal control signal, the target solid state drive is controlled to perform priority storage control on the first priority data sequence.

2. According to claim 1, a storage device control method based on pulse neural network is characterized in that: The first predetermined state characteristic indicators include a first response time, a first I / O throughput and a first load.

3. The storage device control method based on pulse neural network according to claim 1, characterized in that: Also includes: Retrieving a first historical read-write record of the first storage unit; Analyzing the first historical read-write record to obtain a first historical read-write frequency; Extracting a first historical record in the first historical read-write record, wherein the first historical record includes a first historical read-write data volume; Obtaining a first average historical read-write data volume according to the first historical read-write data volume; Weighting the first historical read-write frequency and the first average historical read-write data volume to obtain a first wear level; The first state fitness is adjusted according to the first wear level.

4. The storage device control method based on pulse neural network according to claim 3, characterized in that: Also includes: determining whether the first wear level is within a wear level limit; If not, issue a second control instruction; Performing a comparative analysis on the wear performance of the plurality of storage units based on the second control instruction to obtain a second storage unit; The first data set is stored in the second storage unit instead of the first storage unit.

5. The storage device control method based on pulse neural network according to claim 1, characterized in that: include: Reading a second predetermined state characteristic indicator, and traversing and matching in the first running state based on the second predetermined state characteristic indicator to obtain second state characteristic information; weighting the second state feature information to obtain a first environment index; Adjusting the first state fitness according to the first environmental index; Wherein, the second predetermined state characteristic index includes temperature and vibration.

6. A storage device control system based on a pulse neural network, characterized in that: Steps for implementing a storage device control method based on a pulse neural network as described in any one of claims 1 to 5, wherein the storage device control system based on a pulse neural network comprises: A feature perception module, the feature perception module is used to perform multi-dimensional feature perception and collection on the target solid-state hard disk through a multi-layer perceptron to obtain a target hard disk state, wherein the target hard disk state includes multiple operating states of multiple storage units; a state matching module, the state matching module being used to match a first operating state of a first storage unit among the multiple operating states, the first storage unit being any one of the multiple storage units, and the first operating state being information when the first storage unit writes a first data set; A characteristic information acquisition module, the characteristic information acquisition module is used to read a first predetermined state characteristic indicator, and traverse and match in the first operating state based on the first predetermined state characteristic indicator to obtain first state characteristic information; A control instruction generation module, wherein the control instruction generation module is used to issue a first control instruction when the first state fitness obtained by analyzing the first state characteristic information does not reach the fitness limit; a control signal processing module, the control signal processing module being used to obtain a first real-time control signal of the first storage unit based on the first control instruction, and to encode the first real-time control signal to obtain a first real-time pulse sequence; an optimal control signal determination module, the optimal control signal determination module being used to compare the first real-time pulse sequence with a predetermined pulse sequence to obtain a first comparison result, and to analyze the first comparison result to obtain a first optimal control signal; A storage control execution module is used to control the target solid-state drive to perform storage control on the first data set according to the first optimal control signal.

Images