Performance optimization method and device of solid state disk, electronic equipment, medium and product

Abstract

The invention provides a solid state disk performance optimization method and device, electronic equipment, a medium and a product. The method comprises the steps of obtaining historical operation data of a solid state disk; inputting the test data set of the historical operation data into the target network model to obtain a prediction command number output by the target network model; determining a bandwidth proportion according to the predicted command number and a command number threshold value; wherein the command number threshold value is a preset number; determining a performance optimization result of the solid state disk according to the bandwidth proportion and the dynamic optimization rule; wherein the dynamic optimization rule is a preset rule used for adjusting the bandwidth proportion. According to the technical scheme, based on the target network model obtained through training of the training data set, the verification data set and the long-short-term neural network model of the solid-state disk, the predicted command number for the solid-state disk is output, the bandwidth proportion of the solid-state disk is determined based on the predicted command number, and the performance of the solid-state disk is optimized in real time according to the bandwidth proportion. And precise control of the solid state disk is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of hard disk optimization, and in particular to a method, device, electronic equipment, medium and product for optimizing the performance of a solid state hard disk. Background Art

[0002] With the advent of the big data era, the demand for data analysis and processing has spawned a series of big data processing systems. Big data processing systems require the support of storage technology. Big data applications have posed greater challenges to the performance and reliability of storage systems in all aspects.

[0003] Currently, in the prior art, for the evaluation and optimization of storage performance, a traditional write arbitration control algorithm is usually used to evaluate the storage performance of a hard disk. However, the performance control capability of the traditional write arbitration control algorithm fluctuates greatly under different hardware configurations, and it is impossible to accurately control and optimize hardware with different performance.

[0004] Therefore, there is an urgent need for a solid state drive performance optimization method to accurately evaluate and optimize hard drive performance. Summary of the invention

[0005] The present invention provides a performance optimization method, device, electronic device, medium and product for a solid-state hard disk, which are used to solve the defect that the storage performance of a hard disk is evaluated by a traditional write arbitration control algorithm in the prior art, and it is impossible to accurately control and optimize hardware with different performances. The present invention implements a target network model obtained by training a training data set, a verification data set and a long-term and short-term neural network model based on the solid-state hard disk, outputs the predicted number of commands for the solid-state hard disk, determines the bandwidth ratio of the solid-state hard disk based on the predicted number of commands, thereby optimizing the performance of the solid-state hard disk in real time based on the bandwidth ratio, and achieving precise control of the solid-state hard disk.

[0006] The present invention provides a method for optimizing the performance of a solid state hard disk, comprising the following steps.

[0007] Get the historical operation data of the solid state drive.

[0008] The test data set of the historical operation data is input into the target network model to obtain the predicted number of commands output by the target network model; wherein the target network model is trained based on the training data set, the verification data set and the long-short term neural network model of the historical operation data; the predicted number of commands is used to evaluate the performance of the solid state drive.

[0009] The bandwidth ratio is determined according to the predicted command quantity and the command quantity threshold, wherein the command quantity threshold is a preset quantity; the bandwidth ratio is used to determine whether the performance of the solid state drive needs to be optimized and adjusted.

[0010] The performance optimization result of the solid state drive is determined according to the bandwidth ratio and the dynamic optimization rule; wherein the dynamic optimization rule is a pre-set rule for adjusting the bandwidth ratio.

[0011] According to a performance optimization method for a solid-state hard disk provided by the present invention, historical operation data of the solid-state hard disk is obtained, including: obtaining initial sample operation data of the solid-state hard disk; cleaning the initial sample operation data for outliers and / or missing values ​​to obtain target sample operation data; normalizing the target sample operation data to obtain historical operation data.

[0012] According to a performance optimization method for a solid-state hard disk provided by the present invention, historical operation data includes a training data set, a verification data set and a test data set; after obtaining the historical operation data of the solid-state hard disk, it also includes: determining the training data set, the verification data set and the test data set according to the historical operation data and a data partitioning rule; wherein the data partitioning rule is a pre-set partitioning ratio for partitioning the training data set, the verification data set and the test data set.

[0013] According to a performance optimization method for a solid-state hard disk provided by the present invention, the training process of a target network model is as follows: a time series forward arrangement training data set and a time series reverse arrangement training data set are determined according to a training data set; the time series forward arrangement training data set and the time series reverse arrangement training data set are respectively input into a long-term and short-term neural network model to obtain a bidirectional learning network model; the bidirectional learning network model is trained according to a preset number of training times and error accuracy to obtain an initial network model; a verification data set is input into the initial network model for model optimization to obtain a target network model.

[0014] According to a performance optimization method for a solid-state hard disk provided by the present invention, a bandwidth ratio is determined according to a predicted command quantity and a command quantity threshold, including: obtaining a command write data volume and a command reclaim data volume; obtaining a data volume ratio according to a ratio of a command write data volume to a command reclaim data volume; determining a volume difference according to a difference between the predicted command quantity and the command quantity threshold; and determining a bandwidth ratio according to the data volume ratio and the volume difference.

[0015] According to a performance optimization method for a solid-state hard disk provided by the present invention, a performance optimization result of the solid-state hard disk is determined according to a bandwidth ratio and a dynamic optimization rule, including: determining a preset ratio; wherein the preset ratio is a ratio of a preset write bandwidth expectation to a preset recycle bandwidth expectation; determining a relationship between the preset ratio and the bandwidth ratio; adjusting the bandwidth ratio according to the relationship between the preset ratio and the bandwidth ratio and the dynamic optimization rule to determine a performance optimization result.

[0016] The present invention also provides a performance optimization device for a solid state hard disk, comprising the following modules.

[0017] The data acquisition module is used to obtain the historical operation data of the solid state drive.

[0018] The quantity prediction module is used to input the test data set of historical operation data into the target network model to obtain the predicted number of commands output by the target network model; wherein the target network model is trained based on the training data set, verification data set and long-short term neural network model of the historical operation data; the predicted number of commands is used to evaluate the performance of the solid state drive.

[0019] The ratio determination module is used to determine the bandwidth ratio according to the predicted command quantity and the command quantity threshold; wherein the command quantity threshold is a preset quantity; the bandwidth ratio is used to determine whether the performance of the solid state drive needs to be optimized and adjusted.

[0020] The result determination module is used to determine the performance optimization result of the solid state drive according to the bandwidth ratio and the dynamic optimization rule.

[0021] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, any of the above-mentioned methods for optimizing the performance of a solid state drive is implemented.

[0022] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the performance optimization method of any of the above-mentioned solid-state hard disks is implemented.

[0023] The present invention also provides a computer program product, including a computer program, which implements any of the above-mentioned methods for optimizing the performance of a solid state drive when the computer program is executed by a processor.

[0024] The present invention provides a method, device, electronic device, medium and product for optimizing the performance of a solid-state hard disk, which obtains the historical operation data of the solid-state hard disk; inputs the test data set of the historical operation data into the target network model to obtain the predicted number of commands output by the target network model; wherein the target network model is obtained by training the training data set, the verification data set and the long-short term neural network model based on the historical operation data; determines the bandwidth ratio according to the predicted number of commands and the command number threshold; wherein the command number threshold is a preset number; determines the performance optimization result of the solid-state hard disk according to the bandwidth ratio and the dynamic optimization rule; wherein the dynamic optimization rule is a preset rule for adjusting the bandwidth ratio. The technical scheme of the present invention is used to solve the defect that the storage performance of the hard disk is evaluated by the traditional write arbitration control algorithm in the prior art, and it is impossible to accurately control and optimize the hardware with different performances, and realizes the target network model obtained by training the training data set, the verification data set and the long-short term neural network model based on the solid-state hard disk, outputs the predicted number of commands for the solid-state hard disk, and determines the bandwidth ratio of the solid-state hard disk based on the predicted number of commands, thereby optimizing the performance of the solid-state hard disk in real time based on the bandwidth ratio, and achieving accurate control of the solid-state hard disk. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0026] Figure 1 It is a flow chart of the performance optimization method of the solid state drive provided by the present invention.

[0027] Figure 2 It is a schematic diagram of the model structure provided by the present invention.

[0028] Figure 3 It is a structural schematic diagram of the performance optimization device for a solid state hard disk provided by the present invention.

[0029] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] Combine the following Figure 1-Figure 2 The performance optimization method of the solid-state hard disk provided by the present invention is described. The performance optimization method of the solid-state hard disk provided by the present invention can be applicable to the steady-state performance optimization of the solid-state hard disk. The executor of the method can be an electronic device or a performance optimization device of the solid-state hard disk arranged in the electronic device. The performance optimization device of the solid-state hard disk can be implemented by software, hardware or a combination of both. Figure 1 FIG. 1 is a flow chart of a method for optimizing the performance of a solid state drive provided by the present invention. Figure 1 As shown, the method includes the following steps 101, 102, 103 and 104.

[0032] Step 101: Obtain historical operation data of the solid state drive.

[0033] In this step, the solid state hard disk may be a hard disk of the same batch and the same type produced by the same manufacturer, which is not limited in this embodiment.

[0034] The historical operation data is data updated and counted during the historical operation of the solid-state drive, such as time data, command data, block size data, write scale, and whether the current garbage collection (GC) has been started; among them, the time data refers to the time period data of 10 intensity sampling, and the command data refers to the command mode issued by the upper computer, which is divided into continuous writing and random writing. This embodiment does not limit this.

[0035] In a specific embodiment, the specific implementation method of obtaining historical operation data of a solid-state drive includes: obtaining initial sample operation data of the solid-state drive; cleaning the initial sample operation data for outliers and / or missing values ​​to obtain target sample operation data; normalizing the target sample operation data to obtain historical operation data.

[0036] In this step, the initial sample operation data is the initial data updated and counted during the historical operation of the solid state drive.

[0037] Specifically, the initial sample operation data of the solid-state drive that is updated and counted during historical operation is obtained. After obtaining the initial sample operation data, the initial sample operation data is first cleaned of outliers and / or missing values ​​to reduce the impact of abnormal data on subsequent model training, thereby obtaining target sample operation data; and then the target sample operation data is further normalized so that the target sample operation data has a unified data specification, thereby obtaining the final historical operation data.

[0038] Exemplarily, the target sample running data is normalized, and the normalization function is as follows: ,in, is the sample feature of the historical operation data collected at time t, is the minimum value of the sample feature collected, is the maximum value of the sample features collected.

[0039] The advantage of this setting is that the final historical operation data can be obtained after processing the initial sample operation data, which improves the accuracy of subsequent model training and model application and reduces the impact of one data on subsequent use.

[0040] In a specific embodiment, the historical operation data includes a training data set, a verification data set and a test data set; after obtaining the historical operation data of the solid-state drive, it also includes: determining the training data set, the verification data set and the test data set based on the historical operation data and data partitioning rules; wherein the data partitioning rules are pre-set division ratios for dividing the training data set, the verification data set and the test data set.

[0041] In this step, the training data set is used to train the subsequent target network model; the validation data set is used to verify the trained target network model and determine the optimization performance of the trained target network model; the test data set is used to input the target network model for data testing.

[0042] The data partitioning rule is a pre-set partitioning ratio for partitioning the training data set, the verification data set, and the test data set. For example, the collected historical operation data may be randomly divided into three parts, such as 60% as the training data set, 20% as the verification data set, and 20% as the test data set. This embodiment does not limit this.

[0043] Specifically, after the historical operation data of the solid state drive is acquired, the acquired historical operation data is randomly divided into three parts according to the historical operation data and the data division rule.

[0044] Exemplarily, when 100,080 pieces of historical operation data are obtained, the 100,080 pieces of historical operation data are divided according to a data division rule such as 60% as a training data set, 20% as a verification data set, and 20% as a test data set.

[0045] Step 102: Input the test data set of historical operation data into the target network model to obtain the number of predicted commands output by the target network model.

[0046] In this step, the target network model is obtained by training a training data set, a validation data set and a long-short term neural network model based on historical operation data; the predicted number of commands is used to evaluate the performance of the solid state drive.

[0047] Specifically, after acquiring the historical operation data, a test data set of the historical operation data is input into the target network model, and the input historical operation data is processed by the target network model to obtain the predicted command data volume output by the target network model.

[0048] In a specific implementation, the training process of the target network model is as follows: determine a time series forward arrangement training data set and a time series reverse arrangement training data set based on a training data set; input the time series forward arrangement training data set and the time series reverse arrangement training data set into a long-term and short-term neural network model respectively to obtain a bidirectional learning network model; train the bidirectional learning network model according to a preset number of training times and error accuracy to obtain an initial network model; input a verification data set into the initial network model for model optimization to obtain a target network model.

[0049] In this step, the long-term and short-term neural network model is a traditional long-term and short-term neural network system. Figure 2 It is a schematic diagram of the model structure provided by the present invention, such as Figure 2As shown in the figure, the input of the long-term and short-term neural network model includes X1, X2, X3, ...Xn, etc. X1, X2, X3, ...Xn represent the number of historical running data input. The long-term and short-term neural network model is composed of input gate, forget gate and output gate, in which each neuron has a memory unit and three gate structures: input, output and forget. This network controls the flow of information through these gates and can save and recall long-term information between input data. The target network model is improved on the basis of the traditional long-term and short-term neural network model: it is a fusion of the forward long-term and short-term memory network and the reverse long-term and short-term memory neural network. The improved target network model takes into account the past and future features of the sequence, uses two long-term and short-term memory networks, a forward input sequence, and a reverse input sequence, and then combines the outputs of the two according to certain weights as the final result. The hidden layer of the improved model consists of two parts: the forward cell state and the reverse cell state. The flow sequence enters the hidden layer through the input layer to participate in the forward calculation and reverse calculation respectively. The final output result is obtained by the output layer fusing the output result of the forward long-term and short-term memory network and the output result of the reverse long-term and short-term memory network according to certain weights.

[0050] The preset number of training times is the preset number of training times for the target network model, for example, it can be 60 times, 80 times, 100 times, etc., which is not limited in this embodiment. The error accuracy is the accuracy preset for evaluating whether the training meets the preset error requirement. The preset error requirement can be, for example, the mean absolute error ( ) and the root mean square error ( ) as a measurement indicator, is the average absolute error between the predicted and observed values, It is the expected value of the square of the difference between the predicted value and the true value. The smaller the values ​​of these two errors are, the better the performance of the model is. and The expression is as follows: , , where m represents the total number of sampling times for historical operation data, i represents the i-th sampling, represents the expected number of prediction commands for the i-th time, Represents the actual number of prediction commands for the ith time. If the preset error requirement set by the training goal is reached, the results are output and the target network model is determined. If the preset error requirement set by the training goal is not reached, the preset number of training times is increased or the training goal is redefined. When the predetermined training goal is reached or the number of training times is maximized, the network model training is completed. After many experiments, the best improved long short-term memory network model is obtained to determine the target network model. The unknown number of prediction commands is predicted through a test data set of known historical operating data of the solid-state drive. After a large number of repeated experiments, the recorded results are analyzed and studied. The number of prediction commands is obtained by multiplying the output results of the forward training network and the reverse training network by the weighted matrix fusion.

[0051] Specifically, first, a time series forward arrangement training data set and a time series reverse arrangement training data set are determined according to the training data set, and the time series forward arrangement training data set and the time series reverse arrangement training data set are respectively input into the long-term and short-term neural network model to strengthen and highlight the overall characteristics of the logical address of the long-term and short-term neural network model to obtain a two-way learning network model, and then the two-way learning network model is further trained according to the preset number of training times and error accuracy to obtain an initial network model, and finally, the verification data set is input into the initial network model for model optimization, and the parameters and count values ​​are updated to obtain the target network model.

[0052] Furthermore, in the process of processing the input historical operation data through the target network model to obtain the predicted command data volume output by the target network model, considering the ideal situation that all blocks have no bad blocks and the GC write and host write are both at a uniform speed, the data volume of the GC recovered source super data block (data block bound according to 32 data block units) is recorded as , the amount of data in a super block excluding parity is recorded as Then, the time it takes for GC to produce an idle super block is equal to the time it takes for GC and the host to write a super block. During this period, the amount of data recovered by GC is , and the amount of data written by the host is , which means that the bandwidth ratio of host write bandwidth to GC write bandwidth should be equal to . Since GC will start when the number of free super blocks is less than a certain threshold, we hope that the number of free super blocks will be stable at a value less than this threshold to ensure that the host write does not time out and the GC does not exit early. Therefore, there needs to be a mechanism to make the host write bandwidth ratio larger than the ideal steady state when there are more free super blocks, and the GC write bandwidth ratio smaller than the ideal steady state; and vice versa.

[0053] Step 103: Determine the bandwidth ratio according to the predicted command quantity and the command quantity threshold.

[0054] In this step, the command quantity threshold is a preset quantity; and the bandwidth ratio is used to determine whether it is necessary to optimize and adjust the performance of the solid state drive.

[0055] In a specific embodiment, the implementation method of determining the bandwidth ratio based on the predicted command quantity and the command quantity threshold includes: obtaining the command write data volume and the command retrieved data volume; obtaining the data volume ratio based on the ratio of the command write data volume and the command retrieved data volume; determining the quantity difference based on the difference between the predicted command quantity and the command quantity threshold; determining the bandwidth ratio based on the data volume ratio and the quantity difference.

[0056] In this step, the command quantity threshold is a preset value used to evaluate the predicted command quantity, which is not limited in this embodiment.

[0057] Specifically, the bandwidth ratio refers to the ratio of the expected write bandwidth to the expected recycle bandwidth. , To write bandwidth expectations, To recover bandwidth expectations, The amount of data written for the command, The amount of data recovered for the command, To predict the number of commands, is the command quantity threshold, is a constant coefficient.

[0058] Step 104: Determine the performance optimization result of the solid state drive according to the bandwidth ratio and the dynamic optimization rule.

[0059] In this step, the dynamic optimization rule is a pre-set rule for adjusting the bandwidth ratio.

[0060] In a specific embodiment, the specific implementation method of determining the performance optimization result of the solid-state drive according to the bandwidth ratio and dynamic optimization rules includes: determining a preset ratio; wherein the preset ratio is a ratio of a preset write bandwidth expectation to a preset recovery bandwidth expectation; determining the relationship between the preset ratio and the bandwidth ratio; adjusting the bandwidth ratio according to the relationship between the preset ratio and the bandwidth ratio and the dynamic optimization rules to determine the performance optimization result.

[0061] Specifically, a preset ratio is determined; wherein the preset ratio is a ratio of a preset write bandwidth expectation to a preset recovery bandwidth expectation; the relationship between the preset ratio and the bandwidth ratio is determined; the bandwidth ratio is adjusted according to the relationship between the preset ratio and the bandwidth ratio and a dynamic optimization rule to determine a performance optimization result, and when the bandwidth ratio is less than or equal to the preset ratio, the bandwidth ratio is adjusted according to a first dynamic adjustment rule to determine a performance optimization result; wherein the first dynamic adjustment rule is to lower the write bandwidth expectation and increase the recovery bandwidth expectation; when the bandwidth ratio is greater than the preset ratio, the bandwidth ratio is adjusted according to a second dynamic adjustment rule to determine a performance optimization result; wherein the second dynamic adjustment rule is to increase the recovery bandwidth expectation and lower the write bandwidth expectation.

[0062] The advantage of this setting is that it can dynamically adjust the bandwidth ratio in real time to improve the steady-state performance of the solid-state drive.

[0063] Exemplarily, based on the above considerations, an adjustment factor related to the number of free super blocks, namely, the bandwidth ratio, is added to the bandwidth ratio of the expected host write bandwidth and the bandwidth ratio of the expected GC write bandwidth. ,in, To write bandwidth expectations, To recover bandwidth expectations, The amount of data written for the command, The amount of data recovered for the command, To predict the number of commands, is the command quantity threshold, is a constant coefficient. The value should generally be one order of magnitude higher than , which is determined through steady-state performance testing. Furthermore, the master core calculates the bandwidth ratio of the expected host write bandwidth to the expected GC write bandwidth based on the data volume of the GC source data block counted every second. Then, the bandwidth ratio is sent to the slave core. At the beginning of each second, each microcontroller unit (MCU) sets the current time as the starting time and independently counts the GC write bandwidth and host write bandwidth in the next second. For the slave cores located in the interval At some point, to During this period, the host writes data volume , the amount of data written by GC is . Bandwidth control will block input / output (IO). Therefore, if each core arbitrates independently, different blocking conditions of each core may lead to unexpected deadlock. And by adopting the unified control method of the main core, the bandwidth of each core in the sub-time interval is calculated uniformly. The bandwidth upper limit is a shared variable and is visible to all cores. In the main loop, the main core dynamically adjusts the upper limit in real time according to the current actual ratio of host writes to GC writes. The basic principle is: first, calculate the expected bandwidth ratio of host writes to GC writes, and normalize the ratio to 100,000. Then, after determining If the number of remaining free data blocks is less than the dangerous number, divide the Host write ratio by 2 to prevent timeout.

[0064] The advantage of this setting is that the bandwidth ratio is determined according to the predicted number of prediction commands. In the main loop, the main core dynamically adjusts the upper limit in real time according to the current actual bandwidth ratio of host writes to GC writes, thereby improving the steady-state performance of the SSD.

[0065] The present invention provides a method for optimizing the performance of a solid-state hard disk, which obtains the historical operation data of the solid-state hard disk; inputs the test data set of the historical operation data into the target network model to obtain the predicted number of commands output by the target network model; wherein the target network model is obtained by training the training data set, the verification data set and the long-term and short-term neural network model based on the historical operation data; determines the bandwidth ratio according to the predicted number of commands and the command number threshold; wherein the command number threshold is a preset number; determines the performance optimization result of the solid-state hard disk according to the bandwidth ratio and the dynamic optimization rule; wherein the dynamic optimization rule is a preset rule for adjusting the bandwidth ratio. Based on the above embodiments, the technical scheme of the present invention is used to solve the defect that the storage performance of the hard disk is evaluated by the traditional write arbitration control algorithm in the prior art, and it is impossible to accurately control and optimize the hardware with different performances, and realizes the target network model obtained by training the training data set, the verification data set and the long-term and short-term neural network model based on the solid-state hard disk, outputs the predicted number of commands for the solid-state hard disk, and determines the bandwidth ratio of the solid-state hard disk based on the predicted number of commands, thereby optimizing the solid-state hard disk performance in real time based on the bandwidth ratio, and achieving accurate control of the solid-state hard disk.

[0066] The performance optimization device for a solid state drive provided by the present invention is described below. The performance optimization device for a solid state drive described below and the performance optimization method for a solid state drive described above can be referred to each other.

[0067] Figure 3is a schematic diagram of the structure of the performance optimization device for a solid state hard disk provided by the present invention, referring to Figure 3 As shown, the performance optimization device 300 for a solid state drive includes: a data acquisition module 301 , a quantity prediction module 302 , a ratio determination module 303 and a result determination module 304 .

[0068] In an exemplary embodiment, the data acquisition module 301 is specifically used to: acquire initial sample operation data of the solid state drive; clean the initial sample operation data of outliers and / or missing values ​​to obtain target sample operation data; and normalize the target sample operation data to obtain historical operation data.

[0069] In an exemplary embodiment, the historical operation data includes a training data set, a validation data set, and a test data set.

[0070] In an exemplary embodiment, the device further includes: a data partitioning module. The data partitioning module is used to: after acquiring the historical operation data of the solid state drive, determine the training data set, the verification data set, and the test data set according to the historical operation data and the data partitioning rule; wherein the data partitioning rule is a pre-set partitioning ratio for partitioning the training data set, the verification data set, and the test data set.

[0071] In an exemplary embodiment, the device further includes: a model training module. The model training module is used to: determine a time series forward arrangement training data set and a time series reverse arrangement training data set according to a training data set; input the time series forward arrangement training data set and the time series reverse arrangement training data set into a long-term and short-term neural network model respectively to obtain a two-way learning network model; train the two-way learning network model according to a preset number of training times and error accuracy to obtain an initial network model; input a verification data set into the initial network model for model optimization to obtain a target network model.

[0072] In an exemplary embodiment, the ratio determination module 303 is specifically used to: obtain the command write data volume and the command reclaim data volume; obtain the data volume ratio based on the ratio of the command write data volume to the command reclaim data volume; determine the quantity difference based on the difference between the predicted command quantity and the command quantity threshold; determine the bandwidth ratio based on the data volume ratio and the quantity difference.

[0073] In an exemplary embodiment, the result determination module 304 is specifically used to: determine a preset ratio; wherein the preset ratio is a ratio of a preset write bandwidth expectation to a preset reclaim bandwidth expectation; determine a relationship between the preset ratio and the bandwidth ratio; adjust the bandwidth ratio according to the relationship between the preset ratio and the bandwidth ratio and a dynamic optimization rule to determine a performance optimization result.

[0074] The device of this embodiment can be used to execute the method of any embodiment in the embodiment of the performance optimization method side of the solid-state hard disk. Its specific implementation process and technical effects are similar to those in the embodiment of the performance optimization method side of the solid-state hard disk. For details, please refer to the detailed description in the embodiment of the performance optimization method side of the solid-state hard disk, which will not be repeated here.

[0075] Figure 4 is a schematic diagram of the structure of the electronic device provided by the present invention, such as Figure 4 As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communication interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the performance optimization method of the solid state drive, the method comprising: obtaining the historical operation data of the solid state drive; inputting the test data set of the historical operation data into the target network model to obtain the predicted number of commands output by the target network model; wherein the target network model is obtained by training the training data set, the verification data set and the long-short term neural network model based on the historical operation data; the predicted number of commands is used to evaluate the performance of the solid state drive; the bandwidth ratio is determined according to the predicted number of commands and the command number threshold; wherein the command number threshold is a preset number; the bandwidth ratio is used to determine whether the performance of the solid state drive needs to be optimized and adjusted; the performance optimization result of the solid state drive is determined according to the bandwidth ratio and the dynamic optimization rule; wherein the dynamic optimization rule is a preset rule for adjusting the bandwidth ratio.

[0076] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0077] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the solid-state hard disk performance optimization method provided by the above methods, the method including: obtaining historical operation data of the solid-state hard disk; inputting a test data set of the historical operation data into a target network model to obtain a predicted number of commands output by the target network model; wherein the target network model is obtained by training a training data set, a verification data set and a long-short term neural network model based on the historical operation data; the predicted number of commands is used to evaluate the performance of the solid-state hard disk; the bandwidth ratio is determined based on the predicted number of commands and a command number threshold; wherein the command number threshold is a pre-set number; the bandwidth ratio is used to determine whether the performance of the solid-state hard disk needs to be optimized and adjusted; the performance optimization result of the solid-state hard disk is determined based on the bandwidth ratio and dynamic optimization rules; wherein the dynamic optimization rules are pre-set rules for adjusting the bandwidth ratio.

[0078] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the performance optimization method of the solid-state hard disk provided by the above-mentioned methods, the method comprising: obtaining historical operation data of the solid-state hard disk; inputting a test data set of the historical operation data into a target network model to obtain a predicted number of commands output by the target network model; wherein the target network model is obtained by training a training data set, a verification data set and a long-short term neural network model based on the historical operation data; the predicted number of commands is used to evaluate the performance of the solid-state hard disk; the bandwidth ratio is determined based on the predicted number of commands and a command number threshold; wherein the command number threshold is a pre-set number; the bandwidth ratio is used to determine whether the performance of the solid-state hard disk needs to be optimized and adjusted; the performance optimization result of the solid-state hard disk is determined based on the bandwidth ratio and dynamic optimization rules; wherein the dynamic optimization rules are pre-set rules for adjusting the bandwidth ratio.

[0079] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0080] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing the performance of a solid state drive, characterized in that: include: Get the historical operation data of the solid state drive; Inputting the test data set of the historical operation data into the target network model to obtain the predicted number of commands output by the target network model; wherein the target network model is obtained by training based on the training data set, the verification data set and the long-short term neural network model of the historical operation data; the predicted number of commands is used to evaluate the performance of the solid state drive; Determining the bandwidth ratio according to the predicted command quantity and the command quantity threshold; wherein the command quantity threshold is a preset quantity; and the bandwidth ratio is used to determine whether it is necessary to optimize and adjust the performance of the solid state drive; The performance optimization result of the solid state drive is determined according to the bandwidth ratio and the dynamic optimization rule; wherein the dynamic optimization rule is a pre-set rule for adjusting the bandwidth ratio.

2. The method for optimizing the performance of a solid state drive according to claim 1, wherein: The obtaining of historical operation data of the solid state drive includes: Acquiring initial sample operation data of the solid state drive; Cleaning the initial sample operation data for outliers and / or missing values ​​to obtain target sample operation data; The target sample operation data is normalized to obtain the historical operation data.

3. The method for optimizing the performance of a solid state drive according to claim 1, wherein: The historical operation data includes the training data set, the verification data set and the test data set; After obtaining the historical operation data of the solid state drive, the method further includes: The training data set, the verification data set and the test data set are determined according to the historical operation data and the data partitioning rule; wherein the data partitioning rule is a pre-set partitioning ratio for partitioning the training data set, the verification data set and the test data set.

4. The method for optimizing the performance of a solid state drive according to claim 1, wherein: The training process of the target network model is as follows: Determine a time series forward arrangement training data set and a time series reverse arrangement training data set according to the training data set; Inputting the time series forward arrangement training data set and the time series reverse arrangement training data set into the long-term and short-term neural network model respectively to obtain a bidirectional learning network model; The two-way learning network model is trained according to a preset number of training times and error accuracy to obtain an initial network model; The verification data set is input into the initial network model for model optimization to obtain the target network model.

5. The method for optimizing the performance of a solid state drive according to claim 1, wherein: The determining the bandwidth ratio according to the predicted command quantity and the command quantity threshold comprises: Get the amount of data written by the command and the amount of data recovered by the command; Obtaining a data volume ratio according to a ratio of the amount of data written by the command to the amount of data recovered by the command; determining a quantity difference according to a difference between the predicted command quantity and the command quantity threshold; The bandwidth ratio is determined according to the data volume ratio and the quantity difference.

6. The method for optimizing the performance of a solid state drive according to claim 5, characterized in that: The determining the performance optimization result of the solid state drive according to the bandwidth ratio and the dynamic optimization rule includes: Determine a preset ratio; wherein the preset ratio is a preset ratio of the expected write bandwidth to the expected reclaim bandwidth; Determining a relationship between the preset ratio and the bandwidth ratio; The bandwidth ratio is adjusted according to the relationship between the preset ratio and the bandwidth ratio and the dynamic optimization rule to determine the performance optimization result.

7. A performance optimization device for a solid state hard disk, characterized in that: include: A data acquisition module, used to acquire historical operation data of the solid state drive; a quantity prediction module, used for inputting the test data set of the historical operation data into the target network model to obtain the predicted command quantity output by the target network model; wherein the target network model is obtained by training based on the training data set, the verification data set and the long-short term neural network model of the historical operation data; and the predicted command quantity is used to evaluate the performance of the solid state drive; a ratio determination module, configured to determine a bandwidth ratio according to the predicted command quantity and a command quantity threshold; wherein the command quantity threshold is a preset quantity; and the bandwidth ratio is used to determine whether it is necessary to optimize and adjust the performance of the solid state drive; A result determination module is used to determine the performance optimization result of the solid state drive according to the bandwidth ratio and the dynamic optimization rule.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the performance optimization method of the solid state drive as claimed in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the performance optimization method of the solid state drive as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the performance optimization method of the solid state drive as claimed in any one of claims 1 to 6 is implemented.

Images