WO2020224500A1

WO2020224500A1 - Data processing method and apparatus for solid state disk

Info

Publication number: WO2020224500A1
Application number: PCT/CN2020/087728
Authority: WO
Inventors: 李卫军; 王岩; 杨颖�; 李文江
Original assignee: 深圳大普微电子科技有限公司
Priority date: 2019-05-09
Filing date: 2020-04-29
Publication date: 2020-11-12
Also published as: CN111913649A; CN111913649B

Abstract

A data processing method and apparatus for a solid state disk, for use in predicting I/O information in a first time period in the future by means of a prediction model, so as to perform active management on a solid state disk according to the prediction result to reduce the operation response time of a host. The method according to embodiments of the present invention comprises: predicting I/O information in a first time period in the future by means of a prediction model, the I/O information comprising at least an I/O type and/or I/O intensity; and performing active management on a solid state disk according to the prediction result to reduce the operation response time of a host.

Description

Data processing method and device of solid state hard disk

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office, the application number is 201910385553.6, and the invention title is "a data processing method and device for solid-state hard disks" on May 9, 2019, the entire content of which is incorporated by reference In this application.

Technical field

The present invention relates to the technical field of solid-state hard disk data processing, in particular to a solid-state hard disk data processing method and device.

Background technique

In the SSD solid state drive, the read and write buffer of the SSD is relatively small, generally a few MB to tens of MB. Due to the asymmetric problem of the bus bandwidth of the SSD interface and the storage speed of the SSD solid state drive, the internal storage Buffer often becomes the bottleneck of write performance.

Therefore, how to perform the management of the solid state drive, reduce the response time of the host client's operation, and effectively allocate the limited hardware resources to better utilize the main control performance of the SSD is an urgent problem to be solved.

Summary of the invention

The embodiment of the present invention provides a data processing method and device for a solid state drive, which are used to predict the I/O information of the first time period in the future according to a prediction model, so as to perform active management on the solid state drive based on the prediction result to reduce the number of hosts Terminal operation response time.

The first aspect of the embodiments of the present application provides a data processing method for a solid-state hard disk, including:

Predicting the I/O information for the first time period in the future through the prediction model, where the I/O information includes at least I/O type and/or I/O intensity;

Actively manage the solid state drive based on the predicted result to reduce the response time of the host.

Preferably, the I/O type includes read operation and write operation;

The performing management of the solid state drive according to the prediction result includes:

Get the first read/write ratio of the current time period;

According to the prediction result, obtain the second read-write ratio in the first time period;

Determine whether the second read-write ratio is the same as the first read-write ratio;

If not, adjust the cache allocation in the solid state disk according to the second read-write ratio.

Preferably, the adjusting the cache allocation in the solid state hard disk according to the second read-write ratio includes:

If the intensity of the read operation or the write operation in the second read-write ratio changes, correspondingly adjust the buffer occupied by the read operation or the write operation.

Preferably, the performing management on the solid state disk according to the prediction result includes:

According to the prediction result, obtain the I/O intensity in the first time period;

Obtain the I/O throughput of the solid state drive;

According to the I/O intensity and the I/O throughput in the first time period, it is determined whether the solid state disk is in an idle state.

If the solid state drive is in an idle state;

Judging whether the remaining space of the solid-state hard disk is less than a first threshold;

If it is, a garbage collection event in the solid state drive is triggered.

If the solid state hard disk is not in an idle state;

Obtain the remaining space that can be allocated in the SSD;

Determine whether the I/O write operation intensity in the first time period is greater than the allocatable remaining space;

If it is, a garbage collection event in the solid state drive is triggered.

According to the prediction result, respectively obtain the I/O intensity of the multiple applications in the first time period;

Judging whether there is an I/O intensity difference between two applications in the multiple applications that is greater than a second threshold;

If so, the application with the lower I/O intensity among the two applications is processed first.

Preferably, the prediction model is built into the solid state hard disk and/or the host side to reduce the operation response time of the host side.

An embodiment of the present application also provides a data processing device for a solid state hard disk, including:

The prediction unit is used to predict the I/O information of the first time period in the future through the prediction model, and the I/O information includes at least I/O type and/or I/O size;

The management unit is used to perform management on the solid state drive according to the prediction result, so as to reduce the operation response time of the host side.

Preferably, the I/O type includes read operation and write operation;

Preferably, the management unit is specifically used for:

Get the first read/write ratio of the current time period;

Preferably, the management unit is specifically used for:

Obtain the I/O throughput of the solid state drive;

Preferably, the management unit is specifically used for:

If the solid state drive is in an idle state;

If it is, a garbage collection event in the solid state drive is triggered.

Preferably, the management unit is specifically used for:

If the solid state hard disk is not in an idle state;

Obtain the remaining space that can be allocated in the SSD;

If it is, a garbage collection event in the solid state drive is triggered.

Preferably, the management unit is specifically used for:

Preferably, the prediction model is built in the solid state hard disk and/or the host.

The embodiment of the present application also provides a computer device, including a processor, which is used to implement the data processing method of the solid-state hard disk provided in the first aspect of the embodiment of the present application when the processor executes the computer program stored on the memory.

The embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it is used to implement the data processing method of the solid state hard disk provided in the first aspect of the embodiment of the present application.

It can be seen from the above technical solutions that the embodiments of the present invention have the following advantages:

In the embodiment of the present application, the I/O information of the first time period in the future is predicted according to the prediction model, and the I/O information includes I/O type and I/O intensity. Active management is performed on the solid state drive according to the prediction result to Reduce the operation response time on the host side. Because in the embodiment of this application, the I/O information of the first time period in the future can be predicted based on the prediction model, and the solid state drive can be actively managed based on the predicted result to reduce the operation response time of the host side, thereby improving the performance of the solid state drive. The main control performance improves the user experience of the solid state drive.

Description of the drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only It is an embodiment of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on the provided drawings without creative work.

FIG. 1 is a schematic diagram of an embodiment of a data processing method of a solid-state hard disk in an embodiment of the application;

FIG. 2 is a detailed step of step 101 in the embodiment of FIG. 1 in the embodiment of this application;

FIG. 3 is a detailed step of step 203 in the embodiment of FIG. 2 in the embodiment of this application;

FIG. 4 is a detailed step of step 301 in the embodiment of FIG. 3 in the embodiment of this application;

FIG. 5 is a detailed step of step 303 in the embodiment of FIG. 3 in the embodiment of this application;

Fig. 6 is a schematic diagram of a neuron set in an embodiment of the application;

FIG. 7 is a schematic diagram of the calculation process of each neuron cell in an embodiment of the application;

FIG. 8 is a detailed step of step 102 in the embodiment of FIG. 1 in the embodiment of this application;

FIG. 9 is another detailed step of step 102 in the embodiment of FIG. 1 in the embodiment of this application;

FIG. 10 is another detailed step of step 102 in the embodiment of FIG. 1 in the embodiment of this application;

FIG. 11 is a schematic diagram of an embodiment of a data processing device for a solid-state hard disk in an embodiment of the application.

Detailed ways

In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only It is a part of the embodiments of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

The terms "first", "second", "third", "fourth", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe specific Order or precedence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order other than the content illustrated or described herein. In addition, the terms "including" and "having" and any variations of them are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to the clearly listed Those steps or units may include other steps or units that are not clearly listed or are inherent to these processes, methods, products, or equipment.

In view of the problem that the write buffer buffer becomes a write performance bottleneck caused by the small write buffer buffer and the asymmetrical interface bus bandwidth of the solid state drive SSD and the storage speed of the SSD disk in the prior art, the embodiment of the application proposes A data processing method and device for a solid state hard disk are used to solve the problem that the write buffer Buffer is small and the response time of the host side operation is too long.

To facilitate understanding, the following describes the data processing method of the solid state drive in the embodiment of the present application. For details, please refer to FIG. 1. An embodiment of the data processing method of the solid state drive in the embodiment of the present application includes:

101. Predict I/O information for the first time period in the future through a prediction model, where the I/O information includes at least I/O type and/or I/O intensity;

In view of the problem that the response time of the host side operation is too long due to the small write buffer Buffer in the prior art, the embodiment of the present application can use the data processing device of the solid state hard disk to use the predictive model to analyze the I/O in the first time period in the future. Information is predicted, where the I/O information includes at least I/O type and I/O intensity, so as to perform step 102 according to the predicted result.

Specifically, how to predict the I/O information of the first time period in the future according to the prediction model will be described in detail in the following embodiments, and will not be repeated here.

102. Perform active management of the solid-state hard drive according to the prediction result to reduce the operation response time of the host side.

After obtaining the I/O prediction results for the first time period in the future, the data processing device of the solid-state hard disk performs active management on the solid-state hard disk according to the predicted result, so as to reduce the operation impact time of the host side and improve the user's practical experience of the solid-state hard disk.

Specifically, the process of performing active management of the solid state disk according to the prediction result will be described in detail in the following embodiments, and will not be repeated here.

Based on the embodiment described in Fig. 1, step 101 is described in detail below. Please refer to Fig. 2 for details. Fig. 2 shows the detailed steps of step 101:

201. Obtain the interface protocol command received by the solid state drive SSD;

The interface protocol command is a communication protocol command used between the host and the memory (solid state drive SSD), mainly including SATA, PCIe, NVMe and other protocol interfaces. Preferably, the interface protocol instructions in this embodiment are NVMe instructions.

Specifically, NVMe is an interface specification for connecting storage to a server through a PCI Express bus. Simply put, NVMe makes the communication between the SSD and the host system faster. In the embodiment of the present application, during the data interaction process between the SSD disk and the host system, the NVMe protocol commands are used to implement fast data input and output.

Different from the prior art, SSD uses passive solid-state drive management (passive buffer scheduling, passive garbage collection, etc.) to implement hardware resource configuration and instruction queue management in SSD. Among them, hardware resource configuration includes buffer scheduling and garbage collection. Wait. In the embodiment of the present application, when the solid state drive SSD receives the NVMe protocol instruction sent by the host system, the data processing device of the solid state drive obtains the protocol instruction, and executes step 202 according to the protocol instruction, so as to realize the first time period in the future. Forecast of I/O information.

Specifically, the data processing device of the solid state hard disk may actively obtain the NVMe protocol command from the SSD, or may passively receive the NVMe protocol command sent by the SSD, and there is no specific limitation here.

202. Parse the interface protocol command to obtain I/O information in the protocol command, where the I/O information includes at least an I/O time stamp, an I/O type, and an I/O size;

After the data processing device of the solid state hard disk obtains the interface protocol instruction, it parses the interface protocol instruction to obtain the I/O information in the protocol instruction, and the parsing action can be embodied in accordance with specific protocol specifications (such as NVMe protocol specifications), Read the I/O information in the protocol command, where the I/O information includes at least I/O timestamp, I/O type and I/O size.

Specifically, the I/O timestamp is the generation time of the I/O operation, the I/O type includes read operation or write operation, and the I/O size is the data length of the read operation or write operation.

203. Perform machine learning on the I/O information to predict the I/O information for the first time period in the future.

After the data processing device of the solid state hard disk obtains the I/O information of the current time period, it can learn the I/O information of the past time period according to the machine learning algorithm, so as to predict the I/O information of the first time period in the future .

Among them, machine learning algorithms include but are not limited to neural networks, support vector machines, linear regression, time series, etc. There are no specific restrictions on the machine learning algorithms here.

Based on the embodiment shown in FIG. 2, step 203 in the embodiment in FIG. 2 will be described in detail below. Please refer to FIG. 3 for details. FIG. 3 is a detailed step of step 203 in the embodiment in FIG.

301. Perform preprocessing on the I/O information to obtain a first I/O information table;

The data processing device of the solid state hard disk obtains the I/O information in the instruction by parsing the interface protocol instruction (such as the NVMe protocol instruction), and performs preprocessing on the I/O information to obtain the first I/O information table.

Specifically, the specific process of the preprocessing operation will be described in detail in the following embodiments, and will not be repeated here.

302. Combine multiple adjacent I/O operations in the first I/O information table into one I/O operation.

After acquiring the first I/O information table, the data processing device of the solid-state hard disk merges multiple adjacent I/O operations in the first I/O information table into one I/O operation, where the combination can be understood as Connect the 8 adjacent I/Os processed in the following table 2 and merge them into one I/O input, so that compared with the existing one I/O operation input, multiple I/Os can be input at the same time Calculate once, thereby greatly reducing the amount of calculation and meeting the needs of real-time calculation.

Specifically, 4 adjacent I/O operations may be combined into one I/O operation, or 8 adjacent I/O operations may be combined into one I/O operation. In this embodiment, “multiple "It is mainly determined by the processing performance of the processor (that is, the computing power of the processor). If the computing power of the processor is weak, more adjacent numbers of I/O operations can be combined. If the computing power of the processor is strong , You can combine a smaller number of adjacent I/O operations to adapt to the processing performance of the processor.

303. Use a plurality of combined I/O operation characteristic values for neural network LSTM model learning to obtain an I/O prediction result in the first time period.

After combining multiple I/O operations into one I/O operation, a set of I/O operation characteristic values can be obtained. The specific I/O operation characteristic values are related to I/O timestamp, I/O type and Variables of the I/O size, and in order to predict the I/O information in the first time period in the future, a machine learning algorithm is generally used for prediction. The machine learning algorithm in this embodiment is preferably a neural network learning algorithm.

It should be noted that when using a neural network learning algorithm to predict the I/O information for the first time period in the future, it is necessary to train the known I/O information using a neural network learning model to obtain a neural network training model , And then predict the I/O information in the first time period in the future based on the neural network training model and the current I/O information.

The training process of the neural network learning model according to the known I/O information is described in detail in the prior art, and will not be repeated here, and the neural network is executed for the combined I/O operation characteristic values The specific implementation process of model learning (that is, model inference) to obtain I/O prediction information in the first time period will be described in the following embodiments, and will not be repeated here.

304. Perform post-processing on the I/O prediction result to adjust the prediction result.

Because the prediction algorithm relies more on the selection of assumptions and features, after getting the I/O prediction information, there may be a large deviation in the prediction result, and the prediction result is obviously deviating from the reality, so I/O prediction is required Information performs post-processing, such as deleting the abnormal prediction value in the I/O prediction information, and obtaining the normal prediction result.

In the embodiments of the present application, the process of performing machine learning based on I/O information to predict I/O information for the first time period in the future is described in detail, which improves the implementability of the embodiments of the present application.

Based on the embodiment described in FIG. 3, the preprocessing process of I/O information in step 301 is described in detail below. For details, please refer to FIG. 4. FIG. 4 shows the detailed steps of step 301 in the embodiment in FIG. 3.

401. According to the I/O timestamp, calculate the time interval between multiple sets of I/O operations that meet the first preset number.

Corresponding to the description in step 301, it is assumed that after the data processing device of the solid-state hard disk performs analysis on the NVMe instruction, the I/O information obtained is as shown in Table 1:

Table 1

Then take Table 1 as an example to describe the preprocessing process of I/O information. The preprocessing process is to first perform step 401, that is, to calculate the time between multiple sets of I/O operations that meet the first preset number Specifically, step 401 can be to calculate the time interval between 32 I/O operations, or to calculate the time interval between 16 I/O operations. The selection of the preset number depends on the processing capacity of the host system. To make specific restrictions, the following takes 32 I/O operations as an example.

Specifically, the time stamp of 32 I/Os is calculated every interval, and the interval time is calculated; interval=timestep _i+31 -timestep _i

Among them, timestepi represents the timestamp of the i-th I/O instruction.

402. Calculate the total number of read operations and write operations of the first preset number of I/O operations in each group within the time interval and the corresponding data size;

After obtaining the time intervals between multiple groups of I/O operations that satisfy the first preset number, count the total number of read operations and write operations of the first preset number of I/O operations in each group within the time interval And the corresponding data size.

Further, digital conversion of read and write I/O data, write I/O as 1, and read I/O as 0, and then each group of 32 I/Os are accumulated to calculate the number of read I/O and write I/O quantity, the formula is as follows:

among them,

Specifically, the sum of the number of read operations and write operations of the first preset number of I/O operations in each group within the time interval and the corresponding data size can be calculated according to the above formula.

403. Perform a compression operation on the data size to obtain the first I/O information table.

In order to facilitate regression prediction, the combined data size needs to be compressed to a reasonable range (0-500), so the size of the read I/O corresponding to a set of 32 I/Os needs to be accumulated, and then divided by 200K for compression, similarly accumulate the size of read I/O corresponding to a set of 32 I/Os, and then divide by 200K for compression to obtain the processed first I/O information table as shown in Table 2. .

Specifically, the size of the compression ratio can be customized according to actual needs (for example, the size of the read I/O corresponding to a set of 32 I/Os can be accumulated, and then divided by 400K or 500K for compression), as long as the compression The size of the latter data only needs to meet the data range for regression analysis, and there is no specific restriction here.

Table 2

The time interval in Table 2 above represents the time interval between 32 I/O operations. The total I/O type represents the total number of read operations and write operations in 32 I/O operations. For example, when the sum is 0, then Indicates that the 32 I/O operations are all read operations, and when the sum is 4, it indicates that there are 4 write operations and 28 read operations in the 32 I/O operations, and the read operations and write operations correspond to compression The latter data size is the compressed data size corresponding to the total read operations and total write operations in the total of 32 I/Os.

In the foregoing embodiment, the preprocessing process of the I/O information is described in detail, that is, the generation process of the first I/O information table is described in detail, which improves the implementability of the embodiments of the present application.

Based on the embodiment described in Figure /3, the following is to perform neural network LSTM model learning (ie model inference) on the multiple combined I/O operation feature values in step 303 to obtain the first time period The process of I/O prediction information is described in detail. Please refer to FIG. 5 for details. FIG. 5 shows the detailed steps of step 303 in the embodiment of FIG. 3.

501. Input multiple combined I/O operation eigenvalues to the input fully connected layer of the neural network to map the combined I/O operation eigenvalues into a high-dimensional vector space through linear transformation. /O operation characteristic value is related to the I/O timestamp, the I/O type, and the I/O size;

Corresponding to step 302 of the embodiment described in FIG. 3, multiple adjacent I/O operations in Table 2 are combined into one I/O operation, and then multiple combined I/O operation characteristic values are input to The input fully connected layer of the neural network uses linear transformation to map the merged I/O operation characteristic values into a high-dimensional vector space, that is, to map low-dimensional I/O operation characteristic values to high-dimensional I/O operation characteristics Value, where I/O operation characteristic value is related to I/O timestamp, I/O type and I/O size.

Specifically, the characteristic value of each combined I/O operation can be obtained by the following methods:

For example, if the 8 adjacent I/Os processed in Table 2 are combined into one I/O input, the combination here can be understood as connecting the 8 adjacent I/Os processed in Table 2 to combine them into one I/O O input, because each I/O in Table 2 is combined from the 32 I/Os in Table 1, so 8 adjacent I/Os in Table 2 are combined into one I/O input, which is equivalent To merge 32*8=256 I/O into one I/O input, the specific merge operation can be as follows to obtain a merged I/O operation characteristic value, where each I/O operation characteristic The values are variables related to I/O timestamp, I/O type, and I/O size.

[interval ₁ ,ioTypeSum ₁ ,rSize ₁ ,wSize ₁ ,interval ₂ ,ioTypeSum ₂ ,rSize ₂ ,wSize ₂ ,...,interval ₈ ,ioTypeSum ₈ ,rSize ₈ ,wSize ₈ ]

After obtaining multiple merged I/O characteristic values, input the multiple merged I/O characteristic values into the input fully connected layer of the neural network to transform the merged I/O operation characteristic values through linear transformation Do high-dimensional vector space mapping, that is, map low-dimensional I/O operation characteristic values to high-dimensional I/O operation characteristic values to discover more characteristic values.

Specifically, a set of 32*1 vectors can be transformed into a set of 128*1 vectors through a weight matrix. The specific transformation process can be X _128×1 =W _128×32 ·S _32×1 +b _{128 ×1} , where W _128× ₃₂ is the weight matrix, and b _128×1 is the offset to map the 32-dimensional I/O operation characteristic value to the 128-dimensional I/O operation characteristic value.

502. A neuron set is formed by a plurality of neural network cells LSTM Cells, so that the feature values after the high-dimensional vector space mapping are sequentially input into the neuron set to perform operations, to obtain an operation result;

After obtaining the high-dimensional I/O operation characteristic values, multiple neural network cells LSTM Cells are used to form a neuron set (specifically as shown in FIG. 6, assuming that the embodiment of this application includes 128 neuron cells LSTM Cell), and The characteristic values of high-dimensional I/O operations are sequentially and cyclically input into the neuron set for operation to obtain the operation result.

Specifically, in actual calculations, within a certain range, the greater the number of neuron cells, the more accurate the calculation results, but too much calculation will put too much load on the processor, which will affect the efficiency of real-time output. Therefore, the data volume of neuron cells is generally determined by the accuracy of the calculation results and the processing power of the processor.

The following takes a single neuron cell as an example to describe the calculation process of high-dimensional feature values:

As shown in Figure 7, the calculation process of each neuron cell is shown in Figure 7. The neural network structure of each neuron cell includes input gate, forget gate and output gate, as well as new and past memory states. Among them, the function of each door or state has been described in detail in the prior art, and will not be repeated here.

However, in this embodiment, assuming that there are 128 neuron cells, the calculation result is looped 128 times in sequence to improve the accuracy of the calculation.

503. Input the operation result to the output fully connected layer of the neural network to map the operation result into a prediction result output dimension, where the prediction result output dimension reflects the I/O prediction result of the first time period .

After the high-dimensional feature vector value is processed by multiple neuron cells, the data output from the output gate is input to the output fully connected layer of the neural network to map the calculation result to the output dimension of the prediction result. The output dimension of depends on the user's choice, that is, it can output feature values of any dimension, such as outputting 32-dimensional feature values, or outputting 16-dimensional feature values, or feature values of other dimensions.

Specifically, the embodiment of the present application takes an example of outputting 32-dimensional feature values for description:

The operation result is mapped to 32-dimensional eigenvalues, the specific mapping process is as follows: Y _32×1 = W ₃₂ _{× 128} · X _128×1 + b _32×1 , that is, through the weight matrix W _32×128 , the 128-dimensional The eigenvalue is mapped to a 32-dimensional eigenvalue, where b _32×1 is the matrix offset, which represents the influence of noise in data processing.

The data processing process shown in FIG. 5 will be described in detail below with specific embodiments:

Suppose that in the LSTM network model, one sample input is X={x ₁ ,x ₂ ,...,x _T }, where T is the time step of the input sequence (tentatively T=1024), specifically, corresponding to Step 401, each x _T is the characteristic value of 32*8=256 I/O operations. Because the duration of each I/O operation is about 3 us, the time interval corresponding to 1024 x in the input sample is about 32* 8*3*1024≈1s, and the model output after step 403 is Y={y ₁ ,y ₂ ,...,y ₁₆ ,y ₁₇ ,...,y ₃₂ }, where y ₁ ,y ₂ ,...,y ₁₆ is the predicted value of the read operation size every 50ms in the future, that is to say, the predicted value of the intensity of the read operation in the future 0-50ms, 50-100ms,...,750ms-800ms, and y ₁₇ ,...,y ₃₂ is the predicted value of the size of the write operation every 50ms in the future, that is, the predicted value of the strength of the write operation in the future 0-50ms, 50-100ms,...,750ms-800ms.

It should be noted that when the neural network model is used to predict the I/O information in the first time period in the future, it is generally required that the amount of historical data is greater than the amount of predicted data. For example, the actual I/O information amount within 1s is used in this implementation. Predict the amount of I/O information in the future 800ms. In order to improve the accuracy, it can also predict the amount of I/O information in the future 400ms. Generally, the larger the amount of historical data, the smaller the amount of predicted data, and the corresponding The higher the prediction accuracy rate, on the contrary, the lower the prediction accuracy rate. In the Y value output by the model, y ₁ , y ₂ ,..., y _{16 can} be the predicted value of the write operation at an interval of 50 ms in the future, and y ₁₇ ,..., y ₃₂ are the predicted value of each write operation in the future. The predicted value of the read operation with an interval of 50ms, that is, the output value in the Y value mainly depends on the previous model training process. If the previous model training process, y ₁ , y ₂ ,..., y ₁₆ are the future 50ms is the predicted value of the size of the read operation at an interval. When the later model outputs, y ₁ , y ₂ ,..., y _{16 are} the predicted values of the size of the read operation at every 50ms interval.

Based on the embodiment described in Figure 1, the prediction model and prediction process are described in detail in the above embodiment, and the steps for active management of the solid state drive based on the prediction result will be described in detail below. For details, please refer to Figure 8, Figure 8. It is a detailed step of step 102 in the embodiment of FIG. 1:

801: Obtain the first read-write ratio of the current time period;

Specifically, when the data processing device of the solid state drive performs active management on the solid state drive according to the prediction result, it can obtain the first read/write ratio of the current time period, where the current time period may be the current 100ms, the current 100ms, or the current 200ms, etc. , No specific restrictions here

The first read/write ratio corresponds to the ratio of the number of read operations to the number of write operations in the current time period. In practical applications, it can also be used to directly obtain the number of read operations or write operations in the current time period. Calculate the first read-write ratio, because the first read-write ratio is mainly used to facilitate the comparison with the second read-write ratio in the first time in the future. In actual use, the user can make a decision based on the convenience of comparison.

802. Obtain a second read/write ratio in the first time period according to the prediction result.

After obtaining the first read/write ratio in the current time period, the data processing device of the solid-state hard disk may also calculate the second read/write ratio in the first time period in the future based on the prediction result of step 101, and calculate according to the first read/write ratio. Step 803 is executed for the ratio and the second read-write ratio.

803. Determine whether the second read-write ratio is the same as the first read-write ratio, if not, execute step 804, and if yes, execute step 805;

After obtaining the first read-write ratio and the second read-write ratio, the data processing device of the solid-state hard disk judges whether the second read-write ratio and the first read-write ratio are the same, if not, execute step 804, if yes, execute step 805 .

804. Adjust the cache allocation in the solid state hard disk according to the second read-write ratio.

If the second read-write ratio is different from the first read-write ratio, it means that the read cache or write cache occupied in the first time period will change compared with the current time period. You can use the second read The write ratio is adjusted to the cache allocation in the SSD.

Specifically, if the intensity of the read operation in the next first time period will change, the buffer occupied by the read operation will be adjusted accordingly, and if the intensity of the write operation in the next first time period will change, the amount of the write operation will be adjusted accordingly Cache, so that the read operation or write operation in the first time period can be quickly responded to by the SSD disk, so as to improve the control performance of the solid-state hard disk controller and reduce the response time of the host.

805. Perform other processes.

If the second read-write ratio is the same as the first read-write ratio, other processes are executed, and no specific restrictions are made here.

It should be noted that the prediction model in this embodiment can be built into the solid-state hard disk or the host side, and the prediction model can be fixed, or it can be learned in real time and updated in real time, and there is no specific limitation here.

In the embodiment of the present application, a detailed description of the manner of performing active management of the solid-state hard disk in the cache scheduling is provided, which improves the implementability of the embodiment of the present application.

Based on the embodiment described in FIG. 1, the following describes in detail the steps of performing active management of the solid state drive according to the prediction result. Please refer to FIG. 9 for details. FIG. 9 is another detailed step of step 102 in the embodiment of FIG. 1.

901. Obtain the I/O intensity in the first time period according to the prediction result.

After the data processing device of the solid-state hard disk obtains the I/O prediction result for the first time period in the future, it can obtain the I/O intensity in the first time period according to the prediction result, where the I/O intensity refers to the reading per unit time. The size of the data processed by the operation or write operation.

902. Obtain the I/O throughput of the solid state drive;

Specifically, the I/O throughput of a solid state drive refers to the data intensity of a read operation or a write operation that can be processed by the solid state drive per unit time, which is an indicator of the solid state drive, and is generally marked when the solid state drive is shipped.

903. According to the I/O intensity and the I/O throughput in the first time period, determine whether the solid state disk is in an idle state, if so, perform step 904, and if not, perform step 907.

Specifically, when judging the state of the solid state drive, such as whether the solid state drive is in an idle state during the first time period, it can be determined according to the I/O intensity in the first time period and the I/O throughput of the solid state drive. It is judged that if the ratio of the I/O intensity to the I/O throughput in the first time period is less than the preset threshold, such as 40%, it is considered that the solid state drive is idle in the first time period, and step 904 is performed, otherwise , It is considered that the solid state disk is busy in the first time period, and step 907 is executed.

It is easy to understand that the size of the aforementioned preset threshold can be customized according to actual needs, such as 30% or 20%, etc., and there is no specific limitation here.

904. Determine whether the remaining space of the solid-state hard disk is less than the first threshold, if yes, execute step 905, and if not, execute step 906;

In order to maintain a large enough remaining space in the SSD disk, it is used to process the write operation or read operation in the next first time period, so that the SSD disk can quickly respond to the write operation or read operation, and further obtain the remaining space in the hard disk. It is determined whether the remaining space of the solid-state hard disk is less than the first threshold to determine whether the SSD disk has a large enough remaining space to respond to the read operation or the write operation. If so, step 905 is performed, and if not, step 906 is performed.

905. Trigger the garbage collection event of the solid state hard disk;

If the remaining space of the SSD is less than the first threshold, it indicates that the remaining space in the SSD is insufficient, and a garbage collection event in the SSD is triggered.

Different from the strategy of passively performing garbage collection in the prior art, this application can predict the I/O intensity of the first time period in the future based on the prediction result, and determine whether the solid state drive is idle according to the prediction result. When the solid state drive is in an idle state and the remaining space of the solid state drive is less than the first threshold, the garbage collection event in the solid state drive is actively triggered, thereby improving the main control performance of the solid state drive controller and reducing the operation response time of the host side.

906. Perform other processes.

If the remaining space of the solid state disk is not less than the first threshold, it means that there is enough space in the SSD disk to respond to the I/O operation in the first time period, and other processes are executed, and no specific limitation is made here.

907. Obtain allocatable remaining space of the solid state hard disk;

Corresponding to step 903, if the solid state drive is not in an idle state, and in order to ensure that the solid state drive has enough space to respond to I/O operations in the first time period in the future, the allocatable remaining space in the solid state drive can be further obtained, and the steps are performed 908.

908. Determine whether the I/O write operation intensity in the first time period is greater than the allocatable remaining space, if yes, perform step 905, if not, perform step 906;

After the I/O intensity in the first time period is obtained, it can be further determined whether the I/O write operation intensity in the first time period is greater than the allocatable remaining space, if yes, go to step 905, if not, go to step 906;

Specifically, step 908 is to determine that when the I/O intensity of the solid state drive in the first time period is greater than the remaining space that can be allocated, a garbage collection event in the solid state drive is triggered. If the I/O intensity in the first time period is not If it is greater than the remaining space that can be allocated, other processes are executed, and there is no specific restriction here.

In the embodiment of the present application, the method for performing active management of the solid state drive is described in detail from the perspective of active garbage collection, which improves the implementability of the embodiment of the present application.

Based on the embodiment described in FIG. 1, the following is a detailed description of the steps of performing active management of the solid state drive according to the prediction result. Please refer to FIG. 10 for details. FIG. 10 is another detailed step of step 102 in the embodiment of FIG. 1.

1001. According to the prediction result, respectively obtain the I/O intensity of multiple applications in the first time period;

As another way to perform active management of the solid state drive, the data processing device of the solid state drive can also obtain the I/O intensity of multiple applications in the first time in the future according to the prediction result.

Specifically, if there are 3 applications running in the first time period in the future, the I/O intensity of the first application, the second application, and the third application in the first time period are obtained respectively, and the I/O intensity of each application Intensity executes step 1002.

1002. Determine whether the I/O strength difference of two applications among the multiple applications is greater than a second threshold; if yes, perform step 1003; if not, perform step 1004;

The data processing device of the solid-state hard disk determines whether there is a plurality of applications where the I/O intensity difference between the two applications is greater than the second threshold. Specifically, the size of the second threshold can be customized according to the needs of the user, and no specifics are made here. limit.

Specifically, step 1002 is mainly to judge the I/O intensity of each application according to the I/O intensity of each application, and execute step 1003 according to the judgment result.

1003. Prioritize the application with lower I/O intensity among the two applications;

If the I/O intensity of two applications is greater than the second threshold, the application with the lower I/O intensity of the above two applications is prioritized to improve the user experience of the host.

Specifically, suppose there are two applications in the first time period, and when application A performs a long-time high-intensity read operation, application B only needs to read less data, that is, the I/ of application A and application B If the O intensity difference satisfies the condition of being greater than the second threshold, this embodiment will give priority to the B application, which is equivalent to inserting the B application before A, the average response time of the application will be greatly reduced, and the host will not feel the B Delay, thereby enhancing the user experience of the SSD.

1004. Perform other processes.

If there are no two applications whose I/O intensity is greater than the second threshold, other processes are executed, and there is no specific limitation here.

The data processing method of the solid state drive is described above, and the solid state drive processing device in this embodiment is described below. For details, please refer to FIG. 11. An embodiment of the data processing device of the solid state drive in the embodiment of the present application includes:

The prediction unit 1101 is configured to predict the I/O information of the first time period in the future through the prediction model, the I/O information at least including the I/O type and/or the I/O size;

The management unit 1102 is configured to perform management on the solid state disk according to the prediction result, so as to reduce the operation response time of the host side.

Preferably, the I/O type includes read operation and write operation;

Preferably, the prediction unit 1102 is specifically configured to:

Get the first read/write ratio of the current time period;

Preferably, the management unit 1102 is further configured to:

Preferably, the management unit 1102 is specifically configured to:

Obtain the I/O throughput of the solid state drive;

Preferably, the management unit 1102 is further configured to:

If the solid state drive is in an idle state;

If it is, a garbage collection event in the solid state drive is triggered.

Preferably, the management unit 1102 is further configured to:

If the solid state hard disk is not in an idle state;

Obtain the remaining space that can be allocated in the SSD;

If it is, a garbage collection event in the solid state drive is triggered.

Preferably, the management unit 1102 is further configured to:

It should be noted that the functions of each unit in this embodiment are the same as those described in FIGS. 1 to 10, and will not be repeated here.

In this embodiment of the application, the prediction unit 1101 predicts the I/O information for the first time period in the future, the I/O information includes I/O type and I/O intensity, and the management unit 1102 performs active management on the solid state drive , In order to reduce the operation response time of the host. Because in the embodiment of this application, the I/O information of the first time period in the future can be predicted based on the prediction model, and the solid state drive can be actively managed based on the predicted result to reduce the operation response time of the host side, thereby improving the performance of the solid state drive. The main control performance improves the user experience of the solid state drive.

The above describes the data processing device of the solid-state hard disk in the embodiment of the present invention from the perspective of modularized functional entities, and the following describes the computer device in the embodiment of the present invention from the perspective of hardware processing:

The computer device is used to implement the function of the data processing device of the solid state hard disk. An embodiment of the computer device in the embodiment of the present invention includes:

Processor and memory;

The memory is used to store computer programs, and when the processor is used to execute the computer programs stored in the memory, the following steps can be implemented:

Predicting the I/O information for the first time period in the future through the prediction model, the I/O information at least including the I/O type and/or the I/O size;

Perform management on the solid state drive based on the predicted result to reduce the response time of the host side operation.

In some embodiments of the present invention, the processor may also be used to implement the following steps:

Get the first read/write ratio of the current time period;

Obtain the I/O throughput of the solid state drive;

If the solid state drive is in an idle state;

If it is, a garbage collection event in the solid state drive is triggered.

If the solid state hard disk is not in an idle state;

Obtain the remaining space that can be allocated in the SSD;

If it is, a garbage collection event in the solid state drive is triggered.

It is understandable that when the processor in the computer device described above executes the computer program, it can also implement the functions of the units in the corresponding device embodiments described above, which will not be repeated here. Exemplarily, the computer program may be divided into one or more modules/units, and the one or more modules/units are stored in the memory and executed by the processor to complete the present invention. The one or more modules/units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the data processing device of the solid state hard disk. For example, the computer program may be divided into units in the data processing device of the above-mentioned solid-state hard disk, and each unit can implement specific functions as described in the data processing device of the corresponding solid-state hard disk.

The computer device may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server. The computer device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the processor and the memory are only examples of the computer device and do not constitute a limitation on the computer device. They may include more or fewer components, or combine certain components, or different components, such as The computer device may also include input and output devices, network access devices, buses, and the like.

The processor may be a central processing unit (Central Processing Unit, CPU), other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (ASIC), off-the-shelf Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the computer device, and various interfaces and lines are used to connect various parts of the entire computer device.

The memory may be used to store the computer program and/or module, and the processor implements the computer by running or executing the computer program and/or module stored in the memory, and calling data stored in the memory. Various functions of the device. The memory may mainly include a program storage area and a data storage area. The program storage area may store an operating system, an application program required by at least one function, etc.; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disks, memory, plug-in hard disks, smart media cards (SMC), and secure digital (SD) cards , Flash Card, at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage device.

The present invention also provides a computer-readable storage medium, which is used to implement the functions of a data processing device of a solid-state hard disk, and has a computer program stored thereon. When the computer program is executed by a processor, it can be used to execute The following steps:

In some embodiments of the present invention, when a computer program stored in a computer-readable storage medium is executed by a processor, it may be specifically used to execute the following steps:

Get the first read/write ratio of the current time period;

Obtain the I/O throughput of the solid state drive;

If the solid state drive is in an idle state;

If it is, a garbage collection event in the solid state drive is triggered.

If the solid state hard disk is not in an idle state;

Obtain the remaining space that can be allocated in the SSD;

If it is, a garbage collection event in the solid state drive is triggered.

It can be understood that if the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a corresponding computer readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned corresponding embodiments and methods, and can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can implement the steps of the foregoing method embodiments. Wherein, the computer program includes computer program code, and the computer program code may be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory) , Random Access Memory (RAM, Random Access Memory), electrical carrier signal, telecommunications signal, and software distribution media. It should be noted that the content contained in the computer-readable medium can be appropriately added or deleted in accordance with the requirements of the legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to the legislation and patent practice, the computer-readable medium Does not include electrical carrier signals and telecommunication signals.

Those skilled in the art can clearly understand that, for the convenience and conciseness of description, the specific working process of the above-described system, device, and unit can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

In the several embodiments provided in this application, it should be understood that the disclosed system, device, and method may be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components can be combined or It can be integrated into another system, or some features can be ignored or not implemented. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.

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 units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.

In addition, the functional units in the various embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or software functional unit.

If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solution of the present invention essentially or the part that contributes to the prior art or all or 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 several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program code .

As mentioned above, the above embodiments are only used to illustrate the technical solutions of the present invention, but not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: The technical solutions recorded in the embodiments are modified, or some of the technical features are equivalently replaced; these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

A data processing method for a solid state hard disk, which is characterized in that it comprises:

Predicting the I/O information for the first time period in the future through the prediction model, where the I/O information includes at least I/O type and/or I/O intensity;

Actively manage the solid state drive based on the predicted result to reduce the response time of the host.
The method according to claim 1, wherein the I/O type includes read operation and write operation;

The performing management of the solid state drive according to the prediction result includes:

Get the first read/write ratio of the current time period;

According to the prediction result, obtain the second read-write ratio in the first time period;

Determine whether the second read-write ratio is the same as the first read-write ratio;

If not, adjust the cache allocation in the solid state disk according to the second read-write ratio.
The method according to claim 2, wherein the adjusting the cache allocation in the solid state hard disk according to the second read-write ratio comprises:

If the intensity of the read operation or the write operation in the second read-write ratio changes, correspondingly adjust the buffer occupied by the read operation or the write operation.
The method according to claim 1, wherein the performing management on the solid state disk according to the prediction result comprises:

According to the prediction result, obtain the I/O intensity in the first time period;

Obtain the I/O throughput of the solid state drive;

According to the I/O intensity and the I/O throughput in the first time period, it is determined whether the solid state disk is in an idle state.
The method according to claim 4, wherein the performing management on the solid state drive according to the prediction result comprises:

If the solid state drive is in an idle state;

Judging whether the remaining space of the solid-state hard disk is less than a first threshold;

If it is, a garbage collection event in the solid state drive is triggered.
The method according to claim 4, wherein the performing management on the solid state hard disk according to the prediction result comprises:

If the solid state hard disk is not in an idle state;

Obtain the remaining space that can be allocated in the SSD;

Determine whether the I/O write operation intensity in the first time period is greater than the allocatable remaining space;

If it is, a garbage collection event in the solid state drive is triggered.
The method according to claim 1, wherein the performing management on the solid state disk according to the prediction result comprises:

According to the prediction result, respectively obtain the I/O intensity of the multiple applications in the first time period;

Judging whether there is an I/O intensity difference between two applications in the multiple applications that is greater than a second threshold;

If so, the application with the lower I/O intensity among the two applications is processed first.
The method according to any one of claims 1 to 7, wherein the prediction model is built in the solid state hard disk and/or the host.
A computer device comprising a processor, wherein the processor is used to implement the hard disk data processing method according to any one of claims 1 to 8 when the processor executes a computer program stored on a memory.
A computer-readable storage medium with a computer program stored thereon, wherein the computer program is used to implement the hard disk processing method according to any one of claims 1 to 8 when the computer program is executed by a processor.