Data processing method and apparatus for solid state disk

1. A method for data processing applied to a solid state drive, comprising:

acquiring an interface protocol command received by the solid state drive;

parsing the interface protocol command to obtain I/O information from the interface protocol command, wherein the I/O information comprises at least one I/O timestamp, an I/O type, and an I/O size;

invoking machine learning based on the I/O information to predict I/O information of a first future time period, comprising:

preprocessing the I/O information to obtain a first I/O information table stored in the solid state drive;

combining multiple adjacent I/O operations in the first I/O information table into at least one I/O operation;

obtaining I/O prediction results of the first future time period by using at least one long-short term memory (LSTM) neural network with feature values extracted from the at least one combined I/O operation; and

post-processing the I/O prediction results to adjust the I/O prediction results; and

assessing the I/O prediction results by using a first formula to evaluate an accuracy of the I/O prediction results, wherein the first formula comprises:

r

=

∑

i

=

1

n

⁢

(

x

i

-

x

_

)

⁢

(

y

i

-

y

_

)

∑

i

=

1

n

⁢

(

x

i

-

x

_

)

2

⁢

∑

i

=

1

n

⁢

(

y

i

-

y

_

)

2

wherein r represents an accuracy rate, x i represents an actual intensity of the I/O operation, y i represents a predicted I/O intensity, x represents an average value of a set of x i , and y represents an average value of a set of y i , so that a processor of the solid state drive is configured to proactively execute solid state drive management functions according to the I/O prediction results.

2. The method of claim 1 , wherein preprocessing the I/O information to obtain the first I/O information table, further comprises:

calculating time intervals between multiple groups of a given number of I/O operations according to the I/O timestamp; and

counting total number of read and write operations for each group of the given number of the I/O operations that occur in the time intervals and the corresponding data sizes to obtain the first I/O information table.

3. The method of claim 1 , wherein obtaining the I/O prediction results of the first future time period by using the at least one LSTM neural network with the feature values extracted from the at least one combined I/O operation, further comprises:

inputting the feature values extracted from the at least one combined I/O operation to an input fully connected layer of the at least one LSTM neural network in order to map the feature values to a high dimensional vector space using linear transformations, wherein the feature values are related to the I/O timestamp, the I/O type, and the I/O size;

forming a neuron set with at least one LSTM cell and feeding the feature values transformed from the high dimensional vector space to the neuron set to conduct computation; and

inputting computation result to an output fully connected layer of the at least one LSTM neural network and mapping the computation result to form feature values of an output dimension of a prediction result, wherein the feature values of the output dimension of the prediction result represent the I/O prediction results of the first future time period.

4. The method of claim 3 , wherein the feeding the feature values transformed from the high dimensional vector space to the neuron set to conduct computation, further comprises:

feeding the feature values transformed from the high dimensional vector space to network architecture of each LSTM cell to conduct computation, wherein the network architecture comprises an input gate, a forget gate and an output gate.

5. The method of claim 3 , wherein a corresponding time interval of the feature values extracted from the at least one combined I/O operation is longer than the first future time period.

6. The method of claim 1 , wherein the solid state drive management functions comprises cache management, garbage collection, or command queue management.

7. A computer device comprising a processor, wherein the processor is used to implement a data processing method applied to a solid state drive when executing a computer program, the method comprising:

acquiring an interface protocol command received by the solid state drive;

parsing the interface protocol command to obtain I/O information from the interface protocol command, wherein the I/O information comprises at least one I/O timestamp, an I/O type, and an I/O size; and

invoking machine learning based on the I/O information to predict I/O information of a first future time period, comprising:

preprocessing the I/O information to obtain a first I/O information table stored in the solid state drive;

combining multiple adjacent I/O operations in the first I/O information table into at least one I/O operation;

obtaining I/O prediction results of the first future time period by using at least one long-short term memory (LSTM) neural network with feature values extracted from the at least one combined I/O operation; and

post-processing the I/O prediction results to adjust the I/O prediction results; and

assessing the I/O prediction results by using a first formula to evaluate an accuracy of the I/O prediction results, wherein the first formula comprises:

r

=

∑

i

=

1

n

⁢

(

x

i

-

x

_

)

⁢

(

y

i

-

y

_

)

∑

i

=

1

n

⁢

(

x

i

-

x

_

)

2

⁢

∑

i

=

1

n

⁢

(

y

i

-

y

_

)

2

wherein r represents an accuracy rate, x i represents an actual intensity of the I/O operation, y i represents a predicted I/O intensity, x represents an average value of a set of x i , and y represents an average value of a set of y i , so that a processor of the solid state drive is configured to proactively execute solid state drive management functions according to the I/O prediction results.

8. The computer device of claim 7 , wherein preprocessing the I/O information to obtain the first I/O information table, further comprises:

calculating time intervals between multiple groups of a given number of I/O operations according to the I/O timestamp; and

counting total number of read and write operations for each group of the given number of the I/O operations that occur in the time intervals and the corresponding data sizes to obtain the first I/O information table.

9. The computer device of claim 7 , wherein obtaining the I/O prediction results of the first future time period by using the at least one LSTM neural network with the feature values extracted from the at least one combined I/O operation, further comprises:

inputting the feature values extracted from the at least one combined I/O operation to an input fully connected layer of the at least one LSTM neural network in order to map the feature values to a high dimensional vector space using linear transformations, wherein the feature values are related to the I/O timestamp, the I/O type, and the I/O size;

forming a neuron set with at least one LSTM cell and feeding the feature values transformed from the high dimensional vector space to the neuron set to conduct computation; and

inputting computation result to an output fully connected layer of the at least one LSTM neural network and mapping the computation result to form feature values of an output dimension of a prediction result, wherein the feature values of the output dimension of the prediction result represent the I/O prediction results of the first future time period.

10. The computer device of claim 9 , wherein the feeding the feature values transformed from the high dimensional vector space to the neuron set to conduct computation, further comprises:

feeding the feature values transformed from the high dimensional vector space to network architecture of each LSTM cell to conduct computation, wherein the network architecture comprises an input gate, a forget gate and an output gate.

11. The computer device of claim 9 , wherein a corresponding time interval of the feature values extracted from the at least one combined I/O operation is longer than the first future time period.

12. The computer device of claim 7 , wherein the solid state drive management functions comprises cache management, garbage collection, or command queue management.

13. A non-transitory computer-readable storage medium with a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is configured to implement a data processing method applied to a solid state drive, and the method comprises:

acquiring an interface protocol command received by the solid state drive;

parsing the interface protocol command to obtain I/O information from the interface protocol command, wherein the I/O information comprises at least one I/O timestamp, an I/O type, and an I/O size; and

invoking machine learning based on the I/O information to predict I/O information of a first future time period, comprising:

preprocessing the I/O information to obtain a first I/O information table stored in the solid state drive;

combining multiple adjacent I/O operations in the first I/O information table into at least one I/O operation;

obtaining I/O prediction results of the first future time period by using at least one long-short term memory (LSTM) neural network with feature values extracted from the at least one combined I/O operation; and

post-processing the I/O prediction results to adjust the I/O prediction results; and

assessing the I/O prediction results by using a first formula to evaluate an accuracy of the I/O prediction results, wherein the first formula comprises:

r

=

∑

i

=

1

n

⁢

(

x

i

-

x

_

)

⁢

(

y

i

-

y

_

)

∑

i

=

1

n

⁢

(

x

i

-

x

_

)

2

⁢

∑

i

=

1

n

⁢

(

y

i

-

y

_

)

2

wherein r represents an accuracy rate, x i represents an actual intensity of the I/O operation, y i represents a predicted I/O intensity, x represents an average value of a set of x i , and y represents an average value of a set of y i , so that a processor of the solid state drive is configured to proactively execute solid state drive management functions according to the I/O prediction results.

14. The non-transitory computer-readable storage medium of claim 13 , wherein preprocessing the I/O information to obtain the first I/O information table, further comprises:

calculating time intervals between multiple groups of a given number of I/O operations according to the I/O timestamp; and

counting total number of read and write operations for each group of the given number of the I/O operations that occur in the time intervals and the corresponding data sizes to obtain the first I/O information table.

15. The non-transitory computer-readable storage medium of claim 13 , wherein obtaining the I/O prediction results of the first future time period by using the at least one LSTM neural network with the feature values extracted from the at least one combined I/O operation, further comprises:

inputting the feature values extracted from the at least one combined I/O operation to an input fully connected layer of the at least one LSTM neural network in order to map the feature values to a high dimensional vector space using linear transformations, wherein the feature values are related to the I/O timestamp, the I/O type, and the I/O size;

forming a neuron set with at least one LSTM cell and feeding the feature values transformed from the high dimensional vector space to the neuron set to conduct computation; and

inputting computation result to an output fully connected layer of the at least one LSTM neural network and mapping the computation result to form feature values of an output dimension of a prediction result, wherein the feature values of the output dimension of the prediction result represent the I/O prediction results of the first future time period.