Method and device for predicting operating status, storage medium, electronic device and computer program product

Abstract

Embodiments of the present application relate to the field of communications, and provide a method and device for predicting an operating state, a storage medium, an electronic device, and a computer program product, wherein the method for predicting an operating state includes: obtaining multiple first parameter data of a first central processing unit; determining a first vector cluster and a second vector cluster based on the multiple first parameter data, wherein the first vector cluster is used to indicate the parameter data of the central processing unit in a normal state, and the second vector cluster is used to indicate the parameter data of the central processing unit in a fault state; obtaining second parameter data of a second central processing unit in a solid-state drive, and constructing a first multidimensional vector corresponding to the second parameter data; determining a first similarity between the first multidimensional vector and the first vector cluster, and determining a second similarity between the first multidimensional vector and the second vector cluster; and predicting the operating state of the second central processing unit based on the first similarity and the second similarity.

Description

Technical Field

[0001] Embodiments of the present application relate to the field of communications, and more specifically, to a method and device for predicting an operating state, a storage medium, an electronic device, and a computer program product. Background Art

[0002] With the continuous development of computer technology, multi-core central processing unit (CPU) systems are increasingly used in computers. A multi-core CPU system contains multiple CPUs and can handle multiple concurrent read and write requests at the same time. The system generally has a master CPU and other slave CPUs, and the slave CPUs are activated by the master CPU. If the master CPU fails, the entire system cannot run. If a slave CPU fails, it affects the balance and communication between cores.

[0003] Therefore, the problem that the operating status of the central processing unit cannot be predicted in the related art, resulting in the solid-state drive being unable to operate when any central processing unit fails, has not yet been effectively solved. Summary of the invention

[0004] The embodiments of the present application provide a method and device for predicting an operating status, a storage medium, an electronic device, and a computer program product, so as to at least solve the problem in the related art that the operating status of a central processing unit cannot be predicted, resulting in the inability of a solid-state hard disk to operate when any central processing unit fails.

[0005] According to one embodiment of the present application, a method for predicting an operating status is provided, comprising: obtaining multiple first parameter data of a first central processing unit; determining a first vector cluster and a second vector cluster based on the multiple first parameter data, wherein the first vector cluster is used to indicate the parameter data of the central processing unit in a normal state, and the second vector cluster is used to indicate the parameter data of the central processing unit in a fault state; obtaining second parameter data of a second central processing unit in a solid state drive, and constructing a first multidimensional vector corresponding to the second parameter data; determining a first similarity between the first multidimensional vector and the first vector cluster, and determining a second similarity between the first multidimensional vector and the second vector cluster; predicting the operating status of the second central processing unit based on the first similarity and the second similarity.

[0006] In an exemplary embodiment, obtaining multiple first parameter data of a first central processing unit includes: setting a monitoring thread corresponding to the first central processing unit; and determining a performance parameter set corresponding to the first central processing unit; and determining the multiple first parameter data of the first central processing unit based on the monitoring thread when adjusting the performance parameter of the first central processing unit to each performance parameter in the performance parameter set.

[0007] In an exemplary embodiment, determining a first vector cluster and a second vector cluster based on the multiple first parameter data includes: performing data cleaning on the multiple first parameter data to delete abnormal data in the multiple first parameter data; performing data normalization processing on the multiple first parameter data after data cleaning based on a preset data range to obtain multiple third parameter data; determining each second multidimensional vector corresponding to the first central processor based on each third parameter data; and determining the first vector cluster and the second vector cluster based on the multiple second multidimensional vectors.

[0008] In an exemplary embodiment, determining the first vector cluster and the second vector cluster based on multiple second multidimensional vectors includes: inputting the multiple second multidimensional vectors into a first time series network to obtain a first output value output by an output layer and a second output value output by a hidden layer, wherein the first time series network includes: the output layer and the hidden layer; determining the time correlation between each pair of multidimensional vectors in the multiple second multidimensional vectors based on the first output value and the second output value; and determining the first vector cluster and the second vector cluster based on the time correlation.

[0009] In an exemplary embodiment, before the multiple second multidimensional vectors are input into the first time series network to obtain the first output value of the output layer and the second output value of the hidden layer, it includes: determining a first activation function corresponding to the output layer and a second activation function corresponding to the hidden layer; training the second time series network based on the first activation function and the second activation function to obtain the first time series network.

[0010] In an exemplary embodiment, determining the temporal correlation between each two multidimensional vectors in the plurality of second multidimensional vectors based on the first output value and the second output value includes: determining a loss function corresponding to the first time series network; updating the first output value based on the loss function to generate an updated third output value; and updating the second output value based on the loss function to generate an updated fourth output value; determining the temporal correlation between each two multidimensional vectors in the plurality of second multidimensional vectors based on the third output value and the fourth output value.

[0011] In an exemplary embodiment, determining the first vector cluster and the second vector cluster based on the time correlation includes: determining a distance parameter between each two multidimensional vectors in the multiple second multidimensional vectors; determining a third similarity between each two multidimensional vectors in the multiple second multidimensional vectors based on the distance parameter and the time correlation; and dividing the multiple second multidimensional vectors into the first vector cluster and the second vector cluster based on multiple third similarities.

[0012] In an exemplary embodiment, determining the distance parameter between each two multidimensional vectors in the plurality of second multidimensional vectors includes: determining the distance parameter between each two multidimensional vectors in the plurality of second multidimensional vectors according to a first formula, wherein the first formula is:

[0013] , d is the distance parameter, X n1 , X n2 , …, X nj For every two multidimensional vectors, the first j elements in the vector, X m1 , X m2 , …, X mj are the j elements in the second vector of every two multidimensional vectors, where m, n, and j are all positive integers.

[0014] In an exemplary embodiment, determining the third similarity between each two multidimensional vectors in the plurality of second multidimensional vectors based on the distance parameter and the time correlation includes: acquiring noise data corresponding to the first central processor; and determining a first normalized weight corresponding to the distance parameter and a second normalized weight corresponding to the time correlation; determining the third similarity between each two multidimensional vectors in the plurality of second multidimensional vectors based on the first normalized weight, the second normalized weight and the noise data.

[0015] In an exemplary embodiment, determining a third similarity between every two multidimensional vectors in the plurality of second multidimensional vectors according to the first normalized weight, the second normalized weight, and the noise data comprises: inputting the first normalized weight, the second normalized weight, and the noise data into a second formula to determine the third similarity, wherein the second formula is: , similar is the third similarity, M1 is the first normalized weight, M2 is the second normalized weight, r is the noise data, d is the distance parameter, Ot is the time correlation, A is the credibility corresponding to the distance parameter, and B is the credibility corresponding to the time correlation.

[0016] In an exemplary embodiment, determining the first normalized weight corresponding to the distance parameter includes: determining the first normalized weight by a third formula, wherein the third formula includes: M1=d / (d+Ot), M1 is the first normalized weight, d is the distance parameter, and Ot is the time correlation.

[0017] In an exemplary embodiment, determining a second normalized weight corresponding to the time correlation includes: determining the second normalized weight by a fourth formula, wherein the fourth formula includes: M2=Ot / (d+Ot), M2 is the second normalized weight, d is the distance parameter, and Ot is the time correlation.

[0018] In an exemplary embodiment, constructing a first multidimensional vector corresponding to the second parameter data includes: determining multiple element attributes contained in each first parameter data; determining the multiple element attributes as target element attributes corresponding to the first multidimensional vector; determining multiple elements corresponding to the target element attributes in the second parameter data; and constructing the first multidimensional vector based on the multiple elements.

[0019] In an exemplary embodiment, predicting the operating state of the second central processing unit based on the first similarity and the second similarity includes: determining the size relationship between the first similarity and the second similarity; when the size relationship indicates that the first similarity is greater than the second similarity, determining that the operating state of the second central processing unit is a normal state; when the size relationship indicates that the first similarity is less than or equal to the second similarity, determining that the operating state of the second central processing unit is an abnormal state.

[0020] In an exemplary embodiment, after predicting the operating state of the second central processing unit based on the first similarity and the second similarity, the method further includes: when it is determined that the operating state of the second central processing unit is normal, saving the first multidimensional vector; and updating the first vector cluster based on the first multidimensional vector.

[0021] In an exemplary embodiment, after predicting the operating state of the second central processing unit based on the first similarity and the second similarity, the method also includes: when it is determined that the operating state of the second central processing unit is an abnormal state, determining whether there is a first request being processed in the second central processing unit; when it is determined that the second central processing unit has the first request, controlling the second central processing unit to execute a first operation corresponding to the first request, and determining whether the first operation is completed; when it is determined that the first operation is completed, starting the target process of the solid-state drive; and executing the target operation on the second central processing unit according to the target process to prohibit the second central processing unit from receiving the request sent by the solid-state drive.

[0022] In an exemplary embodiment, after executing the target operation on the second central processing unit according to the target process, the method also includes: when it is determined that the operating state of the second central processing unit is an abnormal state, obtaining multiple second requests sent by the solid-state drive; determining whether the multiple second requests include the target request to be sent to the second central processing unit; when it is determined that the multiple second requests include the target request, sending the target request to a third central processing unit so that the third central processing unit executes the second operation corresponding to the target request.

[0023] In an exemplary embodiment, after executing the target operation on the second central processor according to the target process, the method also includes: terminating communication between the second central processor and other central processors, wherein the other central processors include: central processors other than the second central processor in the solid-state drive; marking the first multidimensional vector and saving the marked first multidimensional vector; and updating the second vector cluster according to the first multidimensional vector.

[0024] According to another embodiment of the present application, a device for predicting an operating status is provided, including: a first acquisition module, used to acquire multiple first parameter data of a first central processing unit; a first determination module, used to determine a first vector cluster and a second vector cluster based on the multiple first parameter data, wherein the first vector cluster is used to indicate the parameter data of the central processing unit in a normal state, and the second vector cluster is used to indicate the parameter data of the central processing unit in a fault state; the second acquisition module is also used to acquire the second parameter data of the second central processing unit in the solid state drive, and construct a first multidimensional vector corresponding to the second parameter data; the second determination module is also used to determine a first similarity between the first multidimensional vector and the first vector cluster, and determine a second similarity between the first multidimensional vector and the second vector cluster; the prediction module is also used to predict the operating status of the second central processing unit based on the first similarity and the second similarity.

[0025] According to another embodiment of the present application, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the steps in the above method embodiment when running.

[0026] According to another embodiment of the present application, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in the above method embodiment.

[0027] According to another embodiment of the present application, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiment are implemented.

[0028] Through the prediction method of the running state of the present application, a plurality of first parameter data in the first central processor are obtained, and a first vector cluster for indicating the parameter data of the central processor in a normal state and a second vector cluster for indicating the parameter data of the central processor in an abnormal state are determined according to the first parameter data. The second parameter data of the second central processor in the solid state drive is obtained, and the first multidimensional vector corresponding to the second parameter data is constructed, and the first similarity and the second similarity of the first multidimensional vector with the first vector cluster and the second vector cluster are respectively determined; the running state of the second central processor is predicted according to the first similarity and the second similarity. According to the embodiment of the present application, the parameter data is divided into a first vector cluster and a second vector cluster by a plurality of first parameter data in the first central processor, and then the motion state of the second central processor can be predicted by comparing the similarity of the first multidimensional vector corresponding to the second central processor with the first vector cluster and the second vector cluster. Through the above embodiment, the problem that the running state of the central processor cannot be predicted in the related art, resulting in the inability of the solid state drive to operate in the case of any central processor failure, can be solved. Then the running state of the core storage is predicted, and in the case of the possibility of failure of any core storage, any core storage can be processed to avoid the situation that the solid state drive cannot operate due to the failure of any core storage. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a hardware structure block diagram of a computer terminal of a method for predicting an operating state according to an embodiment of the present application;

[0030] Figure 2 is a flow chart of a method for predicting an operating state according to an embodiment of the present application;

[0031] Figure 3 is a flow chart of a method for predicting processor failures in a solid-state hard disk multi-core system according to an optional embodiment of the present application;

[0032] Figure 4 It is a structural block diagram of a device for predicting processor failures in a solid-state hard disk multi-core system according to an optional embodiment of the present application;

[0033] Figure 5 is an architecture diagram of a system for processor fault prediction for a solid state drive multi-core system according to an optional embodiment of the present application;

[0034] Figure 6 is a time series network structure diagram according to an optional embodiment of the present application;

[0035] Figure 7 It is a structural block diagram of a device for predicting operating status according to an embodiment of the present application. DETAILED DESCRIPTION

[0036] The embodiments of the present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0037] It should be noted that the terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0038] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a computer terminal as an example, Figure 1 1 is a hardware structure block diagram of a computer terminal of a method for predicting an operating state according to an embodiment of the present application. Figure 1 As shown, the computer terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the above-mentioned computer terminal may also include a transmission device 106 and an input and output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above-mentioned computer terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0039] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the method for determining the interactive state in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories can be connected to the computer terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0040] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of a computer terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0041] Figure 2 is a flow chart of a method for predicting an operating state according to an embodiment of the present application, such as Figure 2 As shown, the process includes the following steps:

[0042] Step S202, obtaining a plurality of first parameter data of the first central processing unit;

[0043] The plurality of first parameter data may be parameter data such as CPU usage rate, temperature, voltage, etc. of the first CPU.

[0044] Step S204, determining a first vector cluster and a second vector cluster according to the plurality of first parameter data, wherein the first vector cluster is used to indicate parameter data of the central processing unit in a normal state, and the second vector cluster is used to indicate parameter data of the central processing unit in a fault state;

[0045] Step S206, obtaining second parameter data of a second central processor in the solid state drive, and constructing a first multidimensional vector corresponding to the second parameter data;

[0046] Step S208, determining a first similarity between the first multidimensional vector and the first vector cluster, and determining a second similarity between the first multidimensional vector and the second vector cluster;

[0047] Step S210: predicting the operating state of the second central processing unit according to the first similarity and the second similarity.

[0048] Through the prediction method of the running state of the present application, a plurality of first parameter data in the first central processor are obtained, and a first vector cluster for indicating the parameter data of the central processor in a normal state and a second vector cluster for indicating the parameter data of the central processor in an abnormal state are determined according to the first parameter data. The second parameter data of the second central processor in the solid state drive is obtained, and the first multidimensional vector corresponding to the second parameter data is constructed, and the first similarity and the second similarity of the first multidimensional vector with the first vector cluster and the second vector cluster are respectively determined; the running state of the second central processor is predicted according to the first similarity and the second similarity. According to the embodiment of the present application, the parameter data is divided into a first vector cluster and a second vector cluster by a plurality of first parameter data in the first central processor, and then the motion state of the second central processor can be predicted by comparing the similarity of the first multidimensional vector corresponding to the second central processor with the first vector cluster and the second vector cluster. Through the above embodiment, the problem that the running state of the central processor cannot be predicted in the related art, resulting in the inability of the solid state drive to operate in the case of any central processor failure, can be solved. Then the running state of the core storage is predicted, and in the case of the possibility of failure of any core storage, any core storage can be processed to avoid the situation that the solid state drive cannot operate due to the failure of any core storage.

[0049] Optionally, the above-mentioned step S202 of obtaining multiple first parameter data of the first central processing unit includes: setting a monitoring thread corresponding to the first central processing unit; and determining a performance parameter set corresponding to the first central processing unit; when adjusting the performance parameter of the first central processing unit to each performance parameter in the performance parameter set, determining the multiple first parameter data of the first central processing unit based on the monitoring thread.

[0050] That is to say, a plurality of first parameter data can be obtained according to the life test, specifically:

[0051] A monitoring thread is set on each first CPU, and the monitoring thread can be responsible for collecting information of the first core processor, such as voltage and temperature of the first CPU, etc. The monitoring thread can also collect information on communication and synchronization between the CPUs.

[0052] During the life test, the performance parameters of the first central processor can be controlled respectively, for example, the first central processor usage rate, temperature, voltage, communication frequency and synchronization delay of the first central processor can be controlled. When different performance parameters are adjusted, the parameter data of the first central processor is monitored through the monitoring thread to obtain a plurality of first parameter data.

[0053] Optionally, the above-mentioned step S204 of determining the first vector cluster and the second vector cluster according to the multiple first parameter data includes: performing data cleaning on the multiple first parameter data to delete abnormal data in the multiple first parameter data; performing data normalization processing on the multiple first parameter data after data cleaning based on a preset data range to obtain multiple third parameter data; determining each second multidimensional vector corresponding to the first central processor according to each third parameter data; and determining the first vector cluster and the second vector cluster according to the multiple second multidimensional vectors.

[0054] It can be understood that to construct the first vector cluster and the second vector cluster, it is necessary to first pre-process the multiple first parameter data, specifically: perform data cleaning on the multiple first parameter data, including: removing outliers, missing values, etc.; perform data normalization on the cleaned multiple first parameter data to obtain multiple third parameter data, that is, put the cleaned multiple first parameter data into a unified range, for example: between 0 and 1. Then, the second multidimensional vector of the first central processing unit is determined based on the pre-processed multiple third parameter data, specifically: the mean, variance, maximum value, minimum value of the central processing unit usage rate, the mean and change rate of the temperature can be determined based on the third parameter data; the mean and fluctuation of the voltage can be used as the characteristics of the data vector to form the second multidimensional vector.

[0055] Wherein, determining the first vector cluster and the second vector cluster according to a plurality of second multidimensional vectors comprises: inputting the plurality of second multidimensional vectors into a first time series network to obtain a first output value output by an output layer and a second output value output by a hidden layer, wherein the first time series network comprises: the output layer and the hidden layer; determining the time correlation between each pair of multidimensional vectors in the plurality of second multidimensional vectors according to the first output value and the second output value; and determining the first vector cluster and the second vector cluster according to the time correlation.

[0056] It is understandable that, considering that the data of each moment of the feature data is associated with the data of the previous moment, when calculating the similarity between vectors, it is necessary to combine the time series network to determine the time correlation, and then determine the first vector cluster and the second vector cluster.

[0057] Specifically: the second multidimensional vector can be input into the first time series network to obtain the first output value corresponding to the output layer of the first time series network and the second output value of the hidden layer, and then the time correlation can be determined according to the first output value and the second output value, and then the first vector cluster and the second vector cluster can be determined.

[0058] It is understandable that determining the temporal correlation between two multi-dimensional vectors can improve the accuracy of similarity calculation.

[0059] Wherein, before the multiple second multidimensional vectors are input into the first time series network to obtain the first output value of the output layer and the second output value of the hidden layer, it includes: determining the first activation function corresponding to the output layer and the second activation function corresponding to the hidden layer; training the second time series network based on the first activation function and the second activation function to obtain the first time series network.

[0060] According to the above technical solution, before inputting the second multidimensional vector into the first time series network, the original second time series network needs to be trained to obtain the trained first time series network. Specifically:

[0061] The first activation function corresponding to the output layer may be determined, the second activation function may be a Rectified Linear Unit Activation Function (ReLU activation function for short), and the first activation function may be a softmax function (a type of activation function).

[0062] By training the first time series network through the first activation function and the second activation function, the generalization ability of the first time series network can be guaranteed and the first time series network can have better fitting ability.

[0063] Wherein, determining the time correlation between each two multidimensional vectors in the plurality of second multidimensional vectors according to the first output value and the second output value includes: determining the loss function corresponding to the first time series network; updating the first output value according to the loss function to generate an updated third output value; and updating the second output value according to the loss function to generate an updated fourth output value; determining the time correlation between each two multidimensional vectors in the plurality of second multidimensional vectors according to the third output value and the fourth output value.

[0064] It is understandable that the loss function corresponding to the first time series network can be determined, and then the parameters of the first time series network can be adjusted according to the loss function, so that the parameters of the first time series network are adjusted to the optimal value, thereby improving the performance of the first time series network. Then, according to the adjusted first time series network, the first output value is updated to the third output value, and the second output value is updated to the fourth output value. Then, according to the updated third output value and the updated fourth output value, the time correlation between each two multidimensional vectors is determined.

[0065] Wherein, determining the first vector cluster and the second vector cluster according to the time correlation includes: determining a distance parameter between each two multidimensional vectors in the multiple second multidimensional vectors; determining a third similarity between each two multidimensional vectors in the multiple second multidimensional vectors according to the distance parameter and the time correlation; and dividing the multiple second multidimensional vectors into the first vector cluster and the second vector cluster based on multiple third similarities.

[0066] In an exemplary embodiment, determining the distance parameter between each two multidimensional vectors in the plurality of second multidimensional vectors includes: determining the distance parameter between each two multidimensional vectors in the plurality of second multidimensional vectors according to a first formula, wherein the first formula is:

[0067] , d is the distance parameter, X n1 , X n2 , …, X nj For every two multidimensional vectors, the first j elements in the vector, X m1 , X m2 , …, X mj are the j elements in the second vector of every two multidimensional vectors, where m, n, and j are all positive integers.

[0068] It is understandable that the distance parameter between every two multidimensional vectors can be determined by the above first formula, and the distance parameter can be calculated, and then the first vector cluster and the second vector cluster can be determined according to the distance parameter and the time correlation to improve the accuracy of dividing the first vector cluster and the second vector cluster.

[0069] Optionally, determining the third similarity between each two multidimensional vectors in the multiple second multidimensional vectors based on the distance parameter and the time correlation includes: obtaining noise data corresponding to the first central processor; and determining a first normalized weight corresponding to the distance parameter and a second normalized weight corresponding to the time correlation; determining the third similarity between each two multidimensional vectors in the multiple second multidimensional vectors based on the first normalized weight, the second normalized weight and the noise data.

[0070] Wherein, determining the third similarity between each two multidimensional vectors in the plurality of second multidimensional vectors according to the first normalized weight, the second normalized weight and the noise data comprises: inputting the first normalized weight, the second normalized weight and the noise data into a second formula to determine the third similarity, wherein the second formula is: , similar is the third similarity, M1 is the first normalized weight, M2 is the second normalized weight, r is the noise data, d is the distance parameter, Ot is the time correlation, A is the credibility corresponding to the distance parameter, and B is the credibility corresponding to the time correlation.

[0071] Among them, determining the first normalized weight corresponding to the distance parameter includes: determining the first normalized weight through a third formula, wherein the third formula includes: M1=d / (d+Ot), M1 is the first normalized weight, d is the distance parameter, and Ot is the time correlation.

[0072] Among them, determining the second normalized weight corresponding to the time correlation includes: determining the second normalized weight through a fourth formula, wherein the fourth formula includes: M2=Ot / (d+Ot), M2 is the second normalized weight, d is the distance parameter, and Ot is the time correlation.

[0073] It can be understood that the technical solution for determining the third similarity between every two multidimensional vectors includes: obtaining the noise data r corresponding to the first central processor, and determining the first normalized weight corresponding to the distance parameter: M1=d / (d+Ot), and the second normalized weight: M2=Ot / (d+Ot).

[0074] Then, the first normalized weight and the second normalized weight are substituted into the second formula to determine the third similarity. According to the above method, the first normalized weight and the second normalized weight corresponding to the distance parameter are determined, and then the third similarity is determined according to the first normalized weight and the second normalized weight, which can improve the accuracy of the third similarity.

[0075] Optionally, the above-mentioned step S206 of constructing the first multidimensional vector corresponding to the second parameter data includes: determining multiple element attributes contained in each first parameter data; determining the multiple element attributes as target element attributes corresponding to the first multidimensional vector; determining multiple elements corresponding to the target element attributes in the second parameter data; and constructing the first multidimensional vector based on the multiple elements.

[0076] It is understandable that the multiple element attributes may include: the temperature and voltage of the CPU, etc. The multiple elements may be: the mean value and the rate of change of the temperature; the mean value and the fluctuation of the voltage, etc.

[0077] Optionally, the above-mentioned step S210 of predicting the operating state of the second central processing unit according to the first similarity and the second similarity includes: determining the size relationship between the first similarity and the second similarity; when the size relationship indicates that the first similarity is greater than the second similarity, determining that the operating state of the second central processing unit is a normal state; when the size relationship indicates that the first similarity is less than or equal to the second similarity, determining that the operating state of the second central processing unit is an abnormal state.

[0078] It can be understood that the greatest similarity is determined between the first similarity and the second similarity, and the operating state corresponding to the greatest similarity is determined as the operating state predicted by the second central processing unit.

[0079] Optionally, after predicting the operating state of the second central processing unit according to the first similarity and the second similarity in step S208, the method further includes:

[0080] (1) When it is determined that the operating state of the second central processing unit is normal, saving the first multi-dimensional vector; and updating the first vector cluster according to the first multi-dimensional vector.

[0081] It is understandable that the operating state of the second central processor may be normal. When it is determined that the operating state of the second central processor is normal, the first multidimensional vector can be saved. On the one hand, when a vector identical to the first multidimensional vector is received, the operating state of the central processor can be directly determined without calculation; on the other hand, the first vector cluster can be updated by the first multidimensional vector, making the first vector cluster more accurate.

[0082] (2) When it is determined that the operating state of the second CPU is abnormal, the following solutions may be executed:

[0083] 1) Determine whether there is a first request being processed in the second central processing unit; if it is determined that the second central processing unit has the first request, control the second central processing unit to execute a first operation corresponding to the first request, and determine whether the first operation is completed; if it is determined that the first operation is completed, start the target process of the solid state drive; perform a target operation on the second central processing unit according to the target process to prohibit the second central processing unit from receiving the request sent by the solid state drive.

[0084] It can be understood that, when it is determined that the second central processing unit is in an abnormal state, the bad core that is about to fail is prevented from continuing to receive requests, and only the request currently being processed by the second central processing unit needs to be completed.

[0085] 2) Acquire multiple second requests sent by the solid-state drive; determine whether the multiple second requests include a target request to be sent to the second central processing unit; if it is determined that the multiple second requests include the target request, send the target request to a third central processing unit so that the third central processing unit executes a second operation corresponding to the target request.

[0086] Wherein, after executing the target operation on the second central processor according to the target process, the method further includes: terminating the communication between the second central processor and other central processors, wherein the other central processors include: central processors other than the second central processor in the solid state drive; marking the first multidimensional vector and saving the marked first multidimensional vector; and updating the second vector cluster according to the first multidimensional vector.

[0087] It is understandable that the CPU that needs to notify other normal CPUs of the impending abnormality will no longer send task messages to the second CPU, and cut off the communication between the normal CPU and the CPU.

[0088] In addition, when it is determined that the operating state of the second central processing unit is abnormal, the first multidimensional vector can be saved. On the one hand, when a vector identical to the first multidimensional vector is received, it is possible to directly determine that the operating state of the central processing unit is abnormal without calculation, and take corresponding measures; on the other hand, the second vector cluster can be updated by the first multidimensional vector, making the second vector cluster more accurate.

[0089] In order to better understand the process of the above-mentioned running state prediction method, the implementation process of the above-mentioned running state prediction method is described below in combination with an optional embodiment, but it is not used to limit the technical solution of the embodiment of the present application.

[0090] Currently, CPU fault handling is all done at the SSD level. In a multi-disk system, two hard disks are pre-assigned to receive each read and write request. When the first hard disk fails, the second hard disk completes the request. Or when the host crashes, the hard disk is replaced.

[0091] The above technical solutions may have the following problems: using a multi-disk system can protect data and prevent data loss, but it requires more disks to be prepared, which is costly. Replacing hard disks after a downtime is not forward-looking and may lead to data loss if problems are not solved. The entire system lacks stability and reliability.

[0092] In response to the above technical problems, the optional embodiment of the present application provides a method, device and system for predicting processor failures in a solid-state hard disk multi-core system. Specifically, by conducting a processor life experiment, the collected historical data is saved and a vector space is established. The real-time collected data is used to construct a vector, and the similarity is calculated with the vector in the vector space to predict whether it is similar to the vector in the vector space, so as to achieve self-repair of the hard disk before a failure occurs, save costs and improve the stability and reliability of the computer system.

[0093] In a first aspect, an optional embodiment of the present application provides a method for predicting processor failures in a solid-state hard disk multi-core system. Figure 3 is a flowchart of a method for predicting processor failures in a solid-state hard disk multi-core system according to an optional embodiment of the present application. Figure 3 As shown, the method includes:

[0094] Step S301, performing a processor life experiment, collecting experimental data through a monitoring module, and transmitting the data to a computing module through a communicator;

[0095] Step S302, the calculation module calculates and integrates the collected experimental data, constructs an N-dimensional vector space, and classifies and saves it in the vector module;

[0096] Step S303, the monitoring module collects relevant information of the current solid state drive in real time and transmits it to the computing module;

[0097] Step S304, the calculation module constructs a vector for the real-time information, and performs similarity calculation with the vector saved by the vector module;

[0098] Step S305: if the real-time information has a high similarity with the category of no failure, the vector of the real-time information is saved in the vector module; if the real-time information has a high similarity with the category of failure, it means that the probability of failure is very high, and the core reduction mechanism of the solid state drive (SSD) is activated for self-repair;

[0099] In a second aspect, an optional embodiment of the present application provides a device for predicting processor failures in a solid-state hard disk multi-core system. Figure 4 is a structural block diagram of a device for predicting processor failures in a solid-state hard disk multi-core system according to an optional embodiment of the present application, such as Figure 4 As shown, the device comprises:

[0100] The monitoring module 42 is independent of each CPU and is used to obtain relevant data information by setting a monitoring thread in the hard disk and each CPU;

[0101] The calculation module 44 integrates the data information, integrates the vectors, and constructs an N (used to indicate the dimension of the vector) dimensional vector;

[0102] Vector module 46, classification and storage of data vector space, and calculation of similarity between vectors;

[0103] Repair module 48, if the probability of failure is high, activates the SSD self-repair mechanism;

[0104] In a third aspect, an optional embodiment of the present application provides a system for predicting processor failures in a solid-state hard disk multi-core system. Figure 5 is an architecture diagram of a system for predicting processor failures in a multi-core solid-state drive system according to an optional embodiment of the present application, such as Figure 5 As shown, the system includes:

[0105] Controller 52: responsible for controlling the operation of the entire system, including coordinating the communication between the monitor and the information processor;

[0106] Monitor 54: includes a monitoring module to monitor the load status of each CPU;

[0107] Information processor 56: includes a calculation module and a vector module, constructs vectors for corresponding information, saves vectors and calculates similarities between vectors;

[0108] Communicator 58: completes internal communication between the monitoring layer, information processing layer and each CPU;

[0109] Multiple CPUs 60: complete read and write requests;

[0110] Repairer 62: includes a repair module. If the self-repair mechanism is triggered, it will schedule between CPUs and shield bad cores.

[0111] like Figure 3 , Figure 4 , Figure 5 As shown, the specific steps of the optional embodiment of the present application include:

[0112] (1) Conduct processor life experiments, collect experimental data through the monitoring module, and transmit it to the computing module through the communicator;

[0113] A monitoring thread is set on each CPU (equivalent to the first central processing unit of this application), which is responsible for collecting CPU information (CPU information is equivalent to multiple first parameter data of this application), including CPU usage, temperature, voltage and other parameters. A monitoring thread is set in the SSD, which is responsible for collecting information on communication and synchronization between each CPU. This information can be transmitted to the monitoring module in real time through the internal communication mechanism.

[0114] In an optional embodiment of the present application, a CPU life experiment is conducted, and raw data is collected by controlling the CPU usage rate, temperature, voltage, communication frequency, and synchronization delay (equivalent to the performance parameter set of the present application). After a CPU fails, the time of the failure is recorded, and the parameter value at which the CPU performance begins to decline before the failure is found. The data at the beginning of the decline is regarded as the data of the failure, and the data before that is regarded as normal data.

[0115] The monitoring module is independent of each CPU and collects performance monitoring data of the multi-core CPU system and the communication and synchronization between each CPU. The parameter data of each CPU can monitor the current CPU usage, communication and synchronization between cores, and can monitor whether the CPU is running normally by comparing with the synchronization of other cores. The monitoring module transmits it to the computing module through the communication mechanism.

[0116] (2) The computing module integrates the collected experimental data, constructs an N-dimensional vector space, and saves it in the vector module. Each piece of detection information from the monitoring module is transmitted to the computing module. During the vector space construction phase, the computing module uses all previous information as a source of data. The data entries consist of CPU parameters, inter-core communication, and synchronization characteristics.

[0117] Constructing vector data is divided into two steps. The first is that the collected raw data (that is, the multiple first parameter data of this application) may contain noise or irregular parts and need to be preprocessed. This includes data cleaning (removing outliers, missing values, etc.) and data normalization (scaling the data to a uniform range, such as between 0-1). The second is to calculate the mean, variance, maximum and minimum values ​​of the CPU usage; the mean and rate of change of the temperature; the mean and fluctuation of the voltage. The communication frequency and synchronization delay between cores. Take these as the characteristics of the vector data, so that each piece of data is a 10-dimensional vector (that is, the second multidimensional vector of this application);

[0118] In the classification of vector data, it is necessary to calculate the similarity of each vector, that is, each piece of data. Considering that the data at each moment of the feature data is actually related to the data at the previous moment, when calculating the similarity between vectors, it is necessary to combine the time series network (that is, the first time series network of this application) to calculate the time correlation, increase the attention to recent data, and improve the accuracy of similarity calculation.

[0119] Figure 6 is a time series network structure diagram according to an optional embodiment of the present application, such as Figure 6As shown, in the time series network, a hidden layer is used. Xt is used as the input of the input layer, St is the output of the hidden layer, and Ot is the output of the output layer (the output of the output layer can indicate the temporal correlation). U is the weight used for the output from the input layer to the hidden layer, b1 is the bias unit used for the output from the input layer to the hidden layer, V is the weight used for the output from the hidden layer to the output layer, b2 is the bias unit used for the output from the hidden layer to the output layer, W represents the parameter weight used for the output of the previous time series and the next time series, t represents time, and the calculation formula of the output layer is: Qt=g(V*St+b2); .

[0120] Among them, f is the activation function of the hidden layer. In order to ensure the generalization ability and better fitting ability of the network, the nonlinear activation function ReLU (equivalent to the second activation function of this application) is selected here. Its characteristics are fast convergence and simple gradient calculation, that is: f=ReLU(x).

[0121] g is the activation function of the output layer, and the softmax function is used to directly output the predicted probability value: (equivalent to the first activation function of this application).

[0122] In the time series network model, the activation function of the hidden layer is the ReLU function, and the activation function of the output layer is the softmax function (an activation function). In this way, there is a loss function L(t) at each position of the sequence. L(t) uses the log-likelihood loss function. The final loss function of the entire model (equivalent to the loss function of this application) is the sum of the loss functions, that is: .

[0123] The loss function can be used to deduce the parameter values.

[0124] After calculating the time correlation Ot between two vector data according to the above time series network, it is also necessary to calculate the distance parameter of the two vector data in the entire vector space:

[0125] Among them, Xn1, Xn2, ..., Xn10 and Xm1, Xm2, Xm10 represent two vectors respectively. In all vector data, two vectors are calculated, and the 10-dimensional features in these two vectors are calculated.

[0126] The calculated distance parameter d is combined with the time correlation Ot to calculate the similarity between the two vectors: .

[0127] Among them, M1=d / (d+Ot) and M2=Ot / (d+Ot) are the normalization of the two parameters, A and B represent the credibility corresponding to the parameters, and r represents noise.

[0128] According to the similarity, all vector data are automatically divided into two vector clusters, one cluster is normal data (equivalent to the first vector cluster of the present application), and the other cluster is fault data (equivalent to the second vector cluster of the present application).

[0129] (3) The monitoring module collects relevant information of the current solid-state drive in real time and transmits it to the computing module.

[0130] After completing the construction of the vector space, the entire vector space and time series vector are saved in the vector module to perform real-time prediction of the SSD CPU. Specifically, the CPU information is collected in real time, including parameters such as CPU usage, temperature, voltage, and information on communication and synchronization between CPUs. This information can be transmitted to the calculation module in real time through the internal communication mechanism.

[0131] (4) The calculation module constructs a vector for the real-time information and calculates the similarity with the vector saved by the vector module.

[0132] It is necessary to calculate the similarity between the real-time information and the two vector clusters in the vector space. According to the calculation rules in the calculation module, the corresponding time correlation parameters and distance parameters are calculated respectively, and finally the similarity is determined. According to the size of the similarity, the vectors are divided into clusters with large similarity.

[0133] (5) If the real-time information has a high similarity with the non-failure category, the vector of the real-time information is saved in the vector module. If the real-time information has a high similarity with the failure category, it means that the probability of failure is very high, and the SSD core reduction mechanism is activated for self-repair.

[0134] The CPU fault prediction is performed on the CPU related data information received in real time. If the real-time information is highly similar to the non-failure type, the vector of the real-time information is saved to the vector module. If the similarity is high to the failure type, it means that the probability of failure is very high. It is judged that the CPU is about to fail, and the repairer is started. The core reduction mechanism of the SSD is used for self-repair and bad cores are shielded to ensure the normal operation of the SSD.

[0135] Furthermore, the specific working principle of the repairer includes:

[0136] 1) Prevent the failed core from receiving any more requests and only need to complete the current requests;

[0137] 2) Notify other normal cores that a bad core is about to appear, stop sending task messages to the bad core, and cut off the communication between normal cores and the bad core.

[0138] 3) For requests sent from the host to the bad core, the repairer is used to transfer them between cores and transfer them to the normal core for execution.

[0139] The optional embodiment of the present application provides a method for predicting processor failures in a solid-state hard disk multi-core system, which performs a processor life experiment, saves the collected historical data, and establishes a vector space. The real-time collected data is used to construct a vector, and the similarity is calculated with the vector in the vector space to predict whether the hard disk can be self-repaired before a failure occurs, saving costs and improving the stability and reliability of the computer system.

[0140] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (e.g., a disk, an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0141] In the present embodiment, a prediction device for an operating state is also provided, and the device is used to implement the above-mentioned embodiment and the preferred implementation mode, and the descriptions that have been made are not repeated. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware is also possible and conceived.

[0142] Figure 7 is a structural block diagram of a device for predicting the operating status according to an embodiment of the present application, such as Figure 7 As shown, the device comprises:

[0143] A first acquisition module 72, used to acquire a plurality of first parameter data of a first central processing unit;

[0144] A first determination module 74, configured to determine a first vector cluster and a second vector cluster according to the plurality of first parameter data, wherein the first vector cluster is used to indicate parameter data of the central processing unit in a normal state, and the second vector cluster is used to indicate parameter data of the central processing unit in a fault state;

[0145] The second acquisition module 76 is further used to acquire second parameter data of the second central processor in the solid state drive, and construct a first multidimensional vector corresponding to the second parameter data;

[0146] The second determination module 78 is further configured to determine a first similarity between the first multidimensional vector and the first vector cluster, and to determine a second similarity between the first multidimensional vector and the second vector cluster;

[0147] The prediction module 80 is further configured to predict the operating state of the second central processing unit according to the first similarity and the second similarity.

[0148] Through the running state prediction device of the present application, a plurality of first parameter data in the first central processor are obtained, and a first vector cluster for indicating the parameter data of the central processor in a normal state and a second vector cluster for indicating the parameter data of the central processor in an abnormal state are determined according to the first parameter data. The second parameter data of the second central processor in the solid state drive is obtained, and the first multidimensional vector corresponding to the second parameter data is constructed, and the first similarity and the second similarity of the first multidimensional vector with the first vector cluster and the second vector cluster are respectively determined; the running state of the second central processor is predicted according to the first similarity and the second similarity. According to the embodiment of the present application, the parameter data is divided into a first vector cluster and a second vector cluster by using a plurality of first parameter data in the first central processor, and then the motion state of the second central processor can be predicted by comparing the similarity of the first multidimensional vector corresponding to the second central processor with the first vector cluster and the second vector cluster. Through the above embodiment, the problem that the running state of the central processor cannot be predicted in the related art, resulting in the inability of the solid state drive to operate in the case of any central processor failure, can be solved. Then the running state of the core storage is predicted, and in the case of the possibility of failure of any core storage, any core storage can be processed to avoid the situation that the solid state drive cannot operate due to the failure of any core storage.

[0149] In an exemplary embodiment, the first acquisition module 72 is also used to set a monitoring thread corresponding to the first central processor; and determine a performance parameter set corresponding to the first central processor; when the performance parameter of the first central processor is adjusted to each performance parameter in the performance parameter set, the multiple first parameter data of the first central processor are determined based on the monitoring thread.

[0150] In an exemplary embodiment, the first determination module 74 is also used to perform data cleaning on the multiple first parameter data to delete abnormal data in the multiple first parameter data; perform data normalization processing on the multiple first parameter data after data cleaning based on a preset data range to obtain multiple third parameter data; determine each second multidimensional vector corresponding to the first central processor according to each third parameter data; determine the first vector cluster and the second vector cluster according to the multiple second multidimensional vectors.

[0151] In an exemplary embodiment, the first determination module 74 is further used to input the multiple second multidimensional vectors into a first time series network to obtain a first output value output by the output layer and a second output value output by the hidden layer, wherein the first time series network includes: the output layer and the hidden layer; determining the time correlation between each pair of the multiple second multidimensional vectors according to the first output value and the second output value; and determining the first vector cluster and the second vector cluster according to the time correlation.

[0152] In an exemplary embodiment, the first determination module 74 is further used to determine a first activation function corresponding to the output layer and a second activation function corresponding to the hidden layer; and train the second time series network based on the first activation function and the second activation function to obtain the first time series network.

[0153] In an exemplary embodiment, the first determination module 74 is also used to determine the loss function corresponding to the first time series network; update the first output value according to the loss function to generate an updated third output value; and update the second output value according to the loss function to generate an updated fourth output value; determine the time correlation between each two multidimensional vectors in the multiple second multidimensional vectors according to the third output value and the fourth output value.

[0154] In an exemplary embodiment, the first determination module 74 is further used to determine a distance parameter between each two multidimensional vectors in the multiple second multidimensional vectors; determine a third similarity between each two multidimensional vectors in the multiple second multidimensional vectors based on the distance parameter and the time correlation; and divide the multiple second multidimensional vectors into the first vector cluster and the second vector cluster based on multiple third similarities.

[0155] In an exemplary embodiment, the first determination module 74 is further configured to determine a distance parameter between every two multidimensional vectors in the plurality of second multidimensional vectors according to a first formula, wherein the first formula is: , d is the distance parameter, X n1 , X n2 , …, X nj For every two multidimensional vectors, the first j elements in the vector, X m1 , X m2 , …, X mj are the j elements in the second vector of every two multidimensional vectors, where m, n, and j are all positive integers.

[0156] In an exemplary embodiment, the first determination module 74 is also used to obtain noise data corresponding to the first central processor; and determine a first normalized weight corresponding to the distance parameter and a second normalized weight corresponding to the time correlation; and determine a third similarity between each two multidimensional vectors in the multiple second multidimensional vectors based on the first normalized weight, the second normalized weight and the noise data.

[0157] In an exemplary embodiment, the first determination module 74 is further configured to input the first normalized weight, the second normalized weight and the noise data into a second formula to determine the third similarity, wherein the second formula is: , similar is the third similarity, M1 is the first normalized weight, M2 is the second normalized weight, r is the noise data, d is the distance parameter, Ot is the time correlation, A is the credibility corresponding to the distance parameter, and B is the credibility corresponding to the time correlation.

[0158] In an exemplary embodiment, the first determination module 74 is further used to determine the first normalized weight through a third formula, wherein the third formula includes: M1=d / (d+Ot), M1 is the first normalized weight, d is the distance parameter, and Ot is the time correlation.

[0159] In an exemplary embodiment, the first determination module 74 is further used to determine the second normalized weight through a fourth formula, wherein the fourth formula includes: M2=Ot / (d+Ot), M2 is the second normalized weight, d is the distance parameter, and Ot is the time correlation.

[0160] In an exemplary embodiment, the first determination module 64 is further used to determine multiple element attributes contained in each first parameter data; determine the multiple element attributes as target element attributes corresponding to the first multidimensional vector; determine multiple elements corresponding to the target element attributes in the second parameter data; and construct the first multidimensional vector based on the multiple elements.

[0161] In an exemplary embodiment, the prediction module 80 is also used to determine the size relationship between the first similarity and the second similarity; when the size relationship indicates that the first similarity is greater than the second similarity, it is determined that the operating state of the second central processing unit is a normal state; when the size relationship indicates that the first similarity is less than or equal to the second similarity, it is determined that the operating state of the second central processing unit is an abnormal state.

[0162] In an exemplary embodiment, the prediction module 80 is further configured to save the first multi-dimensional vector when it is determined that the operating state of the second central processing unit is normal; and update the first vector cluster according to the first multi-dimensional vector.

[0163] In an exemplary embodiment, the prediction module 80 is also used to determine whether there is a first request being processed in the second central processor when it is determined that the operating state of the second central processor is an abnormal state; when it is determined that the second central processor has the first request, control the second central processor to execute a first operation corresponding to the first request, and determine whether the first operation is completed; when it is determined that the first operation is completed, start the target process of the solid-state drive; and perform a target operation on the second central processor according to the target process to prohibit the second central processor from receiving the request sent by the solid-state drive.

[0164] In an exemplary embodiment, the prediction module 80 is also used to obtain multiple second requests sent by the solid-state drive when it is determined that the operating state of the second central processing unit is an abnormal state; determine whether the multiple second requests include a target request to be sent to the second central processing unit; and when it is determined that the multiple second requests include the target request, send the target request to the third central processing unit so that the third central processing unit executes the second operation corresponding to the target request.

[0165] In an exemplary embodiment, the prediction module 80 is also used to terminate the communication between the second central processor and other central processors, wherein the other central processors include: central processors in the solid state drive other than the second central processor; marking the first multidimensional vector and saving the marked first multidimensional vector; and updating the second vector cluster according to the first multidimensional vector.

[0166] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps in the above method embodiment when running.

[0167] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0168] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in the above method embodiment.

[0169] In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0170] An embodiment of the present application further provides a computer program product, including a computer program, which is implemented to perform the steps in the above method embodiment when executed by a processor.

[0171] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0172] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0173] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting an operating state, characterized in that: include: Acquire a plurality of first parameter data of the first central processing unit; Determine a first vector cluster and a second vector cluster according to the plurality of first parameter data, wherein the first vector cluster is used to indicate parameter data of the central processing unit in a normal state, and the second vector cluster is used to indicate parameter data of the central processing unit in a fault state; Acquire second parameter data of a second central processor in the solid state drive, and construct a first multidimensional vector corresponding to the second parameter data; Determining a first similarity between the first multidimensional vector and the first vector cluster, and determining a second similarity between the first multidimensional vector and the second vector cluster; predicting an operating state of the second central processing unit according to the first similarity and the second similarity; Wherein, determining the first vector cluster and the second vector cluster according to the plurality of first parameter data comprises: Performing data cleaning on the plurality of first parameter data to delete abnormal data in the plurality of first parameter data; Performing data normalization processing on the plurality of first parameter data after data cleaning based on a preset data range to obtain a plurality of third parameter data; Determine each second multi-dimensional vector corresponding to the first central processor according to each third parameter data; Inputting a plurality of second multidimensional vectors into a first time series network to obtain a first output value output by an output layer and a second output value output by a hidden layer, wherein the first time series network comprises: the output layer and the hidden layer; Determine the time correlation between any two of the plurality of second multidimensional vectors according to the first output value and the second output value; determining a distance parameter between every two multidimensional vectors in the plurality of second multidimensional vectors; Determine a third similarity between every two multidimensional vectors in the plurality of second multidimensional vectors according to the distance parameter and the time correlation; The plurality of second multi-dimensional vectors are divided into the first vector clusters and the second vector clusters based on a plurality of third similarities.

2. The method according to claim 1, characterized in that Acquiring a plurality of first parameter data of a first central processing unit, including: Setting a monitoring thread corresponding to the first central processor; and determining a set of performance parameters corresponding to the first central processor; In a case where the performance parameter of the first central processor is adjusted to each performance parameter in the performance parameter set, the plurality of first parameter data of the first central processor is determined based on the monitoring thread.

3. The method according to claim 1, characterized in that Before the plurality of second multidimensional vectors are input into the first time series network to obtain the first output value output by the output layer and the second output value output by the hidden layer, the method comprises: Determine a first activation function corresponding to the output layer and a second activation function corresponding to the hidden layer; The second time series network is trained based on the first activation function and the second activation function to obtain the first time series network.

4. The method according to claim 1, characterized in that Determining the time correlation between every two multidimensional vectors in the plurality of second multidimensional vectors according to the first output value and the second output value comprises: Determining a loss function corresponding to the first time series network; updating the first output value according to the loss function to generate an updated third output value; and updating the second output value according to the loss function to generate an updated fourth output value; The temporal correlation between every two multidimensional vectors in the plurality of second multidimensional vectors is determined according to the third output value and the fourth output value.

5. The method according to claim 1, characterized in that Determining a distance parameter between every two multidimensional vectors in the plurality of second multidimensional vectors comprises: The distance parameter between each two multidimensional vectors in the plurality of second multidimensional vectors is determined according to a first formula, wherein the first formula is: , d is the distance parameter, X n1 , X n2 , …, X nj For every two multidimensional vectors, the first j elements in the vector, X m1 , X m2 , …, X mj are the j elements in the second vector of every two multidimensional vectors, where m, n, and j are all positive integers.

6. The method according to claim 1, characterized in that Determining a third similarity between every two multidimensional vectors in the plurality of second multidimensional vectors according to the distance parameter and the time correlation degree comprises: Acquire noise data corresponding to the first central processor; and determine a first normalized weight corresponding to the distance parameter and a second normalized weight corresponding to the time correlation; A third similarity between every two multidimensional vectors in the plurality of second multidimensional vectors is determined according to the first normalized weight, the second normalized weight, and the noise data.

7. The method according to claim 6, characterized in that Determining a third similarity between every two multidimensional vectors in the plurality of second multidimensional vectors according to the first normalized weight, the second normalized weight and the noise data comprises: The first normalized weight, the second normalized weight and the noise data are input into a second formula to determine the third similarity, wherein the second formula is: similar=A*M1*d+B*M2*Ot+r, similar is the third similarity, M1 is the first normalized weight, M2 is the second normalized weight, r is the noise data, d is the distance parameter, Ot is the time correlation, A is the credibility corresponding to the distance parameter, and B is the credibility corresponding to the time correlation.

8. The method according to claim 6, characterized in that Determining a first normalized weight corresponding to the distance parameter includes: The first normalized weight is determined by a third formula, wherein the third formula includes: M1=d / (d+Ot), M1 is the first normalized weight, d is the distance parameter, and Ot is the time correlation.

9. The method according to claim 6, characterized in that Determining a second normalized weight corresponding to the time correlation degree includes: The second normalized weight is determined by a fourth formula, wherein the fourth formula includes: M2=Ot / (d+Ot), M2 is the second normalized weight, d is the distance parameter, and Ot is the time correlation.

10. The method according to claim 1, characterized in that Constructing a first multidimensional vector corresponding to the second parameter data, comprising: Determine a plurality of element attributes included in each first parameter data; Determine the multiple element attributes as target element attributes corresponding to the first multidimensional vector; Determining a plurality of elements corresponding to the target element attribute in the second parameter data; The first multi-dimensional vector is constructed according to the plurality of elements.

11. The method according to claim 1, characterized in that: Predicting the operating state of the second central processing unit according to the first similarity and the second similarity includes: Determine a magnitude relationship between the first similarity and the second similarity; When the magnitude relationship indicates that the first similarity is greater than the second similarity, determining that the operating state of the second central processing unit is a normal state; When the size relationship indicates that the first similarity is less than or equal to the second similarity, it is determined that the operating state of the second central processing unit is an abnormal state.

12. The method according to claim 1, characterized in that After predicting the operating state of the second central processing unit according to the first similarity and the second similarity, the method further includes: When it is determined that the operating state of the second central processing unit is normal, saving the first multi-dimensional vector; The first vector cluster is updated according to the first multidimensional vector.

13. The method according to claim 1, characterized in that After predicting the operating state of the second central processing unit according to the first similarity and the second similarity, the method further includes: In the case where it is determined that the operating state of the second central processor is an abnormal state, determining whether there is a first request being processed in the second central processor; In the case of determining that the second central processor has the first request, controlling the second central processor to execute a first operation corresponding to the first request, and determining whether the first operation is completed; When it is determined that the first operation is completed, starting the target process of the solid state drive; A target operation is performed on the second central processing unit according to the target process to prohibit the second central processing unit from receiving a request sent by the solid state drive.

14. The method according to claim 13, characterized in that After executing the target operation on the second central processing unit according to the target process, the method further includes: When it is determined that the operating state of the second central processing unit is an abnormal state, obtaining a plurality of second requests sent by the solid state drive; determining whether the plurality of second requests include a target request to be sent to the second central processor; When it is determined that the target request is included in the plurality of second requests, the target request is sent to a third central processing unit, so that the third central processing unit executes a second operation corresponding to the target request.

15. The method according to claim 13, characterized in that After executing the target operation on the second central processing unit according to the target process, the method further includes: terminating the communication between the second central processor and other central processors, wherein the other central processors include: central processors other than the second central processor in the solid state drive; The first multidimensional vector is marked and the marked first multidimensional vector is saved; and the second vector cluster is updated according to the first multidimensional vector.

16. A device for predicting an operating state, characterized in that: include: A first acquisition module, used for acquiring a plurality of first parameter data of a first central processing unit; A first determining module, configured to determine a first vector cluster and a second vector cluster according to the plurality of first parameter data, wherein the first vector cluster is used to indicate parameter data of the central processing unit in a normal state, and the second vector cluster is used to indicate parameter data of the central processing unit in a fault state; The second acquisition module is further used to acquire second parameter data of a second central processor in the solid state drive, and construct a first multidimensional vector corresponding to the second parameter data; The second determination module is further used to determine a first similarity between the first multidimensional vector and the first vector cluster, and to determine a second similarity between the first multidimensional vector and the second vector cluster; A prediction module, further configured to predict an operating state of the second central processing unit according to the first similarity and the second similarity; Among them, the first determination module is also used to perform data cleaning on the multiple first parameter data to delete abnormal data in the multiple first parameter data; perform data normalization processing on the multiple first parameter data after data cleaning based on a preset data range to obtain multiple third parameter data; determine each second multidimensional vector corresponding to the first central processor according to each third parameter data; input the multiple second multidimensional vectors into the first time series network to obtain the first output value output by the output layer and the second output value output by the hidden layer, wherein the first time series network includes: the output layer and the hidden layer; determine the time correlation between the two multidimensional vectors in the multiple second multidimensional vectors according to the first output value and the second output value; determine the distance parameter between each two multidimensional vectors in the multiple second multidimensional vectors; determine the third similarity between each two multidimensional vectors in the multiple second multidimensional vectors according to the distance parameter and the time correlation; divide the multiple second multidimensional vectors into the first vector cluster and the second vector cluster based on multiple third similarities.

17. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program implements the steps of the method according to any one of claims 1 to 15 when executed by a processor.

18. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 15 are implemented.

19. A computer program product comprising a computer program, characterized in that The computer program is used by a processor to execute the steps of the method described in any one of claims 1 to 15.

Images