Memory state detection method and apparatus, and communication device and storage medium

Abstract

A memory state detection method and apparatus, and a communication device and a non-volatile readable storage medium. The method comprises: acquiring memory data; dividing the memory data into first memory data and second memory data on the basis of the type of the memory data; determining a preliminary memory health degree score on the basis of a preset memory health degree evaluation model and the first memory data, and processing the second memory data by means of an input and output model, so as to determine a health degree influence factor, wherein the preset memory health degree evaluation model is determined on the basis of first historical memory data, and the input and output model is generated by means of training a preset initial model on the basis of second historical memory data; and determining a memory state on the basis of the preliminary memory health degree score and the health degree influence factor.

Description

Memory status detection method, device, communication equipment and storage medium

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to the Chinese patent application filed with the China Patent Office on July 28, 2023, with application number 202310935420.8, and application name “Memory status detection method, device, communication equipment and storage medium”, all contents of which are incorporated by reference into this application. Technical Field

[0003] The present application relates to the field of data processing technology, and in particular to a memory status detection method, apparatus, communication equipment, and non-volatile readable storage medium. Background Art

[0004] In modern large-scale data centers, millions of servers typically work together to provide high-performance computing and big data storage services. Because these servers run a vast array of tasks, hardware failures can significantly impact server reliability, availability, and serviceability (RAS). In server systems, memory, also known as internal storage, temporarily stores computational data from the CPU (Central Processing Unit) and data exchanged with external storage devices such as hard drives. All computer programs run in memory, so memory anomalies can have a significant impact on the computer. Memory failures are also one of the most common threats to hardware. To prevent memory errors, servers typically equip their memory with advanced Error Correction Code (ECC) mechanisms, such as SEC-DED (single error correction, double error detection) and ChipKill (an ECC memory protection technology). However, relying solely on ECC to ensure memory reliability is far from sufficient. In modern data centers, memory failures have been shown to be a major cause of server downtime or system failure. The increasing computing density and memory capacity have increased the risk of memory failures.

[0005] In the related art, relevant technicians intend to establish a model through a certain memory artificial intelligence algorithm to calculate the memory health based on the number of detected memory correctable errors (CE), frequency of occurrence, operating temperature, number of plug-in and plug-out times, power consumption and other factors. For example, artificial intelligence algorithms based on machine learning and deep learning. However, the above scheme has the problem of inputting all memory-related information into the artificial intelligence algorithm as the same dimension. Therefore, it will lead to an increase in the amount of model training calculations and unstable model establishment, and it is impossible to accurately feedback the current memory status.

[0006] Summary of the Invention

[0007] The purpose of the embodiments of the present application is to provide a memory status detection method, apparatus, communication device, and non-volatile readable storage medium to solve the technical problem in the prior art that the current memory status cannot be accurately fed back. The specific technical solution is as follows:

[0008] In a first aspect of the present application, a memory status detection method is provided. The memory status detection method includes:

[0009] Get memory data;

[0010] dividing the memory data into first memory data and second memory data according to the type of the memory data;

[0011] determining a preliminary memory health score based on a preset memory health assessment model and the first memory data, and processing the second memory data through an input-output model to determine a health influencing factor, wherein the preset memory health assessment model is determined based on the first historical memory data, and the input-output model is generated by training a preset initial model based on the second historical memory data;

[0012] Determine the current memory status based on the preliminary memory health score and health impact factors.

[0013] In some embodiments, the second memory data includes actual input data and actual output data, and the actual input data and the actual output data correspond one to one; processing the second memory data through the input-output model to determine the health impact factor includes:

[0014] Input the actual input data into the input-output model to obtain the predicted output data corresponding to the actual input data;

[0015] Compare the predicted output data with the actual output data to obtain the error value;

[0016] Determine the health impact factor based on the error value.

[0017] In some embodiments, the actual input data includes at least one of the following: average voltage, average operating frequency, and average erase / write speed data; the actual output data includes average memory temperature data; and before inputting the actual input data into the input-output model to obtain predicted output data corresponding to the actual input data, the method includes:

[0018] Preprocessing the actual input data collected within multiple preset time periods to generate data sets corresponding to the multiple preset time periods;

[0019] Filtering target data sets whose memory status is normal within three preset time periods corresponding to the current moment from data sets corresponding to multiple preset time periods;

[0020] Normalize the target data set to obtain training samples.

[0021] In some embodiments, after the step of normalizing the target dataset to obtain training samples, the method includes:

[0022] The preset initial model is trained according to the training samples to obtain the input and output model.

[0023] In some embodiments, determining the health impact factor based on the error value includes:

[0024] When the error value is detected to be less than the target error threshold, the health impact factor is set to 1;

[0025] When it is detected that the error value is greater than the target error threshold, a preset memory health policy is obtained, and a health impact factor is determined according to the memory health policy.

[0026] In some embodiments, the target error threshold is determined based on a root mean square error corresponding to a target mean square error vector between the predicted output data and the actual output data.

[0027] In some embodiments, determining the current memory state based on the preliminary memory health score and the health impact factor includes:

[0028] The target memory health score corresponding to the current moment is determined according to the preliminary memory health score, the health impact factor at the current moment, and the health impact factor within the previous preset time period corresponding to the current moment.

[0029] In some embodiments, the target memory health score is generated by the following formula: SCORE = SCORE_DF / μ N-1 μ N

[0030] In the above formula, SCORE is used to represent the target memory health score, SCORE_DF is used to represent the preliminary memory health score, and μ N Used to represent the health impact factor at the current moment, μ N-1 It is used to represent the health impact factor within the previous preset time period corresponding to the current moment, ε is used to represent the average value of the error values ​​between multiple predicted output data and actual output data, and σ is used to represent the target error threshold.

[0031] In some embodiments, after determining a target memory health score corresponding to the memory at the current moment based on the preliminary memory health score, the health impact factor at the current moment, and the health impact factor within a previous preset time period corresponding to the current moment, the method includes:

[0032] When it is detected that the target memory health score is less than the target health threshold, or the health impact factor is less than or greater than 1, an early warning message is sent to the user so that the user can view the current memory status through a preset display interface.

[0033] In some embodiments, the first memory data is obtained through a register log, and the first memory data includes at least one of the following: a memory hard fault, a number of memory errors, a memory error type, and an enabled error repair operation.

[0034] In some embodiments, the preset memory health assessment model includes:

[0035] When a preset number of memory hard faults are detected, a first preset score is deducted;

[0036] If it is detected that the number of memory errors within a preset time is greater than a preset threshold, a second preset score is deducted;

[0037] Determining a third preset score corresponding to the memory error type according to the memory error type;

[0038] A fourth preset score corresponding to the enabling error recovery operation is determined according to the enabling error recovery operation.

[0039] In some embodiments, the memory error type includes at least one of the following: memory hard error, memory soft error, SRAO (SW Recoverable Action Optional) error, UCNA (Uncorrected No Action) error, SRAR (SW Recoverable Action Required) error and sudden fatal error.

[0040] In some embodiments, determining a third preset score corresponding to the memory error type according to the memory error type includes:

[0041] When the memory error type detected is one of a memory hard error, a memory soft error, an SRAO error, a UCNA error, and an SRAR error, a third preset score is deducted;

[0042] When the detected memory error type is a sudden fatal error, the preliminary memory health score is 0.

[0043] In some embodiments, enabling error repair operations includes at least one of the following: consuming PCLS (Partial Cache-line Sparing), enabling Bank-level ADDDC (Adaptive Double DRAM Device Correction) function, enabling Rank-level ADDDC function, and enabling storm suppression function.

[0044] In some embodiments, the scores corresponding to enabling error repair operations are ranked from small to large as follows: consuming PCLS, enabling Bank-level ADDDC function, enabling Rank-level ADDDC function, and enabling storm suppression function.

[0045] In some embodiments, the data set corresponding to the input-output model includes four columns of vector data, and the input-output model includes three input data corresponding to one output data.

[0046] In some embodiments, the three input data include average voltage, average frequency during operation, and average erase / write speed data corresponding to the memory, and the output data includes average temperature data corresponding to the memory.

[0047] In another aspect of the present application, a memory status detection device is provided, the device comprising:

[0048] Acquisition module, used to obtain memory data;

[0049] A division module, configured to divide the memory data into first memory data and second memory data according to the type of the memory data;

[0050] a first determination module, configured to determine a preliminary memory health score based on a preset memory health assessment model and the first memory data, and to process the second memory data through an input-output model to determine a health influencing factor, wherein the input-output model is generated by training a preset initial model based on the second historical memory data;

[0051] The second determination module is used to determine the current memory state according to the preliminary memory health score and the health impact factor.

[0052] In another aspect of the present application, a communication device is provided, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus;

[0053] Memory for storing computer programs;

[0054] The processor is configured to implement any of the above-mentioned memory status detection methods when executing a program stored in the memory.

[0055] In another aspect of the present application, a non-volatile readable storage medium is provided. The non-volatile readable storage medium stores instructions that, when executed on a computer, enable the computer to execute any of the above-mentioned memory status detection methods.

[0056] In another aspect of the implementation of the present application, a computer program product comprising instructions is provided, which, when executed on a computer, enables the computer to execute any of the above-mentioned memory status detection methods.

[0057] The memory status detection method provided in the embodiment of the present application obtains memory data; divides the memory data into first memory data and second memory data according to the type of memory data; determines a preliminary memory health score according to a preset memory health assessment model and the first memory data, and processes the second memory data through an input-output model to determine a health influencing factor, wherein the preset memory health assessment model is determined based on the first historical memory data, and the input-output model is generated by training a preset initial model based on the second historical memory data; and determines the current memory status according to the preliminary memory health score and the health influencing factor. That is, the embodiment of the present application analyzes various factors that affect the healthy operation of the server memory and divides the memory data corresponding to different influencing factors into two categories, namely, first memory data and second memory data, wherein the first historical storage data corresponding to the first storage data can generate a preset memory health assessment model, and the first storage data can determine the preliminary memory health score corresponding to the memory according to the preset memory health assessment model; secondly, the second historical memory data can be processed to train the preset initial model to generate an input-output model, and by inputting the second memory data into the trained input-output model, the memory health influencing factor can be determined, and then the current memory status can be determined according to the preliminary memory health score and the health influencing factor. The present application divides the memory data into two categories according to the impact of the memory data on the memory, and adjusts the memory health score according to the health influencing factor, so as to effectively and accurately detect the memory health status. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art.

[0059] FIG1 shows a flowchart of the steps of a memory status detection method provided by an embodiment of the present application;

[0060] FIG2 shows a second flow chart of the steps of the memory status detection method provided in an embodiment of the present application;

[0061] FIG3 shows a flowchart of the third step of the memory status detection method provided in an embodiment of the present application;

[0062] FIG4 shows a fourth step flow chart of the memory status detection method provided by an embodiment of the present application;

[0063] FIG5 shows a device block diagram of a memory status detection device provided by an embodiment of the present application;

[0064] FIG6 shows a structural block diagram of a communication device provided in an embodiment of the present application;

[0065] FIG7 shows a schematic diagram of an input-output model training method provided in an embodiment of the present application;

[0066] FIG8 shows a schematic diagram of a three-input single-output ANFIS structure provided in an embodiment of the present application;

[0067] FIG9 shows a schematic diagram of a method for generating an initial FIS network structure provided in an embodiment of the present application;

[0068] FIG10 shows a schematic diagram of a preset display interface provided in an embodiment of the present application;

[0069] FIG11 shows a schematic diagram of another preset display interface provided in an embodiment of the present application. DETAILED DESCRIPTION

[0070] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0071] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, each embodiment of the present application will be described in detail below with reference to the accompanying drawings. However, it will be understood by those skilled in the art that in each embodiment of the present application, many technical details are proposed in order to enable the reader to better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in the present application can be implemented. The division of the following embodiments is for convenience of description and should not constitute any limitation on the specific implementation of the present application. The various embodiments can be combined with each other and referenced to each other under the premise of no contradiction.

[0072] 1 , a flowchart of the first step of a memory status detection method according to an embodiment of the present application is shown. The method may include:

[0073] Step 101, obtaining memory data;

[0074] It should be noted that in the embodiments of this application, the root cause of memory errors is referred to as a fault, irreversible damage in the hardware is referred to as a hard fault, and the observed fault symptoms are referred to as errors. Errors caused by hard faults are called hard errors. Hard faults are inherently persistent and recurring, and cannot be resolved over time or by resetting or restarting the system. In contrast, soft faults and soft errors are random and transient, such as bit flips caused by particle collisions. The above-mentioned number of memory hard faults refers to the number of cells with hard faults in the same memory bank, and is more specifically described as the number of cells with two or more errors. The error frequency refers to the ratio of all memory errors detected within a period of time to the period of time. All memory errors include soft errors and recurring hard errors. The error type refers to the different types of memory errors, such as SW Recoverable Action Optional (SRAO) errors, Uncorrected No Action (UCNA) errors, and SW Recoverable Action Required (SRAR) errors. The repair operation refers to the enabling of certain memory RAS technologies, such as intra-granular redundant row replacement fault row technology (Partial Cache-line Sparing, PCLS) and Adaptive Double DRAM Device Correction (ADDDC) technology.

[0075] Therefore, the memory data in the embodiment of the present application may include all relevant information of the memory that may affect the memory state.

[0076] Step 102, dividing the memory data into first memory data and second memory data according to the type of the memory data;

[0077] It should be noted that after obtaining the memory data in step 101, by analyzing various factors that affect the healthy operation of the server memory, the influencing factors can be divided into two categories.

[0078] The first category is direct factors that can intuitively reflect memory health. Therefore, the memory data corresponding to such influencing factors are divided into first memory data, such as the number of memory hard failures, error frequency, error type, error repair operations, etc., and the above-mentioned first memory data can be obtained through register logs.

[0079] The second category is indirect factors that can indirectly reflect memory abnormalities. Therefore, the memory data corresponding to such influencing factors are divided into second memory data, such as the average memory temperature, average voltage, average frequency during operation, and average erase and write speed measured within a certain time period. When indirect factors are abnormal, they may not directly affect the health of the memory, that is, they may not directly lead to the occurrence of memory errors, but their data abnormalities can, to a certain extent, warn of abnormal memory status. The above-mentioned second memory data can be measured by sensors.

[0080] Step 103: Determine a preliminary memory health score based on a preset memory health assessment model and the first memory data, and process the second memory data through an input-output model to determine a health influencing factor, wherein the preset memory health assessment model is determined based on the first historical memory data, and the input-output model is generated by training a preset initial model based on the second historical memory data.

[0081] It should be noted that, in the present application, after the memory data is divided, different types of memory data may be processed differently.

[0082] Specifically, for the first memory data, a preset memory health assessment model may be generated according to characteristics corresponding to the first memory data, that is, information about direct factors affecting the memory state.

[0083] The preset memory health assessment model can determine the score to be deducted when a certain situation occurs based on the importance of the first memory data. That is, the preset memory health assessment model adopts a deduction strategy, and each time a memory error is detected, points are deducted according to corresponding rules. The score range is set to 0 to 100. A higher score indicates a healthier memory stick, and a lower score indicates a lower failure risk of the memory stick.

[0084] Specifically, as shown in Table 1 below, Table 1 is an exemplary memory status framework corresponding to different memory health scores.

[0085] Table 1 An exemplary memory status framework corresponding to different memory health scores

[0086] It should be noted that the first memory data is obtained through a register log, and the first memory data includes at least one of the following: memory hard fault, memory error number, memory error type, and enabled error repair operation.

[0087] The preset memory health assessment model includes: deducting a first preset score when a preset number of memory hard faults are detected; deducting a second preset score when the number of memory errors detected within a preset time is greater than a preset threshold; determining a third preset score corresponding to the memory error type based on the memory error type; and determining a fourth preset score corresponding to the enabled error repair operation based on the enabled error repair operation.

[0088] Furthermore, the memory error type includes at least one of the following: a memory hard error, a memory soft error, an SRAO error, a UCNA error, a SRAR error, and a sudden fatal error.

[0089] Determining the third preset score corresponding to the memory error type according to the memory error type includes: when the memory error type is detected to be one of a memory hard error, a memory soft error, an SRAO error, an UCNA error, and an SRAR error, deducting the third preset score; when the memory error type is detected to be a sudden fatal error, the preliminary memory health score is 0.

[0090] Furthermore, enabling the error repair operation includes at least one of the following: consuming PCLS, enabling a Bank-level ADDDC function, enabling a Rank-level ADDDC function, and enabling a storm suppression function.

[0091] The scores for enabling error recovery operations are ranked from smallest to largest as follows: consuming PCLS, enabling the bank-level ADDDC function, enabling the rank-level ADDDC function, and enabling the storm suppression function.

[0092] It should be noted that, in the embodiment of the present application, the preset memory health assessment model corresponds to different deduction strategies according to different first memory data.

[0093] Specifically, the first one is the rule for deducting points based on the number of memory hard faults. Memory hard errors are persistent and repetitive errors caused by memory hard faults, and are one of the main factors that dominate memory failures. Usually, the server RAS policy will not allow it to recur, and will process the Cell unit with hard faults by enabling isolation or repair strategies. Memory soft faults and soft errors are usually instantaneous. Under normal circumstances, if the Cell unit with soft errors is not handled in time, the error will not recur. Therefore, this section only considers the impact of hard faults on the physical hardware of the memory bar. In the embodiment of the present application, for example, the deduction rule is set as follows: 1 point is deducted for every 10 hard faults detected, and 1 point is deducted for less than 10 hard faults.

[0094] The second type is the memory error frequency deduction rule. When the number of errors in the memory stick is too large within a certain time period, on the one hand, a large number of memory errors will cause frequent interruption information to interfere with the normal operation of the system; on the other hand, the high frequency of memory errors also reflects the abnormality of the memory stick to a certain extent. It should be noted that the deduction rules set in this section mainly consider the mapping relationship between frequent memory errors and memory abnormalities in a short period of time, and the set threshold should not exceed the memory error storm suppression threshold. In the embodiment of the present application, for example, the deduction rule is set as follows: if the number of memory errors detected every 60 seconds exceeds 500, 1 point will be deducted.

[0095] The third type is the memory error type deduction rule. Memory errors can be divided into correctable errors and uncorrectable errors. Furthermore, correctable errors and uncorrectable errors can be further subdivided. Memory correctable errors can be divided into patrol correctable errors, read / write correctable errors, mirror write-back failure errors, etc. In this application, only correctable errors are divided into two categories: memory hard errors and soft errors, and handled separately. Uncorrectable errors can be divided into ① uncorrectable errors that can be processed selectively - SRAO errors. The error code for SRAO errors is: Error_Type: UCE; MSCODE: 0x0010; ② uncorrectable errors that do not need to be processed - UCNA errors. The error code for UCNA errors is: Error_Type: UCE; MSCODE: 0x0101; ③ uncorrectable errors that must be processed - SRAR errors. The error code for SRAR errors is: Error_Type: UCE; MSCODE: 0x0010. ④ uncorrectable errors that cause system downtime due to physical hardware errors in the memory module - sudden fatal errors. The error code for sudden fatal errors is: Error_Type: UCE. In the embodiment of the present application, for example, the deduction rule is set as follows: 5 points are deducted for each occurrence of UCNA, SRAO, or SRAR error, and the memory health score is reduced to 0 points if a sudden fatal error occurs.

[0096] It should be noted that in addition to the above-mentioned UCNA, SRAO, and SRAR error types, memory error types also include some specific errors, such as row errors, column errors, and Bank errors divided according to the memory structure. These error types divided according to the structure are defined according to the code rules set by relevant memory RAS technicians. Different code rules have different results, so the deduction rules set are also different.

[0097] Fourth, the penalty rule for enabling error repair operations. In memory RAS technology, if a memory hard fault is detected, the OS layer (operating system layer) is typically enabled to perform memory page isolation to isolate the memory page where the current error occurs. Read and write operations on the isolated memory page are no longer permitted. If certain memory pages cannot be isolated for some reason, memory repair technology is enabled to repair the hard error. Repair technology refers to fault-redundant replacement mechanisms for the processor or memory, such as PCLS and ADDDC. Generally, repair resources are limited, and some repair mechanisms affect memory performance. PCLS resources can be called up to 16 times per memory channel and up to once per cache line. ADDDC is an advanced memory RAS feature that supports two sets of VLS registers simultaneously per memory channel. The enablement levels are divided into bank and rank levels. When rank-level ADDDC is enabled, memory performance decreases by approximately 25%-30%. This also reflects, to a certain extent, the high number of accumulated memory errors and the depletion of the remaining repair resources of the memory RAS technology, providing users with a warning of repair resources. When a large number of errors that cannot be processed by memory isolation and repair technology appear in the memory instantly, and the number of errors reaches a certain storm threshold (different server manufacturers have different settings according to different RAS policies), the SMI storm suppression function is enabled, that is, to prevent the BIOS (Basic Input Output System) layer from reporting a large number of System Management Interrupt (SMI) information to the OS layer, affecting the normal operation of the system, and controlling the failure of the memory page isolation function, which means that the memory stick is no longer healthy and needs to be replaced. In the embodiment of the present application, for example, the deduction rule is set as follows: 1 point is deducted for consuming PCLS once, 10 points are deducted for enabling the Bank-level ADDDC function, 20 points are deducted for enabling the Rank-level ADDDC function, and 50 points are deducted for enabling the storm suppression function.

[0098] Specifically, for the second memory data, the characteristics of the second memory data are indirect factor information that has an indirect impact on the memory state. Therefore, in an embodiment of the present application, an input-output model is obtained by training the second memory data, and an influencing factor, namely, a health influencing factor, can be obtained based on the input-output model.

[0099] The above health influencing factors can be used as one of the conditions for ultimately determining the memory status.

[0100] In addition, the training process of the above input and output models can be found in the subsequent content.

[0101] Step 104 : Determine the current memory state based on the preliminary memory health score and the health impact factor.

[0102] It should be noted that after determining the preliminary memory health score and the health impact factor obtained through the input-output model in step 103, the current memory state can be determined based on the above data.

[0103] Specifically, as explained above, the factors that can be mapped to memory health are divided into direct factors and indirect factors. Direct factors are used to establish a preliminary health score strategy, and indirect factors are used as one of the ways to evaluate memory anomalies. At the same time, an input-output model is established through indirect factors, so as to set the health influencing factor and further correct the memory health score. Therefore, when the output result obtained based on the input-output model is compared with the measured data, if the comparison error is less than the set threshold, the feedback indicates that there is no memory anomaly. In this case, the indirect factors do not affect the preliminary health score strategy established by the direct factors. If the feedback result obtained through the indirect factors is a memory anomaly, then its impact on the memory health score must be negative.

[0104] Therefore, the final memory health score can be comprehensively determined by the preliminary memory health score and the health impact factor, thereby determining the current memory status according to the content in Table 1, for example.

[0105] The memory status detection method provided in the embodiment of the present application obtains memory data; divides the memory data into first memory data and second memory data according to the type of memory data; determines a preliminary memory health score according to a preset memory health assessment model and the first memory data, and processes the second memory data through an input-output model to determine a health influencing factor, wherein the preset memory health assessment model is determined based on the first historical memory data, and the input-output model is generated by training a preset initial model based on the second historical memory data; and determines the current memory status according to the preliminary memory health score and the health influencing factor. That is, the embodiment of the present application analyzes various factors that affect the healthy operation of the server memory and divides the memory data corresponding to different influencing factors into two categories, namely, first memory data and second memory data, wherein the first historical storage data corresponding to the first storage data can generate a preset memory health assessment model, and the first storage data can determine the preliminary memory health score corresponding to the memory according to the preset memory health assessment model; secondly, the second historical memory data can be processed to train the preset initial model to generate an input-output model, and by inputting the second memory data into the trained input-output model, the memory health influencing factor can be determined, and then the current memory status can be determined according to the preliminary memory health score and the health influencing factor. The present application divides the memory data into two categories according to the impact of the memory data on the memory, and adjusts the memory health score according to the health influencing factor, so as to effectively and accurately detect the memory health status.

[0106] 2 , a second flowchart of the memory status detection method according to an embodiment of the present application is shown. The method may include:

[0107] Step 201, obtaining memory data;

[0108] Step 202: Divide the memory data into first memory data and second memory data according to the type of the memory data;

[0109] It should be noted that the above steps 201-202 refer to the previous discussion and will not be repeated here.

[0110] Step 203: determining a preliminary memory health score based on a preset memory health assessment model and the first memory data, and inputting actual input data into an input-output model to obtain predicted output data corresponding to the actual input data, comparing the predicted output data with the actual output data to obtain an error value, and determining a health impact factor based on the error value, wherein the preset memory health assessment model is determined based on the first historical memory data, the input-output model is generated by training a preset initial model based on the second historical memory data, and the second memory data includes actual input data and actual output data, and the actual input data and the actual output data correspond one to one;

[0111] Furthermore, for the input and output models, it is necessary to train a preset initial model to obtain the model. Therefore, before step 203, referring to FIG. 4 , FIG. 4 may include the following steps:

[0112] Step 001: pre-processing the actual input data collected during multiple preset time periods to generate data sets corresponding to the multiple preset time periods;

[0113] Step 002: Filter the target data sets whose memory status is normal in the three preset time periods corresponding to the current moment from the data sets corresponding to the multiple preset time periods;

[0114] Step 003: normalize the target data set to obtain training samples.

[0115] Step 004: train the preset initial model according to the training samples to obtain the input and output model.

[0116] It should be noted that in the embodiment of the present application, the processing part of the second memory data is mainly explained. The characteristics of the second memory data are indirect factor information that has an indirect impact on the memory state. Therefore, in the embodiment of the present application, the input-output model is obtained by training the second memory data. According to the input-output model, an influencing factor can be obtained, namely the health influencing factor.

[0117] Figure 7 illustrates the detailed process for acquiring and training training samples for the input-output model. From t0 to (t0 + Δt), measured memory temperature, voltage, runtime frequency, and write / erase speed information are collected every 1 second. The average of every ten data points is calculated to generate the indirect factor information, "Dataset_0." This indirect factor information, "Dataset_0," forms the "Total Indirect Factor Dataset." Δt represents the collection period, which can be set to 1 hour, 5 hours, or 24 hours.

[0118] Furthermore, the data set corresponding to the input-output model includes four columns of vector data, and the input-output model includes three input data corresponding to one output data.

[0119] Furthermore, the three input data include the average voltage, average frequency during operation, and average erase / write speed data corresponding to the memory, and the output data includes the average temperature data of the memory.

[0120] The total indirect factor dataset consists of four columns of vector data. The first three columns represent the memory's average voltage, average runtime frequency, and average erase / write speed, respectively. The last column represents the memory's average temperature. Each data vector is normalized to a value between 0 and 1 to generate training samples. The first three columns of the training sample serve as input for the training input / output model, while the last column serves as output.

[0121] Therefore, after the input-output model is trained, the acquired second memory data is input into the input-output model, wherein the second memory data includes actual input data and actual output data, and the actual input data and the actual output data correspond one to one, wherein the actual input data includes at least one of the following average voltage, average frequency during operation, and average erase speed data, and the actual output data includes average memory temperature data.

[0122] Specifically, the actual input data is input into the input-output model to obtain the predicted output data corresponding to the actual input data. The predicted output data and the actual output data are compared to obtain the error value, and the health influencing factor can be determined based on the error value.

[0123] At time (t0+Δt), a Takagi-Sugeno type FIS (Frame Information Structure) structure can be generated by grid segmentation based on the training samples to determine the initial values ​​of the membership function parameters. Specifically, reference can be made to FIG9 , which is a schematic diagram of the initial FIS network structure generated using the grid segmentation method.

[0124] It should be noted that in the embodiment of the present application, the preset initial model can be an adaptive fuzzy neural network ANFIS (Adaptive Network-based Fuzzy Inference System, adaptive neural fuzzy system). ANFIS is a multi-input single-output system. The ANFIS in this application is a three-input single-output ANFIS structure. Specifically, as shown in Figure 8, the structure of ANFIS consists of five layers, from input to output, namely, fuzzy layer, rule layer, normalization layer, defuzzification layer and output layer.

[0125] Specifically, in ANFIS, the parameters of the membership function are determined by training with a sample data set. The way in which the membership functions combine or interact with each other is called a rule. The if-then rule is described as follows:

[0126] Rule 1: if x=A1,y=B1,z=C1then f1=m1x+p1y+q1z+r1

[0127] Rule 2: if x=A2,y=B2,z=C2then f2=m2x+p2y+q2z+r2

[0128] Rule 3: if x=A3,y=B3,z=C3then f3=m3x+p3y+q3z+r3

[0129] In the above if-then form, x, y, and z are the inputs of the node; f is the output; A i 、B i and C i are fuzzy sets related to input x, y, and z, i = 1, 2, 3; m i 、p i ,q i and r i It is the result parameter, usually called the consequent parameter.

[0130] The first layer: The node function of this layer is the membership function.

[0131] In the above formula 1, O 1,i is the output value of the layer; n is the number of input signals; and is the generalized bell-shaped membership function (gbellmf), defined as

[0132] In the above formula 2, a i 、b i and c i is a conditional parameter, also commonly called antecedent parameter. Changes in the parameter value can change the shape of gbellmf.

[0133] Second layer: The nodes in this layer are labeled as Π. Output value O 2,i It is calculated by multiplying all input member functions.

[0134] In the above formula 3, w i Represents the incentive intensity of the i-th rule.

[0135] The third layer: The nodes in this layer are marked as N. The output results of the previous layer are normalized and the output value is O 3,i is the stimulus intensity after processing.

[0136] In the above formula 4, is the output value of the third layer.

[0137] The fourth layer: This layer creates an adaptive correlation function between the normalized excitation intensity and the result function. 4,iIt is the product of the values ​​of the third layer and the first layer.

[0138] In the above formula 5, m i 、p i ,q i and r i It is the result parameter, usually called the consequent parameter.

[0139] Layer 5: This layer is labeled Σ. It calculates the total output as the sum of all input signals.

[0140] In the above formula 6, O 4,i The sum of all input signals.

[0141] It should be noted that the fuzzy membership function parameters (including antecedent and consequent parameters) in ANFIS are obtained by generating an initial fuzzy model from a large amount of known data and then training it. Through this iterative adaptive learning process, ANFIS trains the antecedent and consequent parameters of the model system, optimizing and adjusting them to ultimately determine the membership function parameter values ​​that best fit the training dataset. With each training iteration, the error between the actual and expected outputs is reduced, and the training process is terminated when a predetermined number of training cycles or error rate is reached.

[0142] During the time period (t0+Δt) to (t0+2Δt), the measured memory temperature, voltage, runtime frequency, and erase / write speed information is collected every 1 second. The average of every ten data points is calculated to generate the indirect factor information "Dataset_1." During this time period, the average voltage, runtime frequency, and erase / write speed data are used as input to the ANFIS model obtained in the previous cycle to obtain the corresponding average temperature output value. This output value is then compared with the measured average memory temperature data. If the comparison error is less than the set threshold, the indirect factor information feedback indicates that there are no memory anomalies during this time period.

[0143] At time (t0+2Δt), the total indirect factor dataset is supplemented by dataset_1, and new training samples are obtained after normalization. Retraining is performed based on the ANFIS model rules obtained in the previous cycle, which can quickly update the ANFIS model rules.

[0144] Similarly, within the time period (t0+2Δt) to (t0+3Δt), the operations corresponding to the above time period (t0+Δt) to (t0+2Δt) and the moment (t0+2Δt) can be repeated to obtain the specific process of acquiring training samples and training the ANFIS model.

[0145] It should be noted that, in order to avoid the data set being too large, a sliding time window can be set so that the total indirect factor data set can contain at most data without abnormalities in the feedback memory collected within the last three time periods; the acquisition of training samples and the training process of the ANFIS model can be set to be performed in a periodic manner with an interval of Δt time period, or it can be set to be performed non-periodically in an active manner. The above is only explained in terms of a periodic manner with an interval of Δt time period.

[0146] Further, in step 203, determining the health impact factor based on the error value includes: when it is detected that the error value is less than the target error threshold, setting the health impact factor to 1; when it is detected that the error value is greater than the target error threshold, obtaining a pre-set memory health policy, and determining the health impact factor based on the memory health policy.

[0147] It should be noted that in the embodiment of the present application, when the feedback result obtained through indirect factors is memory abnormality, its impact on the memory health score must be negative. Therefore, the embodiment of the present application can set health impact factor rules.

[0148] Furthermore, the target error threshold is determined based on a root mean square error corresponding to a target mean square error vector between the predicted output data and the actual output data.

[0149] It should be noted that when the input-output model is trained based on sample data, the minimum mean square error vector of the training data can be obtained, and the root mean square error σ of the minimum mean square error vector can be calculated. According to the preset empirical formula, 3σ can be set as the threshold before the next input-output model rule update, which is used to scalarize the feedback results obtained through indirect factors.

[0150] If the comparison error between the output result obtained based on the input-output model and the measured data is less than 3σ, it means that there is no memory abnormality, that is, the indirect factor does not affect the memory health strategy, and the health impact factor is set to 1; if the comparison error is greater than 3σ, the indirect factor is set to affect the memory health strategy, and the health impact factor value changes with the error size.

[0151] Step 204: Determine the current memory state based on the preliminary memory health score and the health impact factor.

[0152] It should be noted that the above step 204 is described with reference to the previous discussion and will not be repeated here.

[0153] The embodiment of the present application analyzes various factors that affect the healthy operation of the server memory and divides the memory data corresponding to different influencing factors into two categories, namely first memory data and second memory data, wherein the first historical storage data corresponding to the first storage data can generate a preset memory health assessment model, and the first storage data can determine the preliminary memory health score corresponding to the memory according to the preset memory health assessment model; secondly, the second historical memory data can be processed to train the preset initial model to generate an input-output model, and after inputting the second memory data into the trained input-output model, the health influencing factor of the memory can be determined, and then the current memory status can be determined according to the preliminary memory health score and the health influencing factor. The present application divides the memory data into two categories according to the impact of the memory data on the memory, and adjusts the memory health score according to the health influencing factor, so as to effectively and accurately detect the memory health status.

[0154] Furthermore, in this embodiment, the mapping relationship between the measured memory temperature, voltage, operating frequency, and erase / write speed information is obtained based on the input / output model. A method for establishing and updating input / output model rules has been designed, allowing the input / output model to be updated online based on measured data. The predicted results of the input / output model can be used to provide a certain degree of early warning of memory risks.

[0155] In addition, the concept of memory health influencing factor is proposed in the embodiment of the present application, and a detailed method for obtaining the memory health influencing factor and a method for calculating the health score obtained by the health influencing factor and the preliminary memory health evaluation strategy are given.

[0156] 3 , a flowchart of the third step of the memory status detection method provided by an embodiment of the present application is shown. The method may include:

[0157] Step 301, obtaining memory data;

[0158] Step 302: Divide the memory data into first memory data and second memory data according to the type of the memory data;

[0159] Step 303: determining a preliminary memory health score based on a preset memory health assessment model and the first memory data, and processing the second memory data through an input-output model to determine a health impact factor, wherein the preset memory health assessment model is determined based on the first historical memory data, and the input-output model is generated by training a preset initial model based on the second historical memory data;

[0160] It should be noted that the above steps 301-303 refer to the previous discussion and will not be repeated here.

[0161] Step 304 : Determine a target memory health score corresponding to the memory at the current moment based on the preliminary memory health score, the health impact factor at the current moment, and the health impact factor within the previous preset time period corresponding to the current moment.

[0162] It should be noted that in an embodiment of the present application, the target memory health score corresponding to the memory at the current moment is determined by the preliminary memory health score, the health impact factor at the current moment, and the health impact factor within the previous preset time period corresponding to the current moment.

[0163] The health impact factor in the last preset time period corresponding to the current moment refers to the health impact factor in the last time period of the current moment, wherein the time period may refer to the period of the interval Δt described above.

[0164] Specifically, the target memory health score is generated by the following formula:

[0165] In the above formula 7, SCORE is used to represent the target memory health score, SCORE_DF is used to represent the preliminary memory health score, μ N Used to represent the health impact factor at the current moment, μ N-1 It is used to represent the health impact factor within the previous preset time period corresponding to the current moment, ε is used to represent the average value of the error values ​​between multiple predicted output data and actual output data, and σ is used to represent the target error threshold.

[0166] Step 305: Determine the current state of the memory according to the target memory health score.

[0167] Furthermore, step 305 may include: when it is detected that the target memory health score is less than the target health threshold, or the health impact factor is less than or greater than 1, sending an early warning message to the user so that the user can view the current memory status through a preset display interface.

[0168] It should be noted that, in the embodiment of the present application, the health evaluation results of each memory module configured by the server can be displayed through a web interface. The user can query the health evaluation results of each memory module configured by the server by entering the server login address, as shown in the display interface diagram 10.

[0169] In addition, if the target memory health score is detected to be less than the target health threshold, or the health impact factor is less than or greater than 1, an alert message is sent to the user, allowing the user to view the current memory status through a preset display interface. The preset display interface is the aforementioned web interface, where users can view detailed information about each memory bank, as shown in Figure 11. If a memory health score is less than 60 points or the health impact factor is not 1, a proactive alert is issued to the user.

[0170] The embodiment of the present application analyzes various factors that affect the healthy operation of the server memory and divides the memory data corresponding to different influencing factors into two categories, namely first memory data and second memory data, wherein the first historical storage data corresponding to the first storage data can generate a preset memory health assessment model, and the first storage data can determine the preliminary memory health score corresponding to the memory according to the preset memory health assessment model; secondly, the second historical memory data can be processed to train the preset initial model to generate an input-output model, and after inputting the second memory data into the trained input-output model, the health influencing factor of the memory can be determined, and then the current memory status can be determined according to the preliminary memory health score and the health influencing factor. The present application divides the memory data into two categories according to the impact of the memory data on the memory, and adjusts the memory health score according to the health influencing factor, so as to effectively and accurately detect the memory health status.

[0171] In addition, while the embodiment of the present application can effectively perform scalar evaluation of memory health, it can also issue early warnings for memory anomalies based on the memory health score and the health impact factor value. When the memory score is low to the abnormal threshold, the user is prompted to replace healthy hardware.

[0172] 5 , which shows a memory status detection device provided by an embodiment of the present application, the device may include:

[0173] Acquisition module 501, used to acquire memory data;

[0174] A division module 502, configured to divide the memory data into first memory data and second memory data according to the type of the memory data;

[0175] a first determination module 503 configured to determine a preliminary memory health score based on a preset memory health assessment model and the first memory data, and to process the second memory data through an input-output model to determine a health impact factor, wherein the preset memory health assessment model is determined based on the first historical memory data, and the input-output model is generated by training a preset initial model based on the second historical memory data;

[0176] The second determining module 504 is configured to determine the current memory state according to the preliminary memory health score and the health impact factor.

[0177] The embodiment of the present application analyzes various factors that affect the healthy operation of the server memory and divides the memory data corresponding to different influencing factors into two categories, namely first memory data and second memory data, wherein the first historical storage data corresponding to the first storage data can generate a preset memory health assessment model, and the first storage data can determine the preliminary memory health score corresponding to the memory according to the preset memory health assessment model; secondly, the second historical memory data can be processed to train the preset initial model to generate an input-output model, and after inputting the second memory data into the trained input-output model, the health influencing factor of the memory can be determined, and then the current memory status can be determined according to the preliminary memory health score and the health influencing factor. The present application divides the memory data into two categories according to the impact of the memory data on the memory, and adjusts the memory health score according to the health influencing factor, so as to effectively and accurately detect the memory health status.

[0178] The embodiment of the present application further provides a communication device, as shown in FIG6 , including a processor 601, a communication interface 602, a memory 603, and a communication bus 604, wherein the processor 601, the communication interface 602, and the memory 603 communicate with each other via the communication bus 604.

[0179] Memory 603, used for storing computer programs;

[0180] When the processor 601 is used to execute the program stored in the memory 603, it can implement the following steps:

[0181] Get memory data;

[0182] dividing the memory data into first memory data and second memory data according to the type of the memory data;

[0183] determining a preliminary memory health score based on a preset memory health assessment model and the first memory data, and processing the second memory data through an input-output model to determine a health influencing factor, wherein the preset memory health assessment model is determined based on the first historical memory data, and the input-output model is generated by training a preset initial model based on the second historical memory data;

[0184] Determine the current memory status based on the preliminary memory health score and health impact factors.

[0185] The communication bus mentioned in the terminal can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. This communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is used in the figure, but this does not mean that there is only one bus or only one type of bus.

[0186] The communication interface is used for communication between the above terminal and other devices.

[0187] The memory may include random access memory (RAM) or non-volatile memory, such as at least one disk storage. Alternatively, the memory may be at least one storage device located away from the processor.

[0188] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.

[0189] In another embodiment provided in the present application, a non-volatile readable storage medium is also provided. The non-volatile readable storage medium stores instructions, which, when executed on a computer, enable the computer to perform any memory status detection in the above embodiments.

[0190] In another embodiment provided by the present application, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to perform the memory status detection in any one of the above embodiments.

[0191] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When software is used for implementation, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable devices. The computer instructions can be stored in a non-volatile readable storage medium, or transmitted from one non-volatile readable storage medium to another non-volatile readable storage medium. For example, the computer instructions can be transmitted from a website, a computer, a server or a third database by wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) mode to another website, computer, server or the third database. The non-volatile readable storage medium can be any available medium that a computer can access or a data storage device such as a server or the third database that includes one or more available media integrations. Available media may be magnetic media (eg, floppy disks, hard disks, tapes), optical media (eg, DVDs (digital versatile discs)), or semiconductor media (eg, solid state disks (SSDs)).

[0192] It should be noted that, in this document, relational terms such as first and first are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0193] Each embodiment in this specification is described in a related manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiment is generally similar to the method embodiment, so the description is relatively simple. For related parts, refer to the description of the method embodiment.

[0194] The above are only preferred embodiments of the present application and are not intended to limit the scope of protection of the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application are included in the scope of protection of the present application.

Claims

1. A memory status detection method, characterized in that: The memory status detection method comprises: Get memory data; Dividing the memory data into first memory data and second memory data according to the type of the memory data; Determine a preliminary memory health score according to a preset memory health assessment model and the first memory data, and process the second memory data through an input-output model to determine a health influencing factor, wherein the preset memory health assessment model is determined based on the first historical memory data, and the input-output model is generated by training a preset initial model based on the second historical memory data; The current memory state is determined according to the preliminary memory health score and the health impact factor.

2. The method according to claim 1, characterized in that The second memory data includes actual input data and actual output data, and the actual input data and the actual output data correspond one to one; The processing of the second memory data through the input-output model to determine the health impact factor includes: Inputting actual input data into the input-output model to obtain predicted output data corresponding to the actual input data; Comparing the predicted output data with the actual output data to obtain an error value; A health impact factor is determined according to the error value.

3. The method according to claim 2, characterized in that Before the step of inputting the actual input data into the input-output model to obtain the predicted output data corresponding to the actual input data, the method includes: Preprocessing the actual input data collected within multiple preset time periods to generate data sets corresponding to the multiple preset time periods; Filtering, from the data sets corresponding to the multiple preset time periods, a target data set whose memory status is in a normal state within three preset time periods corresponding to the current moment; The target data set is normalized to obtain training samples.

4. The method according to claim 3, characterized in that After the step of normalizing the target data set to obtain training samples, the method includes: The preset initial model is trained according to the training samples to obtain an input-output model.

5. The method according to claim 2, characterized in that: Determining the health impact factor according to the error value includes: When it is detected that the error value is less than the target error threshold, the health impact factor is set to 1; When it is detected that the error value is greater than the target error threshold, a preset memory health policy is obtained, and the health impact factor is determined according to the memory health policy.

6. The method according to claim 5, characterized in that The target error threshold is determined according to a root mean square error corresponding to a target mean square error vector between the predicted output data and the actual output data.

7. The method according to claim 1, characterized in that Determining the current memory state according to the preliminary memory health score and the health impact factor comprises: Determine the target memory corresponding to the current moment according to the preliminary memory health score, the health impact factor at the current moment, and the health impact factor within the previous preset time period corresponding to the current moment. Memory health score; The state of the memory at the current moment is determined according to the target memory health score.

8. The method according to claim 7, characterized in that The target memory health score is generated by the following formula: SCORE = SCORE_DF / μ N-1 ·μ N In the above formula, SCORE is used to represent the target memory health score, SCORE_DF is used to represent the preliminary memory health score, μ N It is used to represent the health impact factor at the current moment, μ N-1 It is used to represent the health impact factor within the previous preset time period corresponding to the current moment, ε is used to represent the average value corresponding to the error values ​​between multiple predicted output data and actual output data, and σ is used to represent the target error threshold.

9. The method according to claim 7, characterized in that: Determining the state corresponding to the memory at the current moment according to the target memory health score includes: When it is detected that the target memory health score is less than the target health threshold, or the health impact factor is less than or greater than 1, a warning message is sent to the user so that the user can view the current memory status through a preset display interface.

10. The memory status detection method according to claim 1, characterized in that: The first memory data is obtained through a register log, and the first memory data includes at least one of the following: a memory hard fault, a memory error number, a memory error type, and an enabled error repair operation.

11. The memory status detection method according to claim 10, characterized in that: The preset memory health evaluation model includes: In the case where a preset number of memory hard failures are detected, deducting a first preset score; When it is detected that the number of memory errors within a preset time is greater than a preset threshold, a second preset score is deducted; Determining a third preset score corresponding to the memory error type according to the memory error type; A fourth preset score corresponding to the enabling error recovery operation is determined according to the enabling error recovery operation.

12. The memory status detection method according to claim 11, characterized in that: The memory error type includes at least one of the following: a memory hard error, a memory soft error, an optional SRAO error, a non-necessary UCNA error, a mandatory SRAR error, and a sudden fatal error.

13. The memory status detection method according to claim 12, characterized in that: Determining, according to the memory error type, a third preset score corresponding to the memory error type comprises: When it is detected that the memory error type is one of a memory hard error, a memory soft error, an SRAO error, a UCNA error, and an SRAR error, deducting the third preset score; When it is detected that the memory error type is a sudden fatal error, the preliminary memory health score is 0.

14. The memory status detection method according to claim 11, characterized in that: The enabling error repair operation includes at least one of the following: consuming redundant rows within the particle to replace faulty rows, enabling Bank-level adaptive dual DRAM device correction ADDDC function, enabling Rank-level adaptive dual DRAM device correction ADDDC function, and enabling storm suppression function.

15. The memory status detection method according to claim 14, characterized in that: The scores corresponding to the enabling error repair operations are ranked from small to large as follows: consuming PCLS, enabling the Bank-level ADDDC function, enabling the Rank-level ADDDC function, and enabling the storm suppression function.

16. The memory status detection method according to claim 1, characterized in that: The data set corresponding to the input-output model includes four columns of vector data, and the input-output model includes three input data corresponding to one output data.

17. The memory status detection method according to claim 16, characterized in that: The three input data include the average voltage, average frequency and average erase speed data corresponding to the memory, and the output data includes the average temperature data corresponding to the memory.

18. A memory status detection device, characterized in that: Applied to a switch, the device comprises: Acquisition module, used to obtain memory data; A division module, used for dividing the memory data into first memory data and second memory data according to the type of the memory data; a first determination module, configured to determine a preliminary memory health score according to a preset memory health assessment model and the first memory data, and to process the second memory data through an input-output model to determine a health influencing factor, wherein the preset memory health assessment model is determined based on the first historical memory data, and the input-output model is generated by training a preset initial model based on the second historical memory data; The second determination module is used to determine the current memory state according to the preliminary memory health score and the health impact factor.

19. A communication device, characterized in that: include: A transceiver, a memory, a processor, and a program stored on the memory and executable on the processor; The processor is used to read the program in the memory to implement the memory status detection method as described in any one of claims 1-17.

20. A non-volatile readable storage medium for storing a program, characterized in that: When the program is executed by a processor, the memory status detection method as described in any one of claims 1 to 17 is implemented.

Images