Memory fault prediction method and related device

By establishing a model to predict the probability of memory failure, updating the sample label and iterative training, and determining the test area with cluster analysis and Laida criterion, the problem of insufficient memory failure prediction accuracy in the prior art is solved, and the system reliability and Availability.

CN119046060APending Publication Date: 2024-11-29HUAWEI TECH CO LTD +1
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202310615819.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-29
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

The prior art is difficult to effectively predict memory failures, which affects the reliability, availability and serviceability of the system and insufficient accuracy of fault prediction.

Method used

By establishing a model to predict the probability of memory failure, and updating the model sample label based on the stress test results, performing multiple iterative training to improve the coverage and accuracy of the model, determining the boundary sample area for clustering analysis and Laida criterion to determine the waiting time Test area, narrow the pressure test range.

Benefits of technology

It significantly improves the accuracy of memory failure prediction, reduces the risk of system failure, and improves the reliability and availability of the system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119046060A_ABST
    Figure CN119046060A_ABST
Patent Text Reader

Abstract

The invention discloses a fault prediction method for a memory, the method is applied to the technical field of storage, and the method comprises the following steps: predicting according to a first model to obtain a first probability that the memory has a fault in a first time period; determining a result of the memory after the pressure test according to the first probability; updating a sample label of the first model based on a result after the pressure test; performing iterative training on a first model based on the updated sample label to obtain a trained first model; and obtaining a fault prediction result of the memory based on the trained first model. Therefore, the accuracy of fault prediction of the memory is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method for predicting a fault of a memory and a related device. Background Art

[0002] Memory failure is one of the most common failures in computer systems, which greatly affects the reliability, availability and serviceability (RAS) of the system. Therefore, it is very important to predict memory failure in a timely manner. Summary of the invention

[0003] The present application provides a method for predicting a fault of a memory, which can improve the accuracy of predicting a fault of the memory.

[0004] In a first aspect, the present application provides a method for predicting a fault of a memory, comprising:

[0005] First, a first probability of a memory failure within a first time period is predicted based on the first model. Next, a result of a stress test on the memory is determined based on the first probability. Then, a sample label of the first model is updated based on the result of the stress test. Finally, the first model is iteratively trained based on the updated sample label to obtain a fault prediction result of the memory based on the trained first model.

[0006] In the present application, after the first model completes the first fault prediction, the memory is stress tested based on the predicted first probability, and the sample labels of the first model are updated based on the results after the stress test, so that the model is iteratively trained for multiple rounds based on the sample labels, thereby greatly improving the coverage and accuracy of the model, and further improving the accuracy of fault prediction for the memory.

[0007] In a possible implementation manner of the first aspect, determining a result of a stress test on a memory according to a first probability includes:

[0008] Determining a first area in the memory to be stress tested according to the first probability;

[0009] A stress test is performed on the first area to obtain a result after the stress test.

[0010] In this possible implementation, a first area in the memory to be stress tested is determined according to the first probability, so as to further test the performance of the memory through the stress test.

[0011] In a possible implementation manner of the first aspect, determining a first area to be stress-tested in a memory according to a first probability includes:

[0012] When the first condition is met, determining a boundary sample according to a comparison result of the first probability and the first range;

[0013] If a region belonging to a boundary sample is screened out from the region where correctable errors occur in the memory, a cluster analysis is performed on the region, and a first region to be subjected to stress testing in the memory is determined based on the Laida criterion.

[0014] In this possible implementation, compared to random sampling, when the first condition is met, boundary samples are determined based on the comparison result of the first probability and the first range, which can be more conducive to the rapid convergence of the first model. Furthermore, since a large range has a great impact on the performance of the system, compared to stress testing all areas in the memory where correctable errors occur, the first area in the memory to be stress tested is determined from the area through cluster analysis and the Laida criterion, thereby further narrowing the scope of the stress test.

[0015] In a possible implementation manner of the first aspect, the first condition includes:

[0016] The first probability is greater than or equal to a first threshold, and the memory is not subjected to stress testing.

[0017] In a possible implementation manner of the first aspect, determining a boundary sample according to a comparison result of the first probability and the first range includes:

[0018] When the first probability belongs to the first range, a region in the memory corresponding to the first probability is determined as a boundary sample.

[0019] In this possible implementation, compared to random sampling, by setting the first range and screening out the area within the first range where the correctable error occurs in the memory, the area with the predicted first probability is helped to further quickly converge the first model.

[0020] In a possible implementation manner of the first aspect, predicting, according to the first model, a first probability that a memory fails within a first time period includes:

[0021] Obtaining error information of correctable errors occurring in the memory;

[0022] Error information of correctable errors occurring in the memory is used as input of the first model, and a first probability of uncorrectable errors occurring in the memory within a first time period is output.

[0023] In a possible implementation manner of the first aspect, updating the sample labels of the first model based on the result after the stress test includes:

[0024] When the result after the stress test does not include uncorrectable errors, the sample labels of the first model are updated based on error information of correctable errors occurring in the memory and the result after the stress test.

[0025] In this possible implementation method, if the record after the stress test does not include uncorrectable errors, it means that the risk of uncorrectable errors occurring in the area of ​​CE belonging to boundary samples in the memory is low. On this basis, the sample labels of the first model are updated by collecting information on the area of ​​CE belonging to boundary samples in the memory and the corresponding results after the stress test, so that the first model can be iteratively trained, the model parameters can be optimized, and the accuracy of model prediction can be improved.

[0026] In a second aspect, the present application provides a memory fault prediction device, comprising:

[0027] A model prediction module, configured to predict, based on a first model, a first probability of a memory failing within a first time period;

[0028] A stress test module, used to determine a result of a stress test on the memory according to a first probability;

[0029] A label updating module, used for updating the sample labels of the first model based on the result after the stress test;

[0030] A model training module, used for iteratively training the first model based on the updated sample labels to obtain a trained first model;

[0031] The model prediction module is also used to obtain a fault prediction result of the memory based on the trained first model.

[0032] In a possible implementation manner of the second aspect, the stress testing module is further used to:

[0033] Determining a first area in the memory to be stress tested according to the first probability;

[0034] A stress test is performed on the first area to obtain a result after the stress test.

[0035] In a possible implementation manner of the second aspect, determining a first area to be stress-tested in a memory according to a first probability includes:

[0036] When the first condition is met, determining a boundary sample according to a comparison result of the first probability and the first range;

[0037] If a region belonging to a boundary sample is screened out from the region where correctable errors occur in the memory, a cluster analysis is performed on the region, and a first region to be subjected to stress testing in the memory is determined based on the Laida criterion.

[0038] In a possible implementation manner of the second aspect, the first condition includes:

[0039] The first probability is greater than a first threshold, and the memory is not stress tested.

[0040] In a possible implementation manner of the second aspect, determining a boundary sample according to a comparison result of the first probability and the first range includes:

[0041] When the first probability belongs to the first range, a region in the memory corresponding to the first probability is determined as a boundary sample.

[0042] In a possible implementation manner of the second aspect, the model prediction module is further used to:

[0043] Obtaining error information of correctable errors occurring in the memory;

[0044] Error information of correctable errors occurring in the memory is used as input of the first model, and a first probability of uncorrectable errors occurring in the memory within a first time period is output.

[0045] In a possible implementation manner of the second aspect, the model prediction module is further used to:

[0046] When the result after the stress test does not include uncorrectable errors, the sample labels of the first model are updated based on error information of correctable errors occurring in the memory and the result after the stress test.

[0047] In the second aspect of the present application, the various units included in the fault prediction device of the memory can be used to implement the steps in the various possible implementation methods of the first aspect. For the specific implementation methods of the second aspect of the embodiment of the present application and certain steps in the various possible implementation methods of the second aspect, as well as the beneficial effects brought about by each possible implementation method, you can refer to the description of the various possible implementation methods in the first aspect, and will not be repeated here one by one.

[0048] In a third aspect, the present application provides a computer device, the computer device comprising a memory and a processor;

[0049] Memory for storing computer programs;

[0050] A processor is used to execute a computer program to implement the method in the above-mentioned first aspect or any possible implementation manner of the first aspect.

[0051] In a fourth aspect, the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the method in the above-mentioned first aspect or any possible implementation manner of the first aspect is implemented.

[0052] In a fifth aspect, the present application provides a computer program product, which, when executed by a computer, implements the method in the above-mentioned first aspect or any possible implementation manner of the first aspect.

[0053] In a sixth aspect, the present application provides a circuit system, which includes a processing circuit, and the processing circuit is configured to execute the method in the above-mentioned first aspect or any possible implementation manner of the first aspect.

[0054] In a seventh aspect, the present application provides a chip system, including a processor and a memory, the memory being used to store a computer program, and the processor being used to call and run the computer program stored in the memory to execute the method of the first aspect or any possible implementation of the first aspect. The chip system may be composed of a chip, or may include a chip and other discrete devices.

[0055] The solutions of the second to seventh aspects mentioned above are used to implement or cooperate with the method of the first aspect or any possible implementation method thereof, and therefore can achieve the same or corresponding beneficial effects as the first aspect, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1a A schematic diagram of a structure of a computing device provided in an embodiment of the present application;

[0057] Figure 1b A schematic diagram of a flow chart of a method for predicting a fault of a memory provided in an embodiment of the present application;

[0058] Figure 2 Another schematic diagram of a flow chart of a method for predicting a fault of a memory provided in an embodiment of the present application;

[0059] Figure 3 Another schematic diagram of a flow chart of a method for predicting a fault of a memory provided in an embodiment of the present application;

[0060] Figure 4 Another schematic diagram of a flow chart of a method for predicting a fault of a memory provided in an embodiment of the present application;

[0061] Figure 5 Another schematic diagram of a flow chart of a method for predicting a fault of a memory provided in an embodiment of the present application;

[0062] Figure 6 Another schematic diagram of a flow chart of a method for predicting a fault of a memory provided in an embodiment of the present application;

[0063] Figure 7 Another schematic diagram of a flow chart of a method for predicting a fault of a memory provided in an embodiment of the present application;

[0064] Figure 8 A schematic diagram of a structure of a fault prediction device for a memory provided in an embodiment of the present application;

[0065] Fig. 9 Another structural schematic diagram of a fault prediction device for a memory provided in an embodiment of the present application. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of this application. It is known to those skilled in the art that with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0067] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0068] The term "and / or" in this application can be a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this application generally indicates that the associated objects before and after are in an "or" relationship.

[0069] It should also be noted that, in some alternative implementations, the functions / acts noted may occur out of the order of the drawings. For example, two figures shown in succession may in fact occur substantially simultaneously or may sometimes be performed in the reverse order, depending on the functions / acts involved.

[0070] In the embodiments of the present application, unless otherwise specified, the meaning of "at least one" refers to one or more, and the meaning of "plurality" refers to two or more. It is understood that in the present application, "when", "if" and "if" all refer to the device making corresponding processing under certain objective circumstances, and do not limit the time, nor do they require that there must be a judgment action when the device is implemented, nor do they mean that there are other limitations. In addition, the special word "exemplary" means "used as an example, embodiment or illustrative". Any embodiment described as "exemplary" is not necessarily interpreted as being superior or better than other embodiments.

[0071] The embodiments of the present application are described below in conjunction with the accompanying drawings. Those skilled in the art will appreciate that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0072] First of all, in order to better understand the solutions of the embodiments of the present application, the relevant terms and concepts that may be involved in the embodiments of the present application are introduced below.

[0073] (1) Baseboard Management Controller

[0074] A baseboard management controller (BMC) is an independent service processor used to manage and monitor various hardware devices in a computer system. For example, the BMC can be used to query, warn, and report errors of hardware devices.

[0075] (2) Basic Input and Output System

[0076] The basic input output system (BIOS) can be understood as a set of programs fixed on a ROM chip on the motherboard of the computer. It stores the most important basic input and output programs of the computer, the self-test program after powering on, and the system self-starting program.

[0077] (3) Stress testing

[0078] The purpose of stress testing memory is to evaluate the performance and stability of the memory to ensure that the memory can meet specific application requirements. For example, by simulating memory usage under high load, errors and problems in the memory can be detected, which helps to ensure the reliability of the memory and data security.

[0079] In order to better understand the solution of the embodiment of the present application, Figure 1a , an example is given to illustrate the application scenario of the memory fault prediction method provided in the embodiment of the present application.

[0080] See also Figure 1a , Figure 1a A schematic diagram of a computing device provided in an embodiment of the present application is shown in FIG. 1 : The computing device 100 includes a processor 101 , a memory 102 , a BIOS 103 , and a BMC 104 .

[0081] In this embodiment, the computing device 100 refers to a computing device with computing, communication and other functions, such as a server, a mobile phone, a computer, a wearable device, etc. The computing device 100 may also be a computing device in a distributed computing system.

[0082] The processor 101 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0083] The memory 102 may include a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), and may also include a volatile memory, such as a random access memory (RAM), a static random access memory (SRAM), a dynamic random access memory (DRAM), a phase change memory (PCM), a resistance random access memory (RRAM), a magnetic random access memory (MRAM), a ferroelectric random access memory (FeRAM), and the like.

[0084] In a specific implementation, as an embodiment, the RAM can be implemented by dual inline memory modules (DIMM) or double data rate synchronous dynamic random access memory (DDR SDRAM), also known as DDR.

[0085] The BMC 104 may call program codes to implement the memory fault prediction method provided in the embodiment of the present application. The specific implementation may refer to the description below.

[0086] Exemplarily, the processor 101 is mainly used to detect whether an error occurs in the memory 102. After the processor 101 detects that an error occurs in the memory 102, the error is reported to the BIOS 103, and the BIOS 103 reports the error to the BMC 104 for processing. After the BMC 104 processes the error, the processing result is fed back to the BIOS 103, and the BIOS 103 issues a command to trigger the processor 101 to perform a corresponding operation, and the BIOS 103 accesses the memory 102 through the processor 101.

[0087] It should be noted that Figure 1a The structure of the computing device shown is only one example. The fault prediction method of the memory provided in the embodiment of the present application can be applied to the structure described above but is not limited to it. For example, the computing device may also include a communication interface, etc. Figure 1a Not shown in FIG.

[0088] It is understandable that as the memory architecture evolves, in order to fit more storage units into the same silicon wafer area and perform faster reading and writing, the memory process line width is getting smaller and smaller, and the operating frequency is getting higher and higher. In order to control heat, the operating voltage is forced to be continuously reduced, which reduces the reliability of the storage medium itself.

[0089] In modern large data centers, there are usually millions of servers working together to provide high-performance computing and big data storage services. A large number of tasks are usually running on these servers, and the occurrence of hardware failures will have a great impact on the reliability, availability and serviceability (RAS) of the servers. Among them, DRAM errors are the most common threat to hardware. Servers are usually equipped with advanced error correction code (ECC) mechanisms to deal with DRAM errors. However, relying solely on ECC to ensure the reliability of DRAM is far from enough. DRAM failures are still one of the main causes of system failures. When DRAM errors exceed the correction capacity of ECC, they become UCE, i.e. DRAM failures, causing further server shutdowns. According to statistics, in a modern data center using DDR4 DRAM, up to 30% of hosts have CE errors and up to 10% of hosts have UCE errors each year. These UCE errors may cause business processes to hang, restart or host downtime, resulting in short-term business interruption, data loss or performance degradation.

[0090] Current computing and storage products are seriously challenged by the quality of storage devices. Although existing RAS technologies can alleviate these challenges, they are limited by the accuracy of judgment. Therefore, it is urgent to effectively analyze storage risks and intelligently improve the efficiency of RAS. Among them, analyzing the characteristics of storage errors, predicting storage UCE errors and managing risk areas is an effective RAS measure that can prevent and control problems.

[0091] based on Figure 1a The computing device shown, at present, memory failure is one of the most common failures of computer systems, which greatly affects the reliability, availability and serviceability (RAS) of the system. Therefore, it is very important to predict the failure of the memory in time.

[0092] In order to improve the accuracy of memory fault prediction, the present application embodiment provides a memory fault prediction method, which is applied to Figure 1a BMC 104 is shown. For details, see Figure 1b , Figure 1bA flow chart of a method for predicting a fault of a memory provided in an embodiment of the present application. The method specifically includes: S1. Predicting a first probability of a fault of the memory within a first time period based on a first model. S2. Determining the result of a stress test on the memory based on the first probability. S3. Updating the sample labels of the first model based on the result after the stress test. S4. Iteratively training the first model based on the updated sample labels to obtain a trained first model. S5. Obtaining a fault prediction result of the memory based on the trained first model.

[0093] It can be understood that after the first model completes the first fault prediction, the memory is stress tested based on the predicted first probability, and the sample labels of the first model are updated based on the results of the stress test, so that the model is iteratively trained for multiple rounds based on the sample labels, thereby greatly improving the coverage and accuracy of the model, and further improving the accuracy of the fault prediction of the memory.

[0094] For ease of explanation, the following specifically describes the memory fault prediction method provided by the embodiment of the present application in combination with the accompanying drawings and application scenarios. A person skilled in the art will appreciate that, with the development of technology and the emergence of new scenarios, the technical solution provided by the embodiment of the present application is also applicable to similar technical problems.

[0095] See also Figure 2 , Figure 2 Another flow chart of a method for predicting a fault of a memory provided in an embodiment of the present application is shown in FIG. Figure 1a It is understood that, if Figure 1a The memory 102 shown includes multiple types, and the BMC monitors the operating status of the memory by detecting the memory 102. For ease of explanation, the memory is taken as an example below, and each DIMM in the memory is taken as a detection object. The BMC monitors the operating status of the memory by detecting each DIMM in the memory.

[0096] like Figure 2 As shown, the method may include the following steps 201 to 219.

[0097] Step 201: The BMC obtains error information of the DIMM.

[0098] In the embodiment of the present application, the error information of the DIMM is information about a correctable error (CE) that occurs during the operation of the DIMM. The error information of the DIMM may include: the time of occurrence of the CE, the physical address information of the CE, the manufacturer and batch of the DIMM, etc. The embodiment of the present application does not limit the specific content of the error information of the DIMM, which is only an exemplary description.

[0099] Exemplarily, the physical address information of the CE is used to indicate the physical location of the CE in the memory. That is, the specific physical location of the CE in the memory can be determined based on the physical address information of the CE. For example, the physical address information of the CE may include: the CPU node number where the CE is located, the socket number where the CE occurs, the channel number, the DIMM number, the memory chip number, the logical array group number, the logical array number, the row number, the column number or the data line number, the burst sequence number, and one or more of the above information.

[0100] The embodiment of the present application does not limit the specific content of the physical address information of the above CE, which is only an exemplary description.

[0101] It is understandable that after a memory failure occurs, the memory itself has an error correction algorithm, such as an error correction code (ECC) to correct the error. The corrected error is called a correctable error (CE). The error correction algorithm has a certain error correction capability, but the capability is limited. If the error correction capability of the error correction algorithm is exceeded, an uncorrectable error (UCE) will occur, causing the device to crash.

[0102] Optionally, the BMC may receive DIMM error information from the BIOS.

[0103] Exemplarily, after the CPU identifies an error, it sends error information to the BIOS, and the BIOS then sends the error information to the BMC.

[0104] It should be noted that the BMC may also receive error information sent by other processor firmware, for example, firmware, management engine (ME) or intelligent management unit (IMU), and the embodiments of the present application are not limited to this.

[0105] Step 202: The BMC uses the error information of the DIMM as an input of a first model and outputs a first probability.

[0106] In the embodiment of the present application, after obtaining the error information of the DIMM, the BMC inputs the error information of the DIMM into the first model, performs a first round of fault prediction, and calculates a first probability P of the DIMM having UCE in a first time period in the future.

[0107] Step 203: The BMC determines whether the first probability is less than a first threshold. If yes, step 204 is executed. If no, step 207 is executed.

[0108] In the embodiment of the present application, the BMC compares the first probability P with the first threshold value P0. If P is less than P0, it can be determined that the risk of UCE occurring in the first time period in the future of the DIMM is low, and observe whether UCE occurs in the first time period in the future to verify the prediction accuracy of the first model. If P is greater than or equal to P0, it is determined that the risk of UCE occurring in the first time period of the DIMM is high, and further checks whether the DIMM has been stress tested.

[0109] It should be noted that, in actual application, the size of the first threshold can be set according to actual needs or experiments and is not limited here.

[0110] It is worth noting that in the related art, the coverage and accuracy of the model are seriously unbalanced and lack practical application value. In an embodiment of the present application, the prediction coverage of the model is adjusted by adjusting the size of the first threshold P0. The lower the value of P0 is set, the more sample labels of the model are obtained, thereby improving the coverage of the first model. On this basis, the first model is iteratively trained through sample labels to further improve the prediction accuracy of the model.

[0111] Step 204, the BMC determines whether a UCE occurs in the DIMM within the first time period. If yes, execute steps 205 to 206. If no, return to step 201.

[0112] In the embodiment of the present application, if the BMC determines that the DIMM has a UCE in the first time period, it means that the prediction accuracy of the first model is low, and the first model needs to be retrained to improve the prediction accuracy. If the BMC determines that the DIMM has not had a UCE in the first time period, it means that the prediction accuracy of the first model is high, and there is no need to retrain, and wait for the next CE to occur in the DIMM to trigger the first model for fault prediction.

[0113] Step 205: The BMC updates the sample tag according to the CE information and the UCE information.

[0114] In an embodiment of the present application, if the BMC determines that UCE has occurred in the DIMM within the first time period, the sample label corresponding to the UCE can be updated according to the CE information and the UCE information, so that the first model learns that after the CE information is input, the area where the CE occurs will experience UCE within the first time period.

[0115] Step 206: BMC inputs the updated sample labels into the first model for iterative training to update the first model.

[0116] In the embodiment of the present application, the BMC re-updates the sample labels and inputs the updated sample labels into the first model, iteratively trains the first model, and updates the parameters of the first model to improve the accuracy of the first model's prediction.

[0117] Exemplarily, if the BMC predicts through the first model that the probability P of the DIMM experiencing UCE in the first time period in the future is less than P0, then it is determined that the risk of the DIMM experiencing UCE in the first time period in the future is low. On this basis, the BMC verifies the accuracy of the first model by detecting whether the DIMM will actually experience UCE in the first time period in the future. If the DIMM experiences UCE in the first time period in the future, it means that the model prediction is incorrect, and the sample label is updated by collecting the error information of the DIMM and the information of UCE, and the updated sample label is input into the first model for iterative training. If the BMC predicts through the first model that the probability P of the DIMM experiencing UCE in the first time period in the future is greater than or equal to P0, then it is determined that the risk of the DIMM experiencing UCE in the first time period in the future is high, and further detects whether the DIMM has been stress tested.

[0118] Step 207, the BMC detects whether the DIMM has been stress tested. If yes, the BMC issues an isolation or replacement command to the area in the DIMM where the CE occurs. If no, execute steps 208 to 209.

[0119] In the embodiment of the present application, if the BMC predicts based on the first model that the risk of UCE occurring in the CE region within the DIMM within the first time period is high, it is further determined whether the region has been stress tested. The range of the region can be a single address, a single row, multiple rows, a single bank, multiple banks, a single particle, multiple particles, a single rank, multiple ranks, and a whole DIMM. If the DIMM has been stress tested, reliability measures are directly taken for the region, such as isolation or replacement.

[0120] If the BMC detects that the DIMM has been stress-tested, the BMC sends an isolation or replacement command to the BIOS, which then reports the command to the CPU to trigger the CPU to perform isolation or replacement actions, and to partially isolate or replace the area where the CE occurs, and no longer use the area to avoid reading and writing the area again. The replacement action can include hardware replacement and software replacement. Hardware replacement can be manual replacement of the DIMM. Software replacement can be, for example, the CPU using a normally operating area to replace an erroneous area for read and write operations, etc., and can be set according to actual needs.

[0121] In an optional implementation, the BMC determines the value of the first flag to detect whether the memory has been subjected to a stress test.

[0122] Specifically, when the BMC detects that the value of the first mark is 1, it determines that the DIMM has been subjected to a stress test. When the BMC detects that the value of the first mark is 0, it determines that the DIMM has not been subjected to a stress test.

[0123] It is understandable that the first mark is used to indicate whether the DIMM has been subjected to a pressure test. In actual application, the value of the first mark can be set according to actual needs.

[0124] Step 208: The BMC issues an isolation command to the area in the DIMM where the CE occurs, and determines a boundary sample according to a comparison result between the first probability and the first range.

[0125] In the embodiment of the present application, when the BMC detects that the DIMM has not been stress tested, the BMC will first isolate the area where CE occurs in the DIMM, and then determine the boundary samples based on the comparison result of the first probability and the first range. The boundary samples can be understood as the set of areas within the first range where the BMC predicts the probability of UCE occurring in the DIMM based on the first model. By setting the first range, the BMC filters out the boundary samples belonging to the first range from the area where CE occurs.

[0126] In an optional implementation, the first range includes a range that is greater than or equal to the first threshold and less than or equal to the first threshold plus a specific value.

[0127] Specifically, the specific value can be set according to actual needs, such as 0.2, etc., and the first range is from PO to PO+0.2.

[0128] It is worth noting that, compared with the random determination method, determining the area of ​​the boundary samples from the boundary range of the first threshold is more conducive to rapid convergence of the model.

[0129] Step 209, the BMC determines whether there is an area belonging to a boundary sample in the area where CE occurs in the DIMM. If yes, step 210 is executed. If not, an isolation or replacement command is issued to the area where CE occurs in the DIMM.

[0130] In the embodiment of the present application, the BMC sets a first range to screen out areas belonging to the boundary samples. For areas that do not belong to the boundary samples, the BMC determines that the probability of UCE is high and takes reliability measures, such as isolation or replacement. For areas where the probability of UCE is close to the first threshold, the BMC re-determines the range of the stress test from the area and performs the stress test to further evaluate the performance of the DIMM.

[0131] In step 210 , the BMC determines a first region to be stress-tested from the regions belonging to the boundary samples.

[0132] In the embodiment of the present application, after determining that there is an area belonging to a boundary sample in the area where CE occurs, the BMC re-determines the first area to be stress-tested from the area.

[0133] In an optional implementation, the BMC performs cluster analysis on the regions belonging to the boundary samples, and determines the first region in the DIMM to be stress tested based on the Laida criterion.

[0134] It is understandable that, in actual operation, the method for determining the pressure test range may be adjusted according to actual needs or experiments. This is only one of the optional methods and is not a limitation.

[0135] Step 211: The BMC performs a stress test on the first area to obtain a result of the stress test.

[0136] In the embodiment of the present application, after the BMC determines the first area to be stress-tested in the area belonging to the boundary sample, the BMC performs the stress test on the first area to be stress-tested to obtain a result after the stress test.

[0137] Step 212, the BMC determines whether the result after the stress test includes UCE. If yes, an isolation or replacement command is issued to the area in the DIMM where CE occurs. If no, steps 213 to 216 are executed.

[0138] In an embodiment of the present application, after the BMC performs a stress test on the first area and obtains the result after the stress test, it determines whether the result after the stress test includes UCE. If the result after the stress test includes UCE, it means that the risk of UCE occurring in the first area is high, and reliability measures need to be taken. The BMC triggers the CPU to isolate or replace the area where CE occurs in the DIMM by issuing an isolation or replacement command. If the result after the stress test does not include UCE, it means that the risk of UCE occurring in the area where CE occurs in the DIMM is low.

[0139] Step 213: The BMC sends a release command to the area in the DIMM where the CE occurs.

[0140] In the embodiment of the present application, if the result after the stress test does not include UCE, it means that the risk of UCE occurring in the area where CE occurs in the DIMM is low. The BMC issues a release command to the area and sends the command to the BIOS, which further sends the command to the CPU so that the CPU executes the operation of releasing the isolated area according to the command. In addition, the BMC marks the area, which indicates that the area has been stress tested.

[0141] It is understandable that after the BMC performs a stress test on the first area, the value of the first mark corresponding to the area is set to 1 to mark that the area has undergone a stress test. The content of the first mark can also be other values ​​or other forms, which can be set according to actual needs, as long as it can ensure that it can distinguish whether a stress test has been performed.

[0142] Step 214 : The BMC updates the sample label of the first model according to the information of the CE of the area within the boundary sample in the DIMM and the result after the stress test.

[0143] In the embodiment of the present application, when the BMC determines that no UCE occurs during the stress test according to the result after the stress test, it issues a command to release the area where the CE occurs and marks the area as having undergone the stress test. Subsequently, the BMC updates the sample label, using the error information of the DIMM and the information of the CE occurring in the result after the stress test as the low-risk sample label.

[0144] Step 215: BMC iteratively trains the first model according to the updated sample labels to update the first model.

[0145] In an embodiment of the present application, the BMC iteratively trains the first model based on the updated sample labels, optimizes the parameters of the first model, and generates an optimized first model.

[0146] It is understandable that in actual operation, multiple rounds of data collection, updating of sample labels, iteration of the first model, and then iterative prediction can be carried out to quickly converge false positives, so as to simultaneously improve the coverage and accuracy of the model and solve the problem that the coverage and accuracy cannot be taken into account at the same time when machine learning algorithms are used in related technologies.

[0147] Step 216, the BMC determines whether a UCE occurs in the DIMM within the first time period. If yes, execute steps 217 to 218. If no, return to step 201.

[0148] In an embodiment of the present application, after the BMC trains and optimizes the first model based on the updated sample labels, in order to further improve the prediction accuracy of the first model, for the areas released after the occurrence of CE in the DIMM, it is still necessary to monitor whether these areas have UCE occurring in the first time period in the future. If the BMC detects that UCE has not occurred in the first time period, there is no need to re-update the sample labels, and continue to wait for the next CE to trigger the fault prediction of the first model. If the BMC detects that UCE has occurred in the first time period, it means that the first model prediction is wrong, and the sample labels need to be re-updated according to the result of the UCE. If the BMC detects that UCE has not occurred in the first time period, it means that the prediction accuracy of the first model is high, and there is no need to re-train, and wait for the next CE to occur in the DIMM to trigger the first model for fault prediction.

[0149] Step 217 : The BMC updates the sample label of the first model according to the information of the CE of the area within the boundary sample in the DIMM and the result after the stress test.

[0150] In the embodiment of the present application, if the BMC detects that UCE occurs in the first time period, it means that the first model prediction is wrong. The BMC updates the sample label, uses the error information of the DIMM and the information of UCE and CE occurring in the result after the stress test as the high-risk sample label, and inputs the updated sample label into the first model for iterative training to generate the trained first model.

[0151] It is understandable that if the BMC detects that UCE occurs in the first time period, it means that CE may occur first during the stress test, and then UCE occurs in the first time period. Therefore, the BMC can modify the sample tag using the DIMM error information and the information of UCE and CE occurring in the result after the stress test to update the sample tag.

[0152] In step 218, the BMC iteratively trains the first model according to the updated sample labels to update the first model.

[0153] It should be noted that, for ease of explanation, the embodiments of the present application are described using each DIMM in the memory as an example. In actual application, other units in the memory or other memory types may be used as examples as research objects, which are not limited here.

[0154] It is worth noting that in the related art, there is a method for predicting the non-burst UCE of memory by counting the CE and UCE information of a large amount of memory for a long time, performing fault statistical analysis on a large amount of memory, and constructing a fault prediction model. However, this method focuses on reflecting the degree of memory failure by counting the number of CEs, and has a low dimension. It is equivalent to replacing a large amount of memory in exchange for system reliability, which increases the workload and cost of operation and maintenance. In an embodiment of the present application, after the first model completes the first fault prediction, the coverage of the model can be flexibly set based on the relationship between the predicted first probability and the first threshold. By reasonably setting the size of the first threshold, the amount of data of the sample label of the model is increased, thereby improving the coverage of the model prediction. On this basis, the memory is stress tested based on the predicted first probability, and the sample label of the first model is updated based on the result after the stress test, so as to perform multiple rounds of iterative training on the model based on the sample label, thereby greatly improving the coverage and accuracy of the model, and thus improving the prediction accuracy of the memory failure.

[0155] In order to more intuitively understand the fault prediction method of the memory provided in the embodiments of the present application, different implementation methods are described below with examples.

[0156] (1) Embodiment 1

[0157] See also Figure 3 , Figure 3 Another flowchart of a method for predicting a memory failure provided by an embodiment of the present application. Assume that the first probability P = 0.1, the first threshold P0 = 0.3, P is less than P0, and the DIMM does not experience UCE in the first time period. Figure 3 As shown, the method may include the following steps 301 to 304.

[0158] Step 301: The BMC obtains error information of the DIMM.

[0159] Step 302: The BMC uses the error information of the DIMM as the input of the first model and outputs a first probability P.

[0160] Step 303: The BMC determines that P is less than P0.

[0161] Step 304 : The BMC determines that no UCE occurs in the DIMM within the first time period, and returns to step 301 .

[0162] In the embodiment of the present application, assuming that no UCE occurs in the DIMM within the first time period, it means that the prediction accuracy of the first model is high, and there is no need to retrain, and wait for the next CE to occur in the DIMM to trigger the first model for fault prediction.

[0163] (2) Embodiment 2

[0164] See also Figure 4 , Figure 4 Another flowchart of a method for predicting a memory failure provided by an embodiment of the present application. Assume that the first probability P = 0.2, the first threshold P0 = 0.3, P is less than P0, and a UCE occurs in the DIMM in the first time period. Figure 4 As shown, the method may include the following steps 401 to 406.

[0165] Step 401: The BMC obtains error information of the DIMM.

[0166] Step 402: The BMC uses the error information of the DIMM as the input of the first model and outputs a first probability P.

[0167] Step 403: The BMC determines that P is less than P0.

[0168] Step 404: The BMC determines that a UCE occurs in the DIMM within the first time period.

[0169] In the embodiment of the present application, assuming that UCE occurs in the DIMM within the first time period, it means that the prediction accuracy of the first model is low, and the first model needs to be retrained to improve the prediction accuracy.

[0170] Step 405: The BMC updates the sample tag according to the CE information and the UCE information.

[0171] Step 406: BMC inputs the updated sample labels into the first model for iterative training to update the first model.

[0172] (3) Embodiment 3

[0173] See also Figure 5 , Figure 5 Another flowchart of a method for predicting a memory failure provided by an embodiment of the present application. Assume that the first probability P = 0.6, the first threshold P0 = 0.3, P is greater than P0, and the area where CE occurs in the DIMM has been stress tested. Figure 5 As shown, the method may include the following steps 501 to 505.

[0174] Step 501: The BMC obtains error information of the DIMM.

[0175] Step 502: The BMC uses the error information of the DIMM as the input of the first model and outputs a first probability P.

[0176] Step 503: The BMC determines that P is greater than P0.

[0177] Step 504: The BMC detects whether the DIMM has been stress tested.

[0178] Step 505: The BMC sends an isolation or replacement command to the area in the DIMM where the CE occurs.

[0179] (4) Embodiment 4

[0180] See also Figure 6 , Figure 6 Another flowchart of a method for predicting a memory failure provided by an embodiment of the present application. Assume that the first probability P = 0.8, the first threshold P0 = 0.3, P is greater than P0, and the area where CE occurs in the DIMM has not been stress tested, but the area where CE occurs in the DIMM does not belong to the boundary sample. Figure 6 As shown, the method may include the following steps 601 to 607.

[0181] Step 601: The BMC obtains error information of the DIMM.

[0182] Step 602: The BMC uses the error information of the DIMM as the input of the first model and outputs a first probability P.

[0183] Step 603: The BMC determines that P is greater than P0.

[0184] Step 604: The BMC detects that the DIMM has not been subjected to a stress test.

[0185] Step 605: The BMC issues an isolation command to the area in the DIMM where the CE occurs, and determines a boundary sample according to a comparison result between the first probability and the first range.

[0186] Step 606 , the BMC determines that the area where CE occurs in the DIMM does not have an area belonging to a boundary sample.

[0187] Step 607: The BMC sends an isolation or replacement command to the area in the DIMM where the CE occurs.

[0188] (5) Embodiment 5

[0189] See also Figure 7 , Figure 7Another flowchart of a method for predicting a memory failure provided by an embodiment of the present application. Assume that the first probability P = 0.8, the first threshold P0 = 0.3, P is greater than P0, and the area where CE occurs in the DIMM has not been stress tested, but there is an area belonging to a boundary sample in the area where CE occurs in the DIMM, and the result after the stress test does not include UCE, and no UCE is observed in the first time period. Figure 7 As shown, the method may include the following steps 701 to 713.

[0190] Step 701: The BMC obtains error information of the DIMM.

[0191] Step 702: The BMC uses the error information of the DIMM as the input of the first model and outputs a first probability P.

[0192] Step 703: The BMC determines that P is greater than P0.

[0193] Step 704: The BMC detects that the DIMM has not been subjected to a stress test.

[0194] Step 705: The BMC issues an isolation command to the area in the DIMM where the CE occurs, and determines a boundary sample according to a comparison result between the first probability and the first range.

[0195] Step 706 , the BMC determines whether there is an area belonging to a boundary sample in the area where CE occurs in the DIMM.

[0196] Step 707 : The BMC determines a first region to be subjected to a stress test from the regions belonging to the boundary samples.

[0197] Step 708: The BMC performs a stress test on the first area to obtain a result of the stress test.

[0198] Step 709: The BMC determines that the result after the stress test does not include UCE.

[0199] Step 710: The BMC sends a release command to the area in the DIMM where the CE occurs.

[0200] Step 711 : The BMC updates the sample label of the first model according to the information of the CE of the area within the boundary sample in the DIMM and the result after the stress test.

[0201] Step 712: BMC iteratively trains the first model according to the updated sample labels to update the first model.

[0202] Step 713 : The BMC determines that no UCE occurs in the DIMM within the first time period, and returns to step 701 .

[0203] In Figures 1 to Figure 7On the basis of the corresponding embodiments, in order to better implement the above solutions of the embodiments of the present application, the following also provides related devices for implementing the above solutions. Figure 8 , Figure 8 A schematic diagram of a structure of a memory fault prediction device provided in an embodiment of the present application, wherein the memory fault prediction device may be a BMC as shown in 1a, or may be located on the BMC as shown in 1a. The memory fault prediction device 8000 includes:

[0204] A model prediction module 8001 is used to predict a first probability of a memory failure within a first time period according to a first model;

[0205] A stress test module 8002, configured to determine a result of a stress test on the memory according to a first probability;

[0206] The label updating module 8003 is used to update the sample labels of the first model based on the result after the stress test;

[0207] A model training module 8004 is used to iteratively train the first model based on the updated sample labels to obtain a trained first model;

[0208] The model prediction module 8005 is also used to obtain a fault prediction result of the memory based on the trained first model.

[0209] In a possible implementation, the stress testing module 8002 is further used to:

[0210] Determining a first area in the memory to be stress tested according to the first probability;

[0211] A stress test is performed on the first area to obtain a result after the stress test.

[0212] In a possible implementation, determining a first area in a memory to be subjected to a stress test according to a first probability includes:

[0213] When the first condition is met, determining a boundary sample according to a comparison result of the first probability and the first range;

[0214] If a region belonging to a boundary sample is screened out from the region where correctable errors occur in the memory, a cluster analysis is performed on the region, and a first region to be subjected to stress testing in the memory is determined based on the Laida criterion.

[0215] In a possible implementation, the first condition includes:

[0216] The first probability is greater than a first threshold, and the memory is not stress tested.

[0217] In a possible implementation, determining a boundary sample according to a comparison result of the first probability and the first range includes:

[0218] When the first probability belongs to the first range, a region in the memory corresponding to the first probability is determined as a boundary sample.

[0219] In a possible implementation, the model prediction module 8005 is further used to:

[0220] Obtaining error information of correctable errors occurring in the memory;

[0221] Error information of correctable errors occurring in the memory is used as input of the first model, and a first probability of uncorrectable errors occurring in the memory within a first time period is output.

[0222] In a possible implementation, the model prediction module 8005 is further used to:

[0223] When the result after the stress test does not include uncorrectable errors, the sample labels of the first model are updated based on error information of correctable errors occurring in the memory and the result after the stress test.

[0224] In this embodiment, the operations performed by each module in the memory fault prediction device 8000 are similar to those described in the aforementioned method embodiment, and can be used to implement the functions of the BMC in the aforementioned method embodiment, and can also achieve the beneficial effects of the aforementioned method embodiment, which will not be repeated here.

[0225] See also Fig. 9 , Fig. 9 Another structural diagram of the fault prediction device of the memory provided in the embodiment of the present application. Fig. 9 As shown, the memory fault prediction device 900 is implemented by a general bus architecture.

[0226] The memory failure prediction device 900 includes at least one processor 901 , a communication bus 902 , a memory 903 , and at least one communication interface 904 .

[0227] The processor 901, the memory 903, and the communication interface 904 communicate via the communication bus 902, and may also communicate via other means such as wireless transmission. The memory 903 is used to store instructions, and the processor 901 is used to execute the instructions stored in the memory 903. The memory 903 stores program codes, and the processor 901 can call the program codes stored in the memory 903 to execute. Figure 2 Steps 201 to 219 in the illustrated embodiment.

[0228] Optionally, the processor 901 is a general-purpose central processing unit (CPU), and may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), or other programmable logic devices (PLD), transistor logic devices, hardware components, or any combination thereof. The above-mentioned PLD is a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0229] The communication bus 902 is used to transmit information between the processor 901, the memory 903 and the communication interface 904. The communication bus 902 is divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0230] Optionally, the memory 903 is a read-only memory (ROM) or other types of static storage devices that can store static information and instructions. Alternatively, the memory 903 is a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions. Alternatively, the memory 903 is an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to this. Optionally, the memory 903 exists independently and is connected to the processor 901 through the communication bus 902. Optionally, the memory 903 and the processor 901 are integrated together.

[0231] The communication interface 904 uses any transceiver-like device for communicating with other devices or communication networks. The communication interface 904 includes a wired communication interface. Optionally, the communication interface 904 also includes a wireless communication interface. Among them, the wired communication interface is, for example, an Ethernet interface. The Ethernet interface is an optical interface, an electrical interface or a combination thereof. The wireless communication interface is a wireless local area network (WLAN) interface, a cellular network communication interface or a combination thereof, etc.

[0232] In a specific implementation, as an embodiment, the processor 901 includes one or more CPUs, such as Fig. 9 CPU0 and CPU1 are shown in the figure.

[0233] In a specific implementation, as an embodiment, the memory fault prediction device 900 includes multiple processors, such as Fig. 9 901 and processor 905 are shown in FIG. Each of these processors is a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The processor here refers to one or more devices, circuits, and / or processing cores for processing data (such as computer program instructions).

[0234] In some embodiments, the memory 903 is used to store program codes for executing the solution of the present application, and the processor 901 executes the program codes stored in the memory 903. That is, the memory fault prediction device 900 implements the above-mentioned embodiment of the traffic scheduling method through the processor 901 and the program codes in the memory 903.

[0235] It is to be understood that the method steps in the embodiments of the present application may be implemented in hardware or in software instructions executable by the processor 901. The software instructions may be composed of corresponding software modules, and the software modules may be stored in a random access memory, a flash memory, a read-only memory, a programmable read-only memory, an erasable programmable read-only memory, an electrically erasable programmable read-only memory, a register, a hard disk, a mobile hard disk, a CD-ROM, or any other form of storage medium known in the art. In addition, the scope of the device described in the present application is not limited thereto, and the structure of the device may not be limited thereto. Fig. 9 A device may be a stand-alone device or may be part of a larger device. For example, a device may be:

[0236] (1) Independent integrated circuit IC, or chip, or chip system or subsystem;

[0237] (2) having a set of one or more ICs, and optionally, the IC set may also include a storage component for storing data and / or instructions;

[0238] (3) Modules that can be embedded in other devices;

[0239] (4) Receivers, terminals, intelligent terminals, wireless devices, handheld devices, mobile units, vehicle-mounted devices, artificial intelligence devices, machine equipment, household equipment, medical equipment, industrial equipment, etc.;

[0240] (5)Others

[0241] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0242] An embodiment of the present application further provides a computer storage medium, which stores a computer program. When the program is executed by a computer, the computer implements any of the above method embodiments.

[0243] The embodiment of the present application also provides a computer program product, which, when executed by a computer, implements the functions of any of the above method embodiments.

[0244] An embodiment of the present application also provides a circuit system, which includes a processing circuit, and the processing circuit is configured to execute any of the above method embodiments.

[0245] The autonomous driving vehicle provided in the embodiment of the present application may specifically be a chip, and the chip includes: a processing unit and a communication unit, wherein the processing unit may be, for example, a processor, and the communication unit may be, for example, an input / output interface, a pin or a circuit, etc. The processing unit may execute computer-executable instructions stored in the storage unit so that the chip executes any of the above-mentioned method embodiments. Optionally, the storage unit is a storage unit in the chip, such as a register, a cache, etc., and the storage unit may also be a storage unit located outside the chip in the wireless access device end, such as a read-only memory or other types of static storage devices that can store static information and instructions, a random access memory, etc. Among them, the processor mentioned in any of the above places may be a general-purpose central processing unit, a microprocessor, an ASIC, or one or more integrated circuits for controlling the execution of the program of the above-mentioned first aspect method.

[0246] It is understood that the devices and methods described in the present application can also be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0247] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0248] The naming or numbering of the steps in this application does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The named or numbered process steps can change the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved. The division of the units in this application is a logical division. There may be other division methods when it is implemented in actual applications. For example, multiple units can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between each other shown or discussed can be through some interfaces, and the indirect coupling or communication connection between units can be electrical or other similar forms, which are not limited in this application. In addition, the unit or subunit described as a separate component may or may not be physically separated, may or may not be a physical unit, or may be distributed in multiple circuit units, and some or all of the units can be selected according to actual needs to achieve the purpose of the present application solution.

[0249] The above is only a specific implementation mode of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or replacements within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. The various digital numbers involved in the embodiments of the present application are only for the convenience of description and are not used to limit the scope of the embodiments of the present application. The size of the sequence number of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic.

[0250] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting a fault of a memory, characterized in that: include: Predicting, according to the first model, a first probability of a memory failure occurring within a first time period; Determine a result of the memory after the stress test is performed according to the first probability; Update the sample labels of the first model based on the result after the stress test; Iteratively training the first model based on the updated sample labels to obtain a trained first model; A fault prediction result of the memory is obtained based on the trained first model.

2. The method according to claim 1, characterized in that The determining, according to the first probability, a result of the stress test on the memory includes: determining a first area in the memory to be subjected to a stress test according to the first probability; Performing a pressure test on the first area to obtain a result after the pressure test.

3. The method according to claim 2, characterized in that The determining, according to the first probability, a first area in the memory to be subjected to a stress test comprises: When a first condition is met, determining a boundary sample according to a comparison result between the first probability and the first range; If an area belonging to the boundary sample is screened out from the area where correctable errors occur in the memory, a cluster analysis is performed on the area, and a first area to be stress-tested in the memory is determined based on the Laida criterion.

4. The method according to claim 3, characterized in that The first condition includes: The first probability is greater than or equal to a first threshold, and the memory is not subjected to stress testing.

5. The method according to claim 3 or 4, characterized in that: The step of determining a boundary sample according to a comparison result between the first probability and the first range includes: When the first probability belongs to the first range, the area in the memory corresponding to the first probability is determined as the boundary sample.

6. The method according to any one of claims 1 to 5, characterized in that The step of predicting a first probability of a memory failure within a first time period according to the first model includes: Acquiring error information of a correctable error occurring in the memory; The error information of the correctable error occurring in the memory is used as the input of the first model, and a first probability of the occurrence of an uncorrectable error in the memory within the first time period is output.

7. The method according to claim 6, characterized in that The updating of the sample labels of the first model based on the result after the stress test includes: When the result after the stress test does not include uncorrectable errors, the sample labels of the first model are updated based on error information of correctable errors occurring in the memory and the result after the stress test.

8. A memory failure prediction device, characterized in that: include: A model prediction module, configured to predict, based on a first model, a first probability of a memory failing within a first time period; A stress testing module, configured to determine a result of stress testing the memory according to the first probability; A label updating module, used for updating the sample labels of the first model based on the result after the stress test; A model training module, used for iteratively training the first model based on the updated sample labels to obtain a trained first model; The model prediction module is also used to obtain a fault prediction result of the memory based on the trained first model.

9. The device according to claim 8, characterized in that The pressure test module is also used for: determining a first area in the memory to be subjected to a stress test according to the first probability; Performing a pressure test on the first area to obtain a result after the pressure test.

10. The device according to claim 9, characterized in that The determining, according to the first probability, a first area in the memory to be subjected to a stress test comprises: When a first condition is met, determining a boundary sample according to a comparison result between the first probability and the first range; If an area belonging to the boundary sample is screened out from the area where correctable errors occur in the memory, a cluster analysis is performed on the area, and a first area to be stress-tested in the memory is determined based on the Laida criterion.

11. The device according to claim 10, characterized in that The first condition includes: The first probability is greater than a first threshold, and the memory is not stress tested.

12. The device according to claim 10 or 11, characterized in that The step of determining a boundary sample according to a comparison result between the first probability and the first range includes: When the first probability belongs to the first range, the area in the memory corresponding to the first probability is determined as the boundary sample.

13. The device according to any one of claims 8 to 12, characterized in that The model prediction module is also used to: Acquiring error information of a correctable error occurring in the memory; The error information of the correctable error occurring in the memory is used as the input of the first model, and a first probability of the occurrence of an uncorrectable error in the memory within the first time period is output.

14. The device according to claim 13, characterized in that The model prediction module is also used to: When the result after the stress test does not include uncorrectable errors, the sample labels of the first model are updated based on error information of correctable errors occurring in the memory and the result after the stress test.

15. A computer device, characterized in that: The computer device includes a memory and a processor; The memory is used to store computer programs; The processor is configured to execute the computer program to implement the method according to any one of claims 1 to 7.

16. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Citation Information

Cited By

  • Fault prediction method and apparatus, device, storage medium, and program product

    WO2026157267A1

  • Memory management method and related apparatus

    WO2026194688A1