A memory failure prediction method, device, system and equipment based on Gaussian distribution

Abstract

The application belongs to the technical field of memory fault prediction, and specifically provides a memory fault prediction method, device, system and equipment based on Gaussian distribution, which comprises the following steps: collecting the time and number of correctable errors and uncorrectable errors of the memory at regular intervals; calculating the time variance of the correctable errors and the uncorrectable errors, and calculating the number variance of the correctable errors when the uncorrectable errors occur; when a new correctable error occurs, obtaining the prediction accuracy rate preset by the system; when the number distribution of the correctable errors is within the range corresponding to the prediction accuracy rate, triggering an alarm information and recording a log. The memory fault is accurately predicted through Gaussian distribution, necessary data support is provided for the safe operation of the server, the memory bar with a fast fault is replaced in time, and great loss is avoided.

Description

Technical Field

[0001] This invention relates to the field of memory fault prediction technology, and specifically to a method, apparatus, system, and device for memory fault prediction based on Gaussian distribution. Background Technology

[0002] With the rapid development of internet services, the availability of infrastructure has received increasing attention from the industry. However, hardware failure has always been a common phenomenon, and the losses caused by it are often enormous. Among the various components of a server, after hard drive failure, memory failure is the second most common type of hardware failure. Furthermore, servers have a large amount of memory, and the most serious consequence of memory failure is that it can directly lead to system crashes and server downtime, which is unacceptable for upper-layer applications.

[0003] Memory failures can be categorized into Uncorrectable Errors (UCE) and Correctable Errors (CE). When hardware detects an error, it reports it to the CPU in two ways. One way is through an interrupt. In this case, if it's a UCE (Uncorrectable Error), it may cause the server to crash immediately. If it's a CE (Correctable Error), the hardware will use some resources to repair the error. However, when too many CEs accumulate in memory and self-repair is no longer possible, a UCE will occur, causing the system to crash and restart. Therefore, we need to identify memory modules with excessive CEs as early as possible and replace them promptly to avoid significant losses.

[0004] In the past, memory failures were mostly discovered and located by combining the MCE (Machine Check Exception) log and the SEL (System Error Log) log recorded by the BMC. However, the biggest problem with this approach is that memory problems cannot be detected in advance and are often only discovered passively after the server crashes and restarts.

[0005] The normal distribution, also known as the Gaussian distribution, has a bell-shaped curve that is low at both ends and high in the middle, and is symmetrical. Because of its bell-shaped curve, it is often referred to as the bell curve. Summary of the Invention

[0006] In the past, memory faults were mostly detected and located by combining the MCE (Machine Check Exception) log and the SEL (System Error Log) log recorded by the BMC. However, the biggest problem with these methods is that they cannot detect memory problems in advance. They are often only discovered passively after the server crashes and restarts. This invention provides a method, device, system, and equipment for predicting memory faults based on Gaussian distribution.

[0007] In a first aspect, the present invention provides a memory fault prediction method based on Gaussian distribution, comprising the following steps:

[0008] Periodically collect the time and frequency of correctable and uncorrectable errors in memory;

[0009] Calculate the time variance of correctable and uncorrectable errors, and calculate the quantity variance of correctable errors when uncorrectable errors occur;

[0010] When a new correctable error is detected, the system's preset prediction accuracy is obtained;

[0011] When the number of correctable errors is determined to be within the range corresponding to the prediction accuracy, an alarm message is triggered and a log is recorded.

[0012] As a preferred embodiment of the technical solution of the present invention, in the steps of calculating the time variance of the occurrence of correctable and uncorrectable errors, and calculating the quantity variance of correctable errors when uncorrectable errors occur, the formula for calculating the time variance of the occurrence of correctable errors is as follows:

[0013] The time intervals when a correctable error occurs in a single memory module are x1, x2, x3, ..., x n The average time to correctable errors is x;

[0014] Correctable error occurrence time variance

[0015] The formula for calculating the variance of the number of correctable errors when uncorrectable errors occur is as follows:

[0016] When an uncorrectable error occurs in a single memory module, the number of correctable errors that have already occurred is y1, y2, y3, ..., y. n The average number of correctable errors that have occurred when an uncorrectable error occurs is y.

[0017] Calculate the variance of the number of correctable errors when uncorrectable errors occur:

[0018]

[0019] As a preferred embodiment of the technical solution of the present invention, the method further includes:

[0020] Set the prediction accuracy and assign different prediction accuracy ranges; including the following:

[0021] When the prediction accuracy reaches the first threshold, the number of correctable errors in a single memory root is within the range of (XA, X+A).

[0022] When the prediction accuracy reaches the second threshold, the number of correctable errors in a single memory segment is within the range of (X-2A, X+2A).

[0023] When the prediction accuracy reaches the third threshold, the number of correctable errors in a single memory segment is within the range of (X-3A, X+3A).

[0024] Among them, the variance of the number of correctable errors when uncorrectable errors occur is A, and the mean of the number of correctable errors when uncorrectable errors occur is X.

[0025] As a preferred embodiment of the technical solution of the present invention, the step of obtaining the system's preset prediction accuracy when a new correctable error is detected includes:

[0026] When a new correctable error is detected, calculate the current number of correctable errors, B.

[0027] When the system sets the prediction accuracy to the first threshold, it determines whether B is distributed within the range of (XA, X+A).

[0028] If so, the steps are as follows: trigger an alarm message and log it.

[0029] If not, if it is determined that B is distributed within the range of (XA-1, X+A+1), a warning message indicating that an uncorrectable error will occur within the set time range will be triggered.

[0030] As a preferred embodiment of the technical solution of the present invention, the set time range in the step of triggering the warning information that an uncorrectable error will occur within a set time range is (CDE, C+D+E).

[0031] C represents the time variance of correctable errors, D represents the average time of occurrence of correctable errors, and E represents the current time.

[0032] As a preferred embodiment of the technical solution of the present invention, the step of periodically collecting the time and number of occurrences of correctable and uncorrectable errors in memory includes:

[0033] The CPU's memory can be periodically read to identify correctable and uncorrectable errors, and the time and number of occurrences can be recorded; alternatively, the CPU's registers can be directly read via I2C to obtain the time and number of occurrences of correctable and uncorrectable memory errors.

[0034] As a preferred embodiment of the technical solution of the present invention, the step of periodically collecting the time and number of occurrences of correctable and uncorrectable errors in memory includes:

[0035] The BIOS collects information on the timing and frequency of correctable and uncorrectable memory errors.

[0036] Secondly, the present invention provides a memory fault prediction device based on Gaussian distribution, including a collection module, a calculation module, a monitoring and processing module, and an alarm triggering module;

[0037] The collection module is used to periodically collect the time and frequency of correctable and uncorrectable errors in memory.

[0038] The calculation module is used to calculate the time variance of the occurrence of correctable and uncorrectable errors, and to calculate the quantity variance of correctable errors when uncorrectable errors occur.

[0039] The monitoring and processing module is used to obtain the system's preset prediction accuracy when new correctable errors are detected.

[0040] The alarm triggering module is used to trigger alarm information and log when the number of correctable errors falls within the range corresponding to the prediction accuracy.

[0041] As a preferred embodiment of the technical solution of the present invention, in the steps of calculating the time variance of the occurrence of correctable and uncorrectable errors, and calculating the quantity variance of correctable errors when uncorrectable errors occur, the formula for calculating the time variance of the occurrence of correctable errors is as follows:

[0042] The time intervals when a correctable error occurs in a single memory module are x1, x2, x3, ..., x n The average time to correctable errors is x;

[0043] Correctable time variance of error occurrence

[0044] The formula for calculating the variance of the number of correctable errors when uncorrectable errors occur is as follows:

[0045] When an uncorrectable error occurs in a single memory module, the number of correctable errors that have already occurred is y1, y2, y3, ..., y. n The average number of correctable errors that have occurred when an uncorrectable error occurs is y.

[0046] Calculate the variance of the number of correctable errors when uncorrectable errors occur:

[0047]

[0048] As a preferred embodiment of the technical solution of the present invention, the device further includes a preset value module, used to set the prediction accuracy and to set different prediction accuracies corresponding to different ranges; the specific settings are as follows:

[0049] When the prediction accuracy reaches the first threshold, the number of correctable errors in a single memory root is within the range of (XA, X+A).

[0050] When the prediction accuracy reaches the second threshold, the number of correctable errors in a single memory segment is within the range of (X-2A, X+2A).

[0051] When the prediction accuracy reaches the third threshold, the number of correctable errors in a single memory segment is within the range of (X-3A, X+3A).

[0052] Among them, the variance of the number of correctable errors when uncorrectable errors occur is A, and the mean of the number of correctable errors when uncorrectable errors occur is X.

[0053] As a preferred embodiment of the technical solution of the present invention, the monitoring and processing module is specifically used to calculate the number of correctable errors B when a new correctable error is detected; when the system sets the prediction accuracy rate to a first threshold, determine whether B is distributed within the range of (XA, X+A); activate the alarm triggering module to trigger alarm information and record logs or trigger early warning information that an uncorrectable error will occur within a set time range.

[0054] As a preferred embodiment of the technical solution of the present invention, the set time range in the step of triggering the warning information that an uncorrectable error will occur within a set time range is (CDE, C+D+E).

[0055] C represents the time variance of correctable errors, D represents the average time of occurrence of correctable errors, and E represents the current time.

[0056] As a preferred embodiment of the technical solution of the present invention, the step of periodically collecting the time and number of occurrences of correctable and uncorrectable errors in memory includes:

[0057] The collection module is specifically used to periodically read the CPU's correctable and uncorrectable memory errors and record the time and number of occurrences; or, it can directly read the CPU's registers via I2C to obtain the time and number of occurrences of correctable and uncorrectable memory errors.

[0058] As a preferred embodiment of the technical solution of the present invention, the step of periodically collecting the time and number of occurrences of correctable and uncorrectable errors in memory includes:

[0059] The collection module is also used to collect the time and number of correctable and uncorrectable memory errors occurring via the BIOS.

[0060] Thirdly, the technical solution of the present invention also provides a memory fault prediction system based on Gaussian distribution, including a master system and a slave system;

[0061] The system periodically reads the CPU's correctable and uncorrectable memory errors and records the time and number of occurrences; or, it directly reads the CPU's registers via I2C to obtain the time and number of occurrences of correctable and uncorrectable memory errors; or, it collects the time and number of occurrences of correctable and uncorrectable memory errors through the BIOS.

[0062] The system will collect and aggregate the data and send it to the main system;

[0063] The master system calculates the time variance of correctable and uncorrectable errors respectively, and calculates the quantity variance of correctable errors when uncorrectable errors occur and sends it to the slave system.

[0064] When the system monitors a new correctable error, it obtains the system's preset prediction accuracy rate; when it determines that the number of correctable errors is within the range corresponding to the prediction accuracy rate, it triggers an alarm message and records it in the log.

[0065] Fourthly, the present invention also provides an electronic device, the electronic device comprising:

[0066] At least one processor; and,

[0067] A memory communicatively connected to the at least one processor; wherein,

[0068] The memory stores computer program instructions that can be executed by at least one processor to enable the at least one processor to perform the Gaussian distribution-based memory fault prediction method as described in the first aspect.

[0069] As can be seen from the above technical solutions, the present invention has the following advantages: it accurately predicts memory failures through Gaussian distribution, provides necessary data support for the safe operation of the server, and promptly detects and replaces faulty memory modules to avoid significant losses.

[0070] Furthermore, the design principle of this invention is reliable, the structure is simple, and it has a very wide range of application prospects.

[0071] Therefore, it is evident that the present invention has outstanding substantive features and significant progress compared with the prior art, and the beneficial effects of its implementation are also obvious. Attached Figure Description

[0072] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0073] Figure 1 This is a schematic flowchart of a method according to an embodiment of the present invention.

[0074] Figure 2 This is a schematic flowchart of a method according to another embodiment of the present invention.

[0075] Figure 3 This is a schematic block diagram of an apparatus according to an embodiment of the present invention. Detailed Implementation

[0076] Memory failures can be categorized into Uncorrectable Errors (UCE) and Correctable Errors (CE). When hardware detects an error, it reports it to the CPU in two ways. One way is through an interrupt. In the case of a UCE (Uncorrectable Error), this could cause the server to crash immediately. If it's a CE (Correctable Error), the hardware will use some resources to repair the error. However, when too many CEs accumulate in memory and self-repair is impossible, a UCE will occur, causing the system to crash and restart. Therefore, we need to detect memory modules with excessive CEs as early as possible and replace them promptly to avoid significant losses. To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0077] like Figure 1 As shown, this embodiment of the invention provides a memory fault prediction method based on Gaussian distribution, including the following steps:

[0078] Step 1: Periodically collect the time and frequency of correctable and uncorrectable errors in memory;

[0079] Step 2: Calculate the time variance of correctable and uncorrectable errors respectively, and calculate the quantity variance of correctable errors when uncorrectable errors occur;

[0080] Step 3: When a new correctable error is detected, obtain the system's preset prediction accuracy.

[0081] Step 4: When the number of correctable errors falls within the range corresponding to the prediction accuracy, trigger an alarm message and log it.

[0082] In this embodiment of the invention, the formula for calculating the time variance of correctable errors is as follows:

[0083] The time intervals when a correctable error occurs in a single memory module are x1, x2, x3, ..., x n The average time to correctable errors is x;

[0084] Correctable time variance of error occurrence

[0085] The formula for calculating the variance of the number of correctable errors when uncorrectable errors occur is as follows:

[0086] When an uncorrectable error occurs in a single memory module, the number of correctable errors that have already occurred is y1, y2, y3, ..., y. n The average number of correctable errors that have occurred when an uncorrectable error occurs is y.

[0087] Calculate the variance of the number of correctable errors when uncorrectable errors occur:

[0088]

[0089] It should be noted here that the steps of this invention require setting a prediction accuracy rate and setting different prediction accuracies to correspond to different ranges; including the following:

[0090] When the prediction accuracy reaches the first threshold, the number of correctable errors in a single memory root is within the range of (XA, X+A).

[0091] When the prediction accuracy reaches the second threshold, the number of correctable errors in a single memory segment is within the range of (X-2A, X+2A).

[0092] When the prediction accuracy reaches the third threshold, the number of correctable errors in a single memory segment is within the range of (X-3A, X+3A).

[0093] Among them, the variance of the number of correctable errors when uncorrectable errors occur is A, and the mean of the number of correctable errors when uncorrectable errors occur is X.

[0094] When a new correctable error is detected, the steps to obtain the system's preset prediction accuracy include:

[0095] When a new correctable error is detected, calculate the current number of correctable errors, B.

[0096] When the system sets the prediction accuracy to the first threshold, it determines whether B is distributed within the range of (XA, X+A).

[0097] If so, the steps are as follows: trigger an alarm message and log it.

[0098] If not, if the distribution of B falls within the range (XA-1, X+A+1), trigger a warning message indicating that an uncorrectable error is about to occur within a set time range. The set time range is (CDE, C+D+E); C is the time variance of correctable errors, D is the average time of occurrence of correctable errors, and E is the current time.

[0099] In this embodiment of the invention, the step of periodically collecting the time and number of correctable and uncorrectable errors occurring in memory includes:

[0100] The CPU's memory can be periodically read for correctable and uncorrectable errors, and the time and number of occurrences can be recorded; or, the CPU's registers can be directly read via I2C to obtain the time and number of occurrences of correctable and uncorrectable memory errors; or the BIOS can be used to collect the time and number of occurrences of correctable and uncorrectable memory errors.

[0101] like Figure 2 As shown, this embodiment of the invention provides a memory fault prediction method based on Gaussian distribution, including the following steps:

[0102] S101: A system with a master-slave structure, where the master system is responsible for memory prediction of the entire data center and the slave system is responsible for memory prediction of a single server.

[0103] S102: Periodically read the CPU's memory CE and UCE from the system and record the time and number of occurrences;

[0104] It should be noted that in this step, for Phytium platform servers, the CE and UCE values ​​are obtained by directly reading the CPU registers via I2C; for Hygon servers, the values ​​are collected through the BIOS and then sent to the slave system.

[0105] S103: The collected data from the system is aggregated and sent to the main system;

[0106] S104: The master system calculates the time variance of CE and UCE occurrences, and calculates the quantity variance of CE when UCE occurs, and sends it to the slave system;

[0107] The formula for calculating the time variance of CE occurrence is as follows:

[0108] The times when a single memory block experiences a Completely Executable (CE) are x1, x2, x3, ..., x. n The average number of times CE occurs is x;

[0109] Time variance of CE occurrence This is used to measure the magnitude of the fluctuation in this set of data, and it is called the time variance of CE.

[0110] The formula for calculating the variance of the number of CEs when a UCE occurs is as follows:

[0111] When a single memory segment experiences a UCE, the number of CEs that have already occurred is y1, y2, y3, ..., y. n The average number of CEs that have occurred when a UCE occurs is y;

[0112] Calculate the variance of the number of CEs when a UCE occurs:

[0113]

[0114] S105: From system monitoring to the appearance of a new CE, if the number of CEs is within the range corresponding to the prediction accuracy, it is determined that the memory needs to be replaced, and the information is recorded and an alert is issued.

[0115] In step S105, assuming the variance of the number calculated in S104 is A and the average number is X, when the system is set to a prediction accuracy of 68.26%, the number of CEs occurring in a single memory is B. When B is distributed in the range of (XA, X+A), it indicates that an alarm needs to be triggered. Similarly, when the prediction accuracy is 95.45%, the distribution range of B is (X-2A, X+2A); when the prediction accuracy is 99.73%, the distribution range of B is (X-3A, X+3A).

[0116] In step S105, the CE value is monitored from the system. When the CE value is equal to the UCE and the CE variance minus 1, it is predicted that a UCE will occur after a certain period of time with remaining CE variance. Therefore, the memory needs to be replaced, and this is recorded and an alert is issued. For example, if the system is set to a prediction accuracy of 68.26%, and the detected CE value of a single memory module is not in the range (XA, X+A) but in the range (XA-1, X+A+1), a UCE is expected to occur soon, and an alert needs to be issued. Assuming that the time variance of CE is calculated as C in S104, the average time of CE occurrence is D, the current time is E, and a UCE is expected to occur within the time range (CDE, C+D+E).

[0117] like Figure 3As shown, this embodiment of the invention provides a memory fault prediction device based on Gaussian distribution, including a collection module, a calculation module, a monitoring and processing module, and an alarm triggering module;

[0118] The collection module is used to periodically collect the time and frequency of correctable and uncorrectable errors in memory.

[0119] The calculation module is used to calculate the time variance of the occurrence of correctable and uncorrectable errors, and to calculate the quantity variance of correctable errors when uncorrectable errors occur.

[0120] The monitoring and processing module is used to obtain the system's preset prediction accuracy when new correctable errors are detected.

[0121] The alarm triggering module is used to trigger alarm information and log when the number of correctable errors falls within the range corresponding to the prediction accuracy.

[0122] It should be noted that in the steps of calculating the time variance of correctable and uncorrectable errors separately, and calculating the quantity variance of correctable errors when uncorrectable errors occur, the formula for calculating the time variance of correctable errors is as follows:

[0123] The time intervals when a correctable error occurs in a single memory module are x1, x2, x3, ..., x n The average time to correctable errors is x;

[0124] Correctable time variance of error occurrence

[0125] The formula for calculating the variance of the number of correctable errors when uncorrectable errors occur is as follows:

[0126] When an uncorrectable error occurs in a single memory module, the number of correctable errors that have already occurred is y1, y2, y3, ..., y. n The average number of correctable errors that have occurred when an uncorrectable error occurs is y.

[0127] Calculate the variance of the number of correctable errors when uncorrectable errors occur:

[0128]

[0129] It should be noted that the device also includes a preset value module, used to set the prediction accuracy and to set different ranges corresponding to different prediction accuracies; the specific settings are as follows:

[0130] When the prediction accuracy reaches the first threshold, the number of correctable errors in a single memory root is within the range of (XA, X+A).

[0131] When the prediction accuracy reaches the second threshold, the number of correctable errors in a single memory segment is within the range of (X-2A, X+2A).

[0132] When the prediction accuracy reaches the third threshold, the number of correctable errors in a single memory segment is within the range of (X-3A, X+3A).

[0133] Among them, the variance of the number of correctable errors when uncorrectable errors occur is A, and the mean of the number of correctable errors when uncorrectable errors occur is X.

[0134] It should be noted that the monitoring and processing module is specifically used to calculate the number B of correctable errors when a new correctable error is detected; when the system sets the prediction accuracy rate to the first threshold, it determines whether B is distributed within the range of (XA, X+A); it activates the alarm trigger module to trigger alarm information and record logs, or triggers a warning information for uncorrectable errors that will occur within a set time range; the set time range is (CDE, C+D+E); C is the time variance of correctable errors, D is the average time of occurrence of correctable errors, and E is the current time.

[0135] It should be noted that the steps for periodically collecting the time and frequency of correctable and uncorrectable errors in memory include:

[0136] The collection module is specifically used to periodically read the CPU's correctable and uncorrectable memory errors and record the time and number of occurrences; or, it can directly read the CPU's registers via I2C to obtain the time and number of occurrences of correctable and uncorrectable memory errors; or it can collect the time and number of occurrences of correctable and uncorrectable memory errors through the BIOS.

[0137] This invention also provides a Gaussian distribution-based memory fault prediction system, including a master system and a slave system;

[0138] The system periodically reads the CPU's correctable and uncorrectable memory errors and records the time and number of occurrences; or, it directly reads the CPU's registers via I2C to obtain the time and number of occurrences of correctable and uncorrectable memory errors; or, it collects the time and number of occurrences of correctable and uncorrectable memory errors through the BIOS.

[0139] The system will collect and aggregate the data and send it to the main system;

[0140] The master system calculates the time variance of correctable and uncorrectable errors respectively, and calculates the quantity variance of correctable errors when uncorrectable errors occur and sends it to the slave system.

[0141] When the system monitors a new correctable error, it obtains the system's preset prediction accuracy rate; when it determines that the number of correctable errors is within the range corresponding to the prediction accuracy rate, it triggers an alarm message and records it in the log.

[0142] Specific prediction methods include:

[0143] S11: Periodically read the CPU's memory CE and UCE from the system and record the time and number of occurrences;

[0144] It should be noted that in this step, for Phytium platform servers, the CE and UCE values ​​are obtained by directly reading the CPU registers via I2C; for Hygon servers, the values ​​are collected through the BIOS and then sent to the slave system.

[0145] S12: The system will collect and summarize the data and send it to the main system;

[0146] S13: The master system calculates the time variance of CE and UCE occurrences, and calculates the quantity variance of CE when UCE occurs, and sends it to the slave system;

[0147] The formula for calculating the time variance of CE occurrence is as follows:

[0148] The times when a single memory block experiences a Completely Executable (CE) are x1, x2, x3, ..., x. n The average number of times CE occurs is x;

[0149] Time variance of CE occurrence This is used to measure the magnitude of the fluctuation in this set of data, and it is called the time variance of CE.

[0150] The formula for calculating the variance of the number of CEs when a UCE occurs is as follows:

[0151] When a single memory segment experiences a UCE, the number of CEs that have already occurred is y1, y2, y3, ..., y. n The average number of CEs that have occurred when a UCE occurs is y;

[0152] Calculate the variance of the number of CEs when a UCE occurs:

[0153]

[0154] S14: From system monitoring to the appearance of a new CE, determine that the number of CEs is within the range corresponding to the prediction accuracy, determine that the memory needs to be replaced, and record and issue an alert;

[0155] In step S14, assuming the variance of the number calculated in S13 is A and the average number is X, when the system is set to a prediction accuracy of 68.26%, the number of CEs occurring in a single memory is B. When B is distributed in the range of (XA, X+A), it indicates that an alarm needs to be triggered. Similarly, when the prediction accuracy is 95.45%, the distribution range of B is (X-2A, X+2A); when the prediction accuracy is 99.73%, the distribution range of B is (X-3A, X+3A).

[0156] In step S14, the CE value is monitored from the system. When the CE value is equal to the UCE and the CE variance minus 1, it is predicted that a UCE will occur after a certain period of time with remaining CE variance. Therefore, the memory needs to be replaced, and this is recorded and an alert is issued. For example, if the system is set to a prediction accuracy of 68.26%, and the CE value of a single memory module is detected to be outside the range (XA, X+A) but within the range (XA-1, X+A+1), a UCE is expected to occur soon, and an alert needs to be issued. Assuming that the time variance of CE is calculated as C in S13, the average time of CE occurrence is D, the current time is E, and a UCE is expected to occur within the time range (CDE, C+D+E).

[0157] This invention also provides an electronic device, comprising: a processor, a communication interface, a memory, and a bus, wherein the processor, communication interface, and memory communicate with each other via the bus. The bus can be used for information transmission between the electronic device and sensors. The processor can call logical instructions in the memory to execute the following method: Step 1: periodically collect the time and number of occurrences of correctable and uncorrectable errors in the memory; Step 2: calculate the time variance of the occurrences of correctable and uncorrectable errors respectively, and calculate the quantity variance of correctable errors when uncorrectable errors occur; Step 3: when a new correctable error is detected, obtain the system's preset prediction accuracy; Step 4: when the number of correctable errors is distributed within the range corresponding to the prediction accuracy, trigger an alarm message and record a log.

[0158] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0159] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the invention by those skilled in the art without departing from the spirit and essence of the invention, and such modifications or substitutions should all be within the scope of the invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the invention should also be covered within the protection scope of the invention. Therefore, the protection scope of the invention should be determined by the scope of the claims.

Claims

1. A memory fault prediction method based on Gaussian distribution, characterized in that, Includes the following steps: Periodically collect the time and frequency of correctable and uncorrectable errors in memory; Calculate the time variance of correctable and uncorrectable errors, and calculate the quantity variance of correctable errors when uncorrectable errors occur; When a new correctable error is detected, the system's preset prediction accuracy is obtained. Specifically, this includes: when a new correctable error is detected, the number of correctable errors B is calculated; when the system's prediction accuracy is set to a first threshold, it is determined whether B is distributed within the range (XA, X+A); if yes, the following steps are executed: triggering an alarm message and recording a log; if no, when it is determined that B is distributed within the range (XA-1, X+A+1), an early warning message indicating that an uncorrectable error will occur within a set time range is triggered, where the variance of the number of correctable errors when an uncorrectable error occurs is A, and the average number of correctable errors when an uncorrectable error occurs is X. When the number of correctable errors is determined to be within the range corresponding to the prediction accuracy, an alarm message is triggered and a log is recorded.

2. The memory fault prediction method based on Gaussian distribution according to claim 1, characterized in that, In the steps of calculating the time variance of correctable and uncorrectable errors, and calculating the quantity variance of correctable errors when uncorrectable errors occur, the formula for calculating the time variance of correctable errors is as follows: The time it takes for a single memory module to develop a correctable error is The average time to correctable errors is ; Correctable time variance of error occurrence ; The formula for calculating the variance of the number of correctable errors when uncorrectable errors occur is as follows: When a single memory module encounters an uncorrectable error, the number of correctable errors that have already occurred is: The average number of correctable errors that occur before an uncorrectable error occurs is ; Calculate the variance of the number of correctable errors when uncorrectable errors occur: 。 3. The memory fault prediction method based on Gaussian distribution according to claim 2, characterized in that, The method also includes: Set the prediction accuracy and assign different prediction accuracy ranges; including the following: When the prediction accuracy reaches the first threshold, the number of correctable errors in a single memory root is within the range of (XA, X+A). When the prediction accuracy reaches the second threshold, the number of correctable errors in a single memory segment is within the range of (X-2A, X+2A). When the prediction accuracy reaches the third threshold, the number of correctable errors in a single memory segment is within the range of (X-3A, X+3A).

4. The memory fault prediction method based on Gaussian distribution according to claim 3, characterized in that, The time range for triggering the warning information that an uncorrectable error is about to occur within a set time range is (CDE, C+D+E). C represents the time variance of correctable errors, D represents the average time of occurrence of correctable errors, and E represents the current time.

5. The memory fault prediction method based on Gaussian distribution according to claim 4, characterized in that, The steps for periodically collecting the time and frequency of correctable and uncorrectable errors in memory include: The CPU's memory can be periodically read to identify correctable and uncorrectable errors, and the time and number of occurrences can be recorded; alternatively, the CPU's registers can be directly read via I2C to obtain the time and number of occurrences of correctable and uncorrectable memory errors.

6. The Gaussian distribution-based memory fault prediction method according to claim 4, characterized in that, The steps for periodically collecting the time and frequency of correctable and uncorrectable errors in memory include: The BIOS collects information on the timing and frequency of correctable and uncorrectable memory errors.

7. A memory fault prediction device based on Gaussian distribution, characterized in that, It includes a data collection module, a calculation module, a monitoring and processing module, and an alarm triggering module; The collection module is used to periodically collect the time and frequency of correctable and uncorrectable errors in memory. The calculation module is used to calculate the time variance of the occurrence of correctable and uncorrectable errors, and to calculate the quantity variance of correctable errors when uncorrectable errors occur. The monitoring and processing module is used to obtain the system's preset prediction accuracy when new correctable errors are detected. Specifically, when a new correctable error is detected, the system calculates the number of correctable errors B. If the system sets the prediction accuracy to a first threshold, it determines whether B falls within the range (XA, X+A). If yes, it executes the following steps: triggering an alarm and logging the information. If no, it determines that B falls within the range (XA-1, X+A+1), triggering a warning message indicating that an uncorrectable error will occur within a set time frame. Here, the variance of the number of correctable errors when an uncorrectable error occurs is A, and the average number of correctable errors when an uncorrectable error occurs is X. The alarm triggering module is used to trigger alarm information and log when the number of correctable errors falls within the range corresponding to the prediction accuracy.

8. A memory fault prediction system based on Gaussian distribution, characterized in that, Includes master and slave systems; The system periodically reads correctable and uncorrectable errors from the CPU's memory and records the time and number of occurrences; or, it directly reads the CPU's registers via I2C to obtain the time and number of occurrences of correctable and uncorrectable errors in memory. Alternatively, the BIOS can collect data on the timing and frequency of correctable and uncorrectable memory errors. The system will collect and aggregate the data and send it to the main system; The master system calculates the time variance of correctable and uncorrectable errors respectively, and calculates the quantity variance of correctable errors when uncorrectable errors occur and sends it to the slave system. When a new correctable error is detected by the system, the system's preset prediction accuracy is obtained. When the number of correctable errors is determined to be within the range corresponding to the prediction accuracy, an alarm is triggered and a log is recorded. The steps for obtaining the system's preset prediction accuracy when a new correctable error is detected include: when a new correctable error is detected, the number of correctable errors B is calculated; when the system sets the prediction accuracy to a first threshold, it is determined whether B is distributed within the range (XA, X+A); if yes, the steps are executed: an alarm is triggered and a log is recorded; if no, when it is determined that B is distributed within the range (XA-1, X+A+1), an early warning message is triggered indicating that an uncorrectable error will occur within a set time range. The variance of the number of correctable errors when an uncorrectable error occurs is A, and the average number of correctable errors when an uncorrectable error occurs is X.

9. An electronic device, characterized in that, The electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores computer program instructions executable by the at least one processor, the computer program instructions being executed by the at least one processor to enable the at least one processor to perform the Gaussian distribution-based memory fault prediction method as described in any one of claims 1 to 6.

Images