CN-121996509-A - Memory diagnosis method, device, equipment and medium

Abstract

The invention discloses a memory diagnosis method, a device, equipment and a medium, which are applied to the technical field of servers and comprise the steps of acquiring a memory data sequence acquired according to a preset time interval in the latest preset time period, extracting memory data characteristics based on the memory data sequence, inputting the memory data characteristics into a target memory state prediction model to obtain a memory state prediction result of the future preset time period corresponding to the memory data sequence, wherein the target memory state prediction model is obtained by training a memory data sequence characteristic sample and a memory state label, and the memory state label represents a memory state of the future preset time period corresponding to the memory data sequence characteristic sample. Therefore, the memory fault can be predicted in real time, so that the problem can be positioned and solved, and the performance of the server is improved.

Inventors

• ZHANG XU
• YANG LEI
• WANG YOUFU

Assignees

• 浪潮计算机科技有限公司

Dates

Publication Date

20260508

Application Date

20260129

Claims (10)

1. 1. A memory diagnostic method, comprising: acquiring a memory data sequence acquired according to a preset time interval within the latest preset time; Extracting memory data characteristics based on the memory data sequence, wherein the memory data characteristics comprise change trend characteristics; Inputting the memory data characteristics into a target memory state prediction model to obtain a memory state prediction result of a future preset duration corresponding to the memory data sequence; The target memory state prediction model is obtained by training a memory data sequence feature sample and a memory state label, and the memory state label represents a memory state of a future preset duration corresponding to the memory data sequence feature sample.
2. 2. The memory diagnostic method of claim 1, further comprising: obtaining a training sample set, wherein the training sample set comprises a memory data sequence characteristic sample and a memory state label, and comprises a positive sample and a negative sample, wherein the positive sample is the memory data sequence characteristic sample with a fault in a memory in a future preset time length corresponding to the sample, the negative sample is the memory data sequence characteristic sample without the fault in the memory in the future preset time length corresponding to the sample, and the memory state label represents whether the memory has the fault in the memory in the future preset time length; and training the initial memory state prediction model by using the training sample set to obtain a target memory state prediction model.
3. 3. The memory diagnostic method of claim 1 wherein extracting memory data features based on the memory data sequence comprises: and extracting one or more types of characteristics of temperature characteristics, power consumption characteristics, error statistics characteristics and memory performance characteristics based on the memory data sequence, wherein at least one type of characteristics comprises a change trend characteristic.
4. 4. The memory diagnostic method of claim 3 wherein any of the temperature characteristics, the power consumption characteristics, the error statistics characteristics, and the memory performance characteristics further comprise mean and/or discrete level characteristics.
5. 5. The memory diagnostic method of claim 1, further comprising: and acquiring memory data according to a preset time interval by using a specific core in a management controller, wherein the management controller comprises a plurality of cores, one core in the plurality of cores is used as the specific core, and other cores are used for executing other functions.
6. 6. The memory diagnostic method of claim 5, wherein collecting memory data at predetermined time intervals using a specific core in the management controller comprises: and collecting memory data through the bus hub among the enhanced integrated circuits according to a preset time interval by utilizing a specific core in the management controller.
7. 7. The memory diagnostic method according to any one of claims 1 to 6, further comprising, after inputting the memory data characteristics into a target memory state prediction model to obtain a memory state prediction result of a future preset duration corresponding to the memory data sequence: matching a target processing strategy based on the memory state prediction result; And adjusting the memory parameters according to the target processing strategy.
8. 8. A memory diagnostic device, comprising: the data acquisition module is used for acquiring a memory data sequence acquired according to a preset time interval in the latest preset time length; The feature extraction module is used for extracting memory data features based on the memory data sequence, wherein the memory data features comprise variation trend features; The state prediction module is used for inputting the memory data characteristics into a target memory state prediction model to obtain a memory state prediction result of a future preset duration corresponding to the memory data sequence; The target memory state prediction model is obtained by training a memory data sequence feature sample and a memory state label, and the memory state label represents a memory state of a future preset duration corresponding to the memory data sequence feature sample.
9. 9. An electronic device, comprising: A memory for storing a computer program; A processor for executing the computer program to implement the steps of the memory diagnostic method as claimed in any one of claims 1 to 7.
10. 10. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program which, when executed by a processor, implements the steps of the memory diagnostic method according to any of claims 1 to 7.

Description

Memory diagnosis method, device, equipment and medium Technical Field The present invention relates to the field of server technologies, and in particular, to a memory diagnosis method, apparatus, device, and medium. Background The memory of the server is one of the core components of the server system, and the running efficiency and stability of the server are directly affected, so that the memory diagnosis scheme is necessary to discover the memory faults in time, and at present, the existing memory diagnosis scheme cannot timely perform fault early warning, and faults are checked again to influence the performance of the server. It can be seen how to predict memory failures in real time in order to locate and solve the problem, and improving server performance is a problem that needs to be solved by those skilled in the art. Disclosure of Invention The embodiment of the invention aims to provide a memory diagnosis method, a device, equipment and a medium, which can predict memory faults in real time so as to locate and solve problems and improve the performance of a server. The specific scheme is as follows: in a first aspect, the present invention provides a memory diagnosis method, including: acquiring a memory data sequence acquired according to a preset time interval within the latest preset time; Extracting memory data characteristics based on the memory data sequence, wherein the memory data characteristics comprise change trend characteristics; Inputting the memory data characteristics into a target memory state prediction model to obtain a memory state prediction result of a future preset duration corresponding to the memory data sequence; The target memory state prediction model is obtained by training a memory data sequence feature sample and a memory state label, and the memory state label represents a memory state of a future preset duration corresponding to the memory data sequence feature sample. Optionally, the method further comprises: obtaining a training sample set, wherein the training sample set comprises a memory data sequence characteristic sample and a memory state label, and comprises a positive sample and a negative sample, wherein the positive sample is the memory data sequence characteristic sample with a fault in a memory in a future preset time length corresponding to the sample, the negative sample is the memory data sequence characteristic sample without the fault in the memory in the future preset time length corresponding to the sample, and the memory state label represents whether the memory has the fault in the memory in the future preset time length; and training the initial memory state prediction model by using the training sample set to obtain a target memory state prediction model. Optionally, extracting the memory data feature based on the memory data sequence includes: and extracting one or more types of characteristics of temperature characteristics, power consumption characteristics, error statistics characteristics and memory performance characteristics based on the memory data sequence, wherein at least one type of characteristics comprises a change trend characteristic. Optionally, any one of the temperature feature, the power consumption feature, the error statistics feature, and the memory performance feature further includes a mean value and/or a discrete degree feature. Optionally, the method further comprises: and acquiring memory data according to a preset time interval by using a specific core in a management controller, wherein the management controller comprises a plurality of cores, one core in the plurality of cores is used as the specific core, and other cores are used for executing other functions. Optionally, the collecting the memory data by using a specific core in the management controller according to a preset time interval includes: and collecting memory data through the bus hub among the enhanced integrated circuits according to a preset time interval by utilizing a specific core in the management controller. Optionally, the method further comprises: matching a target processing strategy based on the memory state prediction result; And adjusting the memory parameters according to the target processing strategy. In a second aspect, the present invention provides a memory diagnostic device, including: the data acquisition module is used for acquiring a memory data sequence acquired according to a preset time interval in the latest preset time length; The feature extraction module is used for extracting memory data features based on the memory data sequence, wherein the memory data features comprise variation trend features; The state prediction module is used for inputting the memory data characteristics into a target memory state prediction model to obtain a memory state prediction result of a future preset duration corresponding to the memory data sequence; The target memory state prediction model is obtained by training a memory data seq