IDC machine room operation and maintenance system and method based on visual inspection

Abstract

The invention discloses an IDC machine room operation and maintenance system and method based on visual inspection, and belongs to the field of machine room equipment operation and maintenance, and the system comprises a data collection module which is used for collecting data from an IDC moving ring, a physical machine out-of-band management system, a virtual machine cloud platform, a switch and a firewall monitoring system; the image processing module is used for generating an image from the collected data and performing information extraction by using an OCR (Optical Character Recognition) technology; the AI reasoning module is used for loading a pre-trained large model and analyzing and reasoning the extracted image information; the model training module is used for marking the inference result and the image and taking the inference result and the image as a new training sample, and is used for finely adjusting an original model; and the visual presentation module is used for displaying the reasoning result and the operation and maintenance suggestions in a UI form. According to the method, the AI vision technology is combined with large model reasoning, so that the operation and maintenance efficiency and the intelligent level of the IDC machine room are improved.

Description

Technical Field

[0001] The present invention relates to the field of operation and maintenance of computer room equipment, and more specifically, to an IDC computer room operation and maintenance system and method based on visual detection. Background Art

[0002] With the rapid development of Internet technology, as the core facility for data storage and processing, the operation and maintenance management of IDC computer rooms has become particularly important. However, traditional operation and maintenance methods mainly rely on manual inspections and single-device monitoring systems, which have problems such as heavy workload, low efficiency, and insufficient accuracy. In recent years, with the development of artificial intelligence technology, visual detection technology has shown great potential in the field of automated monitoring, but its application in IDC computer room operation and maintenance is not yet mature.

[0003] There are the following technical problems in the application of existing artificial intelligence visual detection technology in IDC computer room operation and maintenance:

[0004] 1) Accuracy and reliability: Existing visual detection systems may have errors in identifying and classifying objects in images, especially in complex or ever-changing environments. Therefore, improving the accuracy and reliability of the system is the key.

[0005] 2) Real-time performance: IDC computer rooms require real-time monitoring to quickly respond to any potential problems. Existing visual detection systems may not be able to meet the requirements of real-time monitoring in terms of processing speed.

[0006] 3) Environmental adaptability: The environment of IDC computer rooms may affect the performance of visual detection systems, such as light changes, equipment occlusion, etc. The system needs to be able to adapt to these environmental changes.

[0007] 4) Data processing ability: With the increase in data volume, existing visual detection systems may encounter bottlenecks in processing a large amount of image data and require more efficient data processing and analysis methods.

[0008] 5) Integration and compatibility: Existing visual detection systems may need to be integrated with existing monitoring systems, which may encounter compatibility problems and require the development of solutions that can be seamlessly integrated.

[0009] 6) Cost-effectiveness: Deploying and maintaining visual detection systems may require relatively high costs, and it is necessary to evaluate their cost-effectiveness to ensure the rationality of investment.

[0010] Therefore, designing an IDC computer room operation and maintenance system based on visual detection has important research value and application prospects. Summary of the Invention

[0011] The purpose of the present invention is to overcome the deficiencies of the prior art and provide an IDC computer room operation and maintenance system and method based on visual detection. By combining AI visual technology with large model reasoning, the operation and maintenance efficiency and intelligent level of the IDC computer room are improved.

[0012] The purpose of the present invention is achieved through the following solutions:

[0013] An IDC computer room operation and maintenance system based on visual detection, comprising:

[0014] A data acquisition module, used to collect data from the IDC power environment, physical machine out-of-band management system, virtual machine cloud platform, and switch and firewall monitoring system;

[0015] An image processing module, used to generate images from the collected data and perform information extraction using OCR technology;

[0016] An AI inference module, used to load a pre-trained large model and analyze and infer the extracted image information;

[0017] A model training module, used to label the inference results and images and use them as new training samples to fine-tune the original model;

[0018] A visualization presentation module, used to display the inference results and operation and maintenance suggestions in the form of a UI.

[0019] Further, in the data acquisition module, a multi-threaded mechanism is specifically used to collect data.

[0020] Further, the multi-threaded mechanism for collecting data includes:

[0021] SNMP polling: Use the SNMP protocol to regularly obtain the operating status of devices;

[0022] NetFlow analysis: Obtain network traffic data through the NetFlow protocol and analyze network traffic patterns;

[0023] API call: Obtain real-time data through the RESTful API of the device.

[0024] Further, in the image processing module, the generation of images from the collected data specifically includes the following sub-steps:

[0025] Convert the color image to a grayscale image through grayscale processing, and then convert the grayscale image to a black and white image through binarization processing and perform denoising processing.

[0026] Further, in the image processing module, the information extraction using OCR technology specifically includes the following sub-steps:

[0027] The text content is identified using the moe model, and the content of the abnormal information in the text is understood.

[0028] Further, in the AI inference module, the pre-trained large model is loaded to analyze and infer the extracted image information, specifically including:

[0029] After the user input on the page, the background uses the moe model to route to experts in relevant fields, and the experts then perform inferences through the model.

[0030] Further, in the model training module, the fine-tuning is performed, specifically including the following sub-steps:

[0031] Distributed training fine-tuning, where the fine-tuning content corresponds to each computing power card, and relevant knowledge is merged into the moe model in sequence according to the load balancing algorithm.

[0032] Further, in the visualization module, the inference results and operation and maintenance suggestions are displayed in the form of UI, specifically including the following sub-steps: Output based on the moe model and send text messages.

[0033] An implementation method of an IDC computer room operation and maintenance system based on visual detection, based on the IDC computer room operation and maintenance system based on visual detection as described in any one of the above, includes the following steps:

[0034] First, perform hardware deployment, install visual detection devices in the IDC computer room, and connect them to various management systems;

[0035] In terms of software configuration, configure the data acquisition module and the image processing module to ensure the acquisition and conversion of real-time data, and then load the pre-trained large model and set the update mechanism.

[0036] An operation and maintenance method of an IDC computer room operation and maintenance system based on visual detection, based on the IDC computer room operation and maintenance system based on visual detection as described in any one of the above, includes the following steps:

[0037] Step 1, monitor the device status in real time, and discover potential risks through visual detection;

[0038] Step 2, combine the inference results of the large model, formulate an operation and maintenance plan and present it through the UI;

[0039] Step 3, regularly fine-tune the model to adapt to changes in equipment and environment

[0040] The beneficial effects of the present invention include:

[0041] The solution of the present invention uses AI visual detection technology to greatly reduce the workload of manual inspections.

[0042] The solution of the present invention realizes intelligence. Through large model reasoning and continuous optimization of training, the accuracy and comprehensiveness of problem diagnosis are improved.

[0043] The solution of the present invention has the advantage of scalability, can adapt to different types of IDC computer rooms and devices, and realizes customized operation and maintenance solutions.

[0044] The solution of the present invention breaks through the limitations of the traditional operation and maintenance mode. Through the combination of software and hardware and the application of AI technology, the high efficiency and intelligence of IDC computer room operation and maintenance are realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0046] Figure 1 It is the architecture diagram of the system according to the embodiment of the present invention;

[0047] Figure 2 It is the step flow chart of the method according to the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] All the features disclosed in all the embodiments in this specification, or all the steps in the methods or processes implicitly disclosed, except for the mutually exclusive features and / or steps, can be combined and / or extended, replaced in any way.

[0049] In a preferred embodiment, as Figure 1 and Figure 2 shown, the present invention specifically provides a system and its corresponding method for IDC (Internet Data Center) computer room operation and maintenance using artificial intelligence vision detection technology.

[0050] First, as the first aspect of the present invention, in the system embodiment, it includes the following modules:

[0051] A data acquisition module for collecting data from the IDC power environment, physical machine out-of-band management system, virtual machine cloud platform, and switch and firewall monitoring systems.

[0052] An image processing module for generating images from the collected data and using OCR (Optical Character Recognition) technology for information extraction.

[0053] An AI reasoning module for loading a pre-trained large model and analyzing and reasoning the extracted image information.

[0054] The model training module is used to label the inference results and images and use them as new training samples for fine-tuning the original model.

[0055] The visualization and presentation module is used to display the inference results and operation and maintenance suggestions in the form of a UI.

[0056] In other embodiments based on the above system embodiment, more specifically, data is collected once every 1 minute in the data collection module, and a multi-threaded mechanism is adopted to obtain data in the following ways.

[0057] SNMP polling: Use the SNMP protocol to regularly obtain the operating status of devices (such as CPU usage, memory usage, port status, etc.).

[0058] NetFlow analysis: Obtain network traffic data through the NetFlow protocol and analyze network traffic patterns.

[0059] API call: Obtain real-time data through the RESTful API of the device.

[0060] In the image processing module, the following specific processing is carried out:

[0061] Grayscale conversion: Convert the color image into a grayscale image to reduce the computational complexity;

[0062] Binarization: Convert the grayscale image into a black and white image to facilitate the separation of text from the background; Specifically, an adaptive threshold method can be used for image segmentation.

[0063] Denoising: Remove the noise in the image (such as salt and pepper noise, Gaussian noise). Specifically, a Gaussian filtering method can be used.

[0064] Skew correction: Detect and correct the text skew in the image. Specifically, the Hough transform method can be used.

[0065] In the image OCR text recognition, specifically use the moe model to recognize the text content and understand the content of the abnormal information in the text. It should be noted that the recognition effect depends to a large extent on the effect fine-tuned by model training.

[0066] In the AI inference module, through the input of the page user, the background moe model is routed to experts in related fields, and these experts then perform inferences through the model.

[0067] In the model training module, distributed training fine-tuning is specifically carried out. The fine-tuning content corresponds to each computing power card, and relevant knowledge is merged into the moe model in turn according to the load balancing algorithm. Further, the fine-tuning first adopts sft + lora, and then uses the clip discriminator. For example, the similarity between the content fine-tuned by sft and the content inferred by the discriminator is calculated, and the result is corrected to output the fine-tuned weight and parameter information. The advantage is that it can reduce memory usage and has strong performance.

[0068] In the visualization module, the user inputs, and the moe model outputs and sends text messages to output the refined results.

[0069] As the second aspect of the present invention, in the method embodiment, the following steps are executed: First, perform hardware deployment, install vision detection equipment in the IDC computer room, and dock with various management systems. In terms of software configuration, configure the data acquisition module and the image processing module to ensure the acquisition and conversion of real-time data. Load the pre-trained large model and set the automatic update mechanism.

[0070] In other method embodiments of the present invention, the following operation and maintenance process is specifically executed:

[0071] Step 1, monitor the device status in real time and discover potential risks through vision detection;

[0072] Step 2, combine the inference results of the large model, formulate an operation and maintenance plan and present it through the UI;

[0073] Step 3, regularly fine-tune the model to adapt to device and environmental changes.

[0074] It should be noted that within the protection scope defined in the claims of the present invention, the following embodiments can be combined and / or extended, replaced in any logical manner from the above specific implementation manners, such as the disclosed technical principles, disclosed technical features or implicitly disclosed technical features.

[0075] Embodiment 1

[0076] An IDC computer room operation and maintenance system based on vision detection includes:

[0077] A data acquisition module for collecting data from the IDC power environment, physical machine out-of-band management system, virtual machine cloud platform, and switch and firewall monitoring system;

[0078] An image processing module for generating images from the collected data and using OCR technology for information extraction.

[0079] An AI inference module for loading a pre-trained large model and analyzing and inferring the extracted image information;

[0080] The model training module is used to label the inference results and images and use them as new training samples for fine-tuning the original model.

[0081] The visualization and presentation module is used to display the inference results and operation and maintenance suggestions in the form of a UI.

[0082] Embodiment 2

[0083] Based on Embodiment 1, in the data acquisition module, a multi-threaded mechanism is specifically used to acquire data.

[0084] Embodiment 3

[0085] Based on Embodiment 2, the multi-threaded mechanism for data acquisition includes:

[0086] SNMP polling: Regularly obtain the operating status of the device using the SNMP protocol.

[0087] NetFlow analysis: Obtain network traffic data through the NetFlow protocol and analyze the network traffic pattern.

[0088] API call: Obtain real-time data through the RESTful API of the device.

[0089] Embodiment 4

[0090] Based on Embodiment 1, in the image processing module, the specific steps for generating an image from the acquired data are as follows:

[0091] Convert the color image to a grayscale image through grayscale processing, and then convert the grayscale image to a black-and-white image through binary processing and perform denoising processing.

[0092] Embodiment 5

[0093] Based on Embodiment 1, in the image processing module, the specific steps for information extraction using OCR technology are as follows:

[0094] Use the moe model to identify the text content and understand the content of the abnormal information in the text.

[0095] Embodiment 6

[0096] Based on Embodiment 5, in the AI inference module, the pre-trained large model is loaded to analyze and infer the extracted image information, specifically including:

[0097] After the user inputs through the page, the background uses the moe model to route to experts in relevant fields, and the experts then perform inferences through the model.

[0098] Embodiment 7

[0099] Based on Embodiment 5, in the model training module, the fine-tuning is specifically carried out according to the following sub-steps:

[0100] Distributed training fine-tuning, where the fine-tuning content corresponds to each computing power card, and relevant knowledge is merged into the moe model in sequence according to the load balancing algorithm.

[0101] Embodiment 8

[0102] Based on Embodiment 5, in the visualization module, the inference results and operation and maintenance suggestions are displayed in the form of UI, specifically including the following sub-steps: Output based on the moe model and send text messages.

[0103] Embodiment 9

[0104] A method for implementing an IDC computer room operation and maintenance system based on visual detection, based on the IDC computer room operation and maintenance system according to any one of Embodiments 1 to 8, includes the following steps:

[0105] First, perform hardware deployment, install visual detection devices in the IDC computer room, and dock with various management systems;

[0106] In terms of software configuration, configure the data acquisition module and the image processing module to ensure the acquisition and conversion of real-time data, and then load the pre-trained large model and set the update mechanism.

[0107] Embodiment 10

[0108] An operation and maintenance method for an IDC computer room operation and maintenance system based on visual detection, based on the IDC computer room operation and maintenance system according to any one of Embodiments 1 to 8, includes the following steps:

[0109] Step 1, monitor the device status in real time and discover potential risks through visual detection;

[0110] Step 2, combine the inference results of the large model, formulate an operation and maintenance plan and present it through the UI;

[0111] Step 3, regularly fine-tune the model to adapt to changes in devices and the environment.

[0112] The units involved in the embodiments of the present invention can be implemented in software or in hardware, and the described units can also be set in the processor. Among them, the names of these units do not constitute a limitation to the unit itself in some cases.

[0113] According to one aspect of an embodiment of the present invention, there is provided a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the methods provided in the above various optional implementation manners.

[0114] As another aspect, an embodiment of the present invention further provides a computer-readable medium. The computer-readable medium may be included in the electronic device described in the above embodiment; or may exist separately without being assembled into the electronic device. The above computer-readable medium carries one or more programs. When the above one or more programs are executed by an electronic device, the electronic device implements the methods described in the above embodiment.

Claims

1. An IDC computer room operation and maintenance system based on visual detection, characterized in that, Including: A data acquisition module, which is used to collect data from the IDC power environment monitoring system, the out-of-band management system of physical machines, the virtual machine cloud platform, and the switch and firewall monitoring systems; An image processing module, which is used to generate images from the collected data and perform information extraction using OCR technology; An AI inference module, which is used to load a pre-trained large model and analyze and infer the extracted image information; A model training module, which is used to label the inference results and images and use them as new training samples to fine-tune the original model; A visualization presentation module, which is used to display the inference results and operation and maintenance suggestions in the form of a UI.

2. The IDC computer room operation and maintenance system based on visual detection according to claim 1, characterized in that, In the data acquisition module, a multi-threaded mechanism is specifically used to collect data.

3. The IDC computer room operation and maintenance system based on visual detection according to claim 2, characterized in that, The multi-threaded mechanism for collecting data includes: SNMP polling: Regularly obtain the operating status of devices using the SNMP protocol; NetFlow analysis: Obtain network traffic data through the NetFlow protocol and analyze network traffic patterns; API call: Obtain real-time data through the RESTful API of the device.

4. The IDC computer room operation and maintenance system based on visual detection according to claim 1, characterized in that In the image processing module, the generation of images from the collected data specifically includes the following sub-steps: Convert the color image to a grayscale image through grayscale processing, and then convert the grayscale image to a black-and-white image through binary processing and perform denoising processing.

5. The IDC computer room operation and maintenance system based on visual detection according to claim 1, wherein In the image processing module, the use of OCR technology for information extraction specifically includes the following sub-steps: Use the moe model to identify the text content and understand the content of abnormal information in the text.

6. The IDC computer room operation and maintenance system based on visual detection according to claim 5, characterized in that In the AI inference module, the loading of the pre-trained large model and the analysis and inference of the extracted image information specifically include: After the page user inputs, the background uses the moe model to route to experts in relevant fields, and the experts then perform inferences through the model.

7. The IDC computer room operation and maintenance system based on visual detection according to claim 5, characterized in that, In the model training module, the fine-tuning specifically includes the following sub-steps: Distributed training for fine-tuning, where the fine-tuning content corresponds to each computing power card, and relevant knowledge is merged into the moe model in sequence according to the load balancing algorithm.

8. The IDC computer room operation and maintenance system based on visual detection according to claim 5, characterized in that, In the visualization module, the display of the inference results and operation and maintenance suggestions in the form of a UI specifically includes the following sub-steps: Output based on the moe model and send text messages.

9. A method for implementing an IDC computer room operation and maintenance system based on visual detection, characterized in that, For the IDC computer room operation and maintenance system based on visual detection according to any one of claims 1 to 8, it includes the following steps: First, perform hardware deployment, install visual detection devices in the IDC computer room, and connect them to various management systems; In terms of software configuration, configure the data acquisition module and the image processing module to ensure the acquisition and conversion of real-time data, and then load the pre-trained large model and set up an update mechanism.

10. A maintenance method for an IDC computer room operation and maintenance system based on visual detection, characterized in that, For the IDC computer room operation and maintenance system based on visual detection according to any one of claims 1 to 8, it includes the following steps: Step 1, Monitor the device status in real time and discover potential risks through visual detection; Step 2, Combine the inference results of the large model, formulate an operation and maintenance plan and present it through the UI; Step 3, Regularly fine-tune the model to adapt to device and environmental changes.

Images