A method for resource auditing and alarming based on intelligent identification of data center equipment ports

Abstract

This invention provides a method for resource auditing and alarming based on intelligent identification of data center equipment ports, belonging to the field of automated data center equipment inspection technology. This invention automatically collects equipment images through data center cameras and uses a deep learning algorithm image recognition model to automatically and intelligently identify equipment port occupancy information. The platform connects to the resource system via an interface to obtain standard information on equipment port occupancy, performs automated data comparison and two-way resource auditing, and provides automated data audit comparison statistical reports, issuing alarm prompts for abnormal port occupancy. This improves data center inspection efficiency and quickly and effectively identifies potential equipment risks.

Description

Technical Field

[0001] This invention relates to the field of automated inspection technology for data center equipment, and in particular to a method for resource auditing and alarming based on intelligent identification of data center equipment ports. Background Technology

[0002] Auditing data on data center equipment resources, especially ports and card slots, is extremely time-consuming. Existing technologies for auditing port or card slot resources generally follow these common practices:

[0003] 1. Network cameras are typically CCD or CMOS digital cameras, with a resolution generally of 1280*1024. The network cameras capture and transmit images of the entire network and store them on the system server. Storage is partitioned by device and time, and data transfer is usually handled via FTP. The system allows setting image deletion times, etc.

[0004] 2. Retrieve equipment data from the resource system and compare the equipment images collected on-site with the equipment information in the resource system one by one.

[0005] 3. Manually record any discrepancies and notify relevant personnel to make corrections.

[0006] 4. If a risk occurs in the computer room, all staff must be mobilized to check the surveillance footage one by one, which is time-consuming and laborious, and may not be able to locate the fault accurately in time, causing the best time for repair to be missed.

[0007] The shortcomings and deficiencies of existing technical solutions:

[0008] 1. Low efficiency in resource auditing. When auditing equipment resources, it is necessary to compare the equipment pictures taken on site with the data information in the asset management system one by one, which results in low auditing efficiency.

[0009] 2. Real-time auditing is not possible. When on-site equipment resources or system resources change, timely monitoring and early warnings are not available. Auditing can only be performed after the next on-site photo is taken.

[0010] 3. Consumes significant storage resources. Each time a resource audit is conducted, on-site photos need to be taken and stored. A large number of identical images exist, making it impossible to perform deduplication and filtering, thus consuming significant storage resources.

[0011] 4. The results of resource audits lack systematic support and are not displayed in reports. Summary of the Invention

[0012] To address the above technical issues, this invention provides a method for resource auditing and alarming based on intelligent identification of data center equipment ports. It employs deep learning image recognition technology for intelligent identification of equipment ports, and performs data auditing and alarm prompts based on the identification results.

[0013] The technical solution of this invention is:

[0014] A method for resource auditing and alarming based on intelligent identification of equipment ports in a data center is proposed. This method automatically captures images of equipment via cameras in the data center and uses a deep learning-based image recognition model to automatically and intelligently identify equipment port occupancy information. The platform interfaces with the resource system to obtain standard information on equipment port occupancy, performs automated data comparison and two-way resource auditing, and provides automated data audit comparison statistical reports with alarm prompts for abnormal port occupancy. This method completes the automated inspection and intelligent early warning process of the data center through steps such as camera data acquisition, application of R-CNN and fast R-CNN image detection algorithms, unified encoding of equipment port detection, obtaining standard port occupancy information from the resource system, generation of automated data audit comparison reports, and alarm prompts. This improves the efficiency of data center inspections and quickly and effectively identifies potential equipment risks.

[0015] Furthermore,

[0016] By acquiring camera data, applying image detection algorithms R-CNN and fast R-CNN, using unified encoding for device port detection, obtaining standard port occupancy information from the resource system, generating automated data audit comparison reports, and providing alarm prompts, the process of automated inspection and intelligent early warning for the data center is completed.

[0017] Based on the image information of the equipment collected by the data center camera, the fast R-CNN algorithm is used to automatically identify the device port, and the identification result is converted into OneHot encoding.

[0018] Furthermore,

[0019] When comparing codes, if the codes of both parties match, the correct code information is saved. After the next image is acquired, the code converted from the current image is simply compared with the previously saved code information. If the codes match, the currently acquired image is deleted without needing to be saved. An audit is only initiated when the code of the on-site image is inconsistent with the previous one.

[0020] Based on the audit report results, automated analysis is performed. By intelligently identifying device port occupancy information and resource system device information, automated data auditing is conducted to generate an audit report. The audit report is analyzed to identify potentially abnormal devices and issue alarms as soon as possible.

[0021] Furthermore,

[0022] The steps are as follows:

[0023] 1) Data acquisition: The camera takes a picture of the device port every preset time interval and transmits it to the intranet server;

[0024] 2) Image detection algorithms detect device port occupancy information. The main algorithms used are R-CNN (deep learning) algorithm and fast R-CNN algorithm.

[0025] 3) Device port detection results are uniformly encoded and transmitted using one-hot encoding.

[0026] 4) Connect with the resource system to conduct resource data audits for the entire equipment; regularly compare equipment information;

[0027] 5) If the port occupancy information of the network device in the currently collected image is correct after system audit, then save the correct image code. Compare the image code of the next collected image with the image code of the currently collected image. If they are inconsistent, it means that the port occupancy information has changed. For abnormal comparison results, record the data and send email and SMS alarms; if the codes are consistent, then delete the current image.

[0028] Furthermore,

[0029] Data auditing includes code comparison, data report analysis, and data alarm notification.

[0030] Encoding comparison: The one-hot encoding data of port occupancy intelligent identification is compared with the preset information of the port information of the network device and the card slot information of the transmission device in the resource system;

[0031] Data report analysis: Statistical reports are generated and analyzed based on the actual usage of ports and resource presets.

[0032] Data alarm prompts, intelligent analysis of report data, prediction of port anomalies or card slot anomalies and alerts via SMS and email, and data comparison is performed once a day.

[0033] Furthermore,

[0034] The data audit process is as follows:

[0035] 1) After image recognition, first query the device's resource data through the API interface based on the device's unique code (primary key), and convert it into the OneHot code on the resource side according to the port location;

[0036] 2) Image recognition uses OneHot unified encoding, and bit-by-bit comparison audit is performed based on the OneHot encoding of the data from the resource side;

[0037] 3) If the comparison results are consistent, delete the image, end the process, and proceed to analyze the next image;

[0038] 4) If the comparison results are inconsistent, save the image code, save the current image, and record the differences based on the image's recognition code and the original code data of the resource;

[0039] 5) Perform two-way audit and intelligent analysis on the report, send an alert, and end the process.

[0040] The beneficial effects of this invention are

[0041] 1. By uniformly encoding and storing device port occupancy information, data loss can be prevented compared to storing it separately. At the same time, resource data is also converted into a unified code, transforming the data audit process into a comparison of codes as a whole, which greatly improves the efficiency of resource audit.

[0042] 2. During the audit process, the encoding of the currently acquired image is compared with the correct encoding. If the encodings match, the currently acquired image is deleted, and the comparison continues with the next acquired image. This eliminates the need to save device images for every audit, reducing image storage resources. In resource data auditing, audits are only initiated when the image encoding changes on-site or when the resource data encoding changes, and only the device or resource encoding information is transmitted, significantly reducing the amount of data transmitted in intermediate data transfers.

[0043] 3. Automated Resource Audit. By intelligently identifying device port occupancy information and resource system device information, the system automatically performs two-way data audits, generates audit reports, analyzes these reports to identify potentially abnormal devices, and issues alerts.

[0044] Socioeconomic benefits:

[0045] This method automatically captures images of devices using cameras in the data center and, based on a deep learning image recognition model, automatically and intelligently identifies device port occupancy information. The platform connects to the resource system via an interface to obtain standard information on device port occupancy, performs automated data comparison and two-way resource auditing, and provides automated data audit comparison and statistical reports, issuing alarms for abnormal port occupancy.

[0046] The economic benefits include: intelligent data collection, saving two maintenance personnel per data center and 200,000 yuan in personnel costs annually. The savings in maintenance manpower are based on the size and number of data centers. Social benefits include: firstly, leveraging existing data center camera systems eliminates the need for significant additional investment in data center upgrades; secondly, the application of intelligent technology addresses pain points such as low maintenance efficiency, poor recognition accuracy, and delayed alarm analysis; and finally, automated reporting and auditing technology eliminates the need for manual data auditing, enabling efficient and high-quality resource inspection and early warning analysis. This technology can be widely adopted in all data centers nationwide. Attached Figure Description

[0047] Figure 1 This is a diagram of the overall architecture of the present invention;

[0048] Figure 2 This is a diagram showing the camera installation location of the present invention;

[0049] Figure 3 This is a diagram of the R-CNN network structure of the present invention.

[0050] Figure 4 This is a diagram of the Fast R-CNN network structure of the present invention;

[0051] Figure 5 This is a data audit flowchart of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0053] This solution uses camera-captured device images and deep learning algorithms to automatically and intelligently detect device port occupancy. The data is interfaced with a resource system for automatic data comparison and two-way resource auditing. This solution consists of the following parts:

[0054] 1. Data Acquisition: The camera takes a picture of the device port at preset intervals and transmits it to the intranet server. For example, it takes a picture every hour.

[0055] 2. Image detection algorithm: Detects device port occupancy information. The main algorithms used are R-CNN (deep learning) algorithm and fast R-CNN algorithm.

[0056] 3. The device port detection results are uniformly encoded and transmitted using one-hot encoding.

[0057] 4. Integrate with the resource system to conduct overall resource data audits for the equipment. Regularly compare equipment information.

[0058] 5. If the port occupancy information of the network device in the currently captured image is correct according to system audit, then save the correct image code. Compare the code of the next captured image with the code of the currently captured image. If they are inconsistent, it indicates that the port occupancy information has changed. For abnormal comparison results, record the data and send email and SMS alerts. If the codes are consistent, then delete the current image.

[0059] The overall architecture of this solution is as follows: Figure 1 As shown:

[0060] The system is mainly divided into 6 layers:

[0061] Raw equipment layer: refers to the equipment in the computer room, mainly network equipment and transmission equipment.

[0062] Data Acquisition Layer: For each device, a photo is taken at a preset interval, the interval length of which can be configured. Images are acquired using cameras located on the server rack. Figure 2 As shown.

[0063] The cameras are installed on the other side of the rack. The number installed depends on the coverage area of ​​the cameras and the actual conditions of the server room. Recommended camera parameters are as follows:

[0064] Dynamic pixels: 5 million

[0065] Static resolution: 2592*1944

[0066] Focusing mode: Auto

[0067] Image sensor: CMOS

[0068] Maximum frame rate: 30

[0069] It is recommended to use doorless cabinets, as this will result in clearer images, which is beneficial for image recognition and resource auditing of network and transmission equipment.

[0070] Data storage layer: This layer refers to the storage of photos taken by the camera at the network device's terminal port and the transmission device's card slot. The data is stored on a file server within the intranet, and it is recommended that the images be retained for one week. They will be automatically deleted after one week.

[0071] AI Intelligent Detection Layer: This layer uses R-CNN and fast R-CNN algorithms for image detection, detecting and identifying whether network device port information is occupied. R-CNN is based on algorithms such as Convolutional Neural Network (CNN), linear regression, and Support Vector Machine (SVM) to achieve object detection technology.

[0072] The R-CNN network structure is as follows: Figure 3 As shown:

[0073] The core steps of the R-CNN algorithm are as follows:

[0074] 1. Terminal location region extraction: Based on the selective search algorithm, 500 (or 800, 1000, etc.) candidate regions are extracted from each image. The final region similarity is generated by weighting the similarity based on color, texture, size, and overlap, and the position of the terminal in each image is modeled.

[0075] 2. Image feature extraction: First, the 500 candidate regions are divided into regions of the same size according to the image size. Then, AlexNet 5-layer CNN is used to extract image features to generate the terminal port region.

[0076] 3. Linear classification: The SVM algorithm is used for region 2 classification, and the classification result is whether the terminal port is occupied.

[0077] 4. Perform one-hot encoding based on the area location and whether the terminal port is occupied.

[0078] The Fast R-CNN network structure is as follows: Figure 4 As shown:

[0079] The core steps of the Fast R-CNN algorithm are as follows:

[0080] 1. Terminal location region extraction: Based on the selective search algorithm, 500 candidate regions are extracted for each image. The final region similarity is generated by weighting the similarity based on color, texture, size, and overlap, and the position of the terminal in each image is modeled.

[0081] 2. Image feature extraction: consistent with R-CNN.

[0082] 3. Fully connected: Transforms feature maps of the same dimension into ROI (regions of interest) feature vectors.

[0083] 4. Classification and Regression: A softmax function is used with a fully connected layer in a neural network to classify whether a region's location terminal is occupied. The activation function is ReLU, i.e., f(x) = Max(0,x).

[0084] Data Result Encoding: Based on the terminal identification situation, this scheme uses one-hot encoding for unified encoding. One-hot encoding represents categorical variables as binary vectors. Each integer value is represented as a binary vector, and except for the integer index, it is a zero value, marked as 1. In this scheme, one-hot encoding is applied to devices of the same type. Assuming that a device has a maximum of 9 terminal ports, a occupied terminal port is represented by 1, and an unoccupied terminal port is represented by 0. An example table using one-hot encoding is shown below:

[0085] Port 1 Port 2 Port 3 Port 4 Port 5 Port 6 Port 7 Port 8 Port 9 Equipment 1 1 1 0 0 1 1 0 0 0 Equipment 2 0 0 1 1 1 0 0 0 1 Equipment 3 1 1 1 1 1 1 1 0 0

[0086] Device 1 is represented as [110011000], Device 2 is represented as [001110001], and Device 3 is represented as [111111100].

[0087] Data auditing includes code comparison, data report analysis, and data alarm notification.

[0088] Encoding Comparison: The one-hot encoded data of port occupancy intelligent identification is compared with the preset information of the network device's port information and the transmission device's card slot information in the resource system. Using the resource data as a reference, discrepancies between the actual used ports or card slots are identified. Based on the actual port or card slot usage, anomalies in the preset resource data are deduced.

[0089] Data report analysis: Statistical reports are generated and analyzed based on the actual usage of ports and resource presets.

[0090] Data alarm prompts, intelligent analysis of report data, prediction of port anomalies or card slot anomalies and alerts via SMS and email, and data comparison is performed once a day.

[0091] The overall data audit process is as follows: Figure 5 As shown:

[0092] 1. After image recognition, the device's resource data is first queried through the API interface based on the device's unique code (primary key), and then converted into OneHot code on the resource side according to the port location.

[0093] 2. Image recognition uses OneHot unified encoding, and bit-by-bit comparison audit is performed based on the OneHot encoding of the data from the resource side.

[0094] 3. If the comparison results are consistent, delete the image, end the process, and proceed to analyze the next image.

[0095] 4. If the comparison results are inconsistent, save the image code, save the current image, and record the differences based on the image's recognition code and the original code data of the resource.

[0096] 5. Perform two-way audit and intelligent analysis on the reports, send alerts, and end the process.

[0097] This invention does not impose specific limitations on resource devices, including but not limited to network device ports, transmission device card slots, etc. During the resource audit process, the on-site collected images and resource system data are converted into similar codes, and the resource audit is transformed into a comparison of codes as a whole. Moreover, the audit is only initiated when the local code changes, thereby reducing intermediate data transmission.

[0098] The above description is merely a preferred embodiment of the present invention and is used only to illustrate the technical solution of the present invention, and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. A method for resource auditing and alarming based on intelligent identification of data center equipment ports, characterized in that, The platform automatically collects images of devices through cameras in the data center and uses a deep learning algorithm image recognition model to automatically and intelligently identify device port occupancy information. The platform connects to the resource system through an interface to obtain standard information on device port occupancy, performs automated data comparison and two-way audit of resources, and provides automated data audit comparison and statistical reports, giving alarm prompts for abnormal port occupancy. By acquiring camera data, applying image detection algorithms R-CNN and fast R-CNN, using unified encoding for device port detection, obtaining standard port occupancy information by connecting to the resource system, generating automated data audit comparison reports and alarm prompts, the process of automated inspection and intelligent early warning of the data center is completed. Based on the image information of the equipment collected by the camera in the data center, the fast R-CNN algorithm is used to automatically identify the device port and convert the identification result into OneHot encoding; When comparing codes, if the codes of both parties match, the correct code information is saved. After the next image is acquired, the code converted from the current image is simply compared with the previously saved code information. If the codes match, the currently acquired image is deleted without needing to be saved. An audit is only initiated when the code of the on-site image is inconsistent with the previous one. Data auditing includes code comparison, data report analysis, and data alarm notification. Encoding comparison: The one-hot encoded data of port occupancy intelligent identification is compared with the preset information of the network device port information and the card slot information of the transmission device in the resource system; Data report analysis: Statistical reports are generated and analyzed based on the actual port usage and resource presets. Data alarm prompts, intelligent analysis of report data, prediction of port or card slot abnormalities and alerts via SMS and email, and data comparison is performed daily.

2. The method according to claim 1, characterized in that, Based on the audit report results, automated analysis is performed. By intelligently identifying device port occupancy information and resource system device information, automated data auditing is conducted to generate an audit report. The audit report is analyzed to identify potentially abnormal devices and issue alarms as soon as possible.

3. The method according to claim 2, characterized in that, The steps are as follows: 1) Data acquisition: The camera takes a picture of the device port every preset time interval and transmits it to the intranet server; 2) Image detection algorithm, which detects device port occupancy information. The algorithms used include R-CNN deep learning algorithm and fastR-CNN algorithm; 3) Device port detection results are uniformly encoded and transmitted using one-hot encoding. 4) Connect with the resource system to conduct resource data audits for the entire equipment; regularly compare equipment information; 5) If the port occupancy information of the network device in the currently collected image is correct after system audit, then save the correct image code and compare the image code of the next collected image with the image code of the currently collected image. If they are inconsistent, it means that the port occupancy information has changed. For abnormal comparison results, record the data and send email and SMS alarms. If the encoding matches, delete the current image.

4. The method according to claim 3, characterized in that, The data audit process is as follows: 1) After image recognition, the device's resource data is first queried through the API interface based on the device's unique code, i.e., the primary key. Based on the port location, the data is converted into OneHot encoding on the resource side. 2) Image recognition uses OneHot unified encoding, and bit-by-bit comparison audit is performed based on the OneHot encoding of the data from the resource side; 3) If the comparison results are consistent, delete the image, end the process, and proceed to analyze the next image; 4) If the comparison results are inconsistent, save the image code, save the current image, and record the differences based on the image's recognition code and the original code data of the resource; 5) Perform two-way audit and intelligent analysis on the report, send an alert, and end the process.

Images