Network additional storage system

Abstract

The invention discloses a network attached storage system, which comprises a data storage module, an end side AI reasoning module and an intelligent retrieval module, and is characterized in that the end side AI reasoning module comprises a local NPU; the end side AI reasoning module is used for carrying out intelligent reasoning on any file stored in the data storage module by utilizing a plurality of neural network models deployed in a local NPU (Network Processing Unit) and determining intelligent index information corresponding to the file, and the intelligent index information of any file is used for indicating file contents and characteristics of the file; and the intelligent retrieval module is used for determining a file display interface of the system according to the intelligent index information corresponding to the at least one file stored in the data storage module. According to the method and the device, intelligent management and offline AI service of the system can be realized by utilizing the end-side AI reasoning module, so that the file management efficiency is improved, the storage file does not need to be uploaded to an external server, and the privacy and the data security of a user are ensured.

Description

Technical Field

[0001] This disclosure relates to the field of data management and analysis technology, and in particular to a network-attached storage system. Background Technology

[0002] Commonly used NAS solutions in the current technology include traditional NAS systems and private cloud systems that combine NAS with cloud AI. Traditional NAS systems lack intelligent management capabilities, offering only basic file storage, sharing, and backup functions. They cannot automatically identify and classify files, nor provide intelligent functions such as content indexing, intelligent retrieval, and automatic summarization, requiring manual file management by users, resulting in low efficiency. While private cloud systems that combine NAS with cloud AI can provide intelligent management functions, they require uploading stored files to cloud servers, posing a risk of leakage of user privacy and sensitive data. Therefore, commonly used NAS solutions in the current technology fail to meet users' modern data management needs for "efficiency, intelligence, privacy, and security." Summary of the Invention

[0003] In view of this, this disclosure proposes a technical solution for a network-attached storage system.

[0004] According to one aspect of this disclosure, a network-attached storage system is provided, comprising: a data storage module, an edge-side artificial intelligence (AI) inference module, and an intelligent retrieval module, wherein the edge-side AI inference module includes a local neural network processor (NPU); the edge-side AI inference module is used to perform intelligent inference on any file stored in the data storage module using multiple neural network models deployed within the local NPU to determine the intelligent index information corresponding to the file, wherein the intelligent index information of any file is used to indicate the file content and features of the file; the intelligent retrieval module is used to determine the file display interface of the system based on the intelligent index information corresponding to at least one file stored in the data storage module.

[0005] In one possible implementation, the edge AI inference module is configured to: for any file stored in the data storage module, determine at least one target neural network model corresponding to the file from among the plurality of neural network models based on the data type of the file; and determine the intelligent index information corresponding to the file based on the file and all its corresponding target neural network models.

[0006] In one possible implementation, the system further includes a user interaction module; the user interaction module is used to display the file display interface using any terminal device corresponding to the system.

[0007] In one possible implementation, the user interaction module is further configured to determine a user interaction instruction based on input data from any terminal device corresponding to the system; the intelligent retrieval module is configured to determine the processing result of the user interaction instruction by utilizing the data storage module and / or the edge AI inference module in response to the user interaction instruction.

[0008] In one possible implementation, the user interaction command includes a file retrieval command, and the processing result includes a file retrieval result; the intelligent retrieval module is used to determine the file retrieval result based on the file retrieval command and the intelligent index information corresponding to all files stored in the data storage module.

[0009] In one possible implementation, the user interaction instruction includes an index information editing instruction, and the processing result includes the adjusted index information of any file; the intelligent retrieval module is used to adjust the intelligent index information corresponding to at least one file stored in the data storage module according to the index information editing instruction, and determine the adjusted index information of the file.

[0010] In one possible implementation, the user interaction instruction includes an AI content generation instruction, and the processing result includes an AI content generation result; the intelligent retrieval module is used to determine AI content generation requirement information based on the AI ​​content generation instruction; the edge AI inference module is used to perform intelligent inference based on the AI ​​content generation requirement information and at least one file stored in the data storage module to determine the AI ​​content generation result.

[0011] In one possible implementation, the system further includes a resource management module; the resource management module is used to: monitor the system in real time and determine the real-time operating data of the system; and adjust the task priority of the edge AI inference module for intelligent inference of any file stored in the data storage module based on the real-time operating data.

[0012] In one possible implementation, the resource management module is further configured to determine the system's operational status analysis results based on the real-time operational data.

[0013] In one possible implementation, the system further includes a power supply module; the power supply module is used to supply power to the system based on a preset operating mode.

[0014] This disclosed network-attached storage system includes a data storage module, an edge AI inference module, and an intelligent retrieval module. The data storage module provides basic persistent data storage functionality for the entire NAS system. For any file stored in the data storage module, the edge AI inference module can utilize multiple neural network models deployed within its local NPU to perform intelligent inference on the file, determining the corresponding intelligent index information to indicate the file's content and characteristics. This enables automatic and intelligent file management without uploading files to external devices or servers, directly preventing the leakage of user privacy and sensitive data, improving the speed of intelligent inference and file management, and avoiding limitations imposed by network bandwidth and network traffic costs on the implementation of intelligent inference functions. Furthermore, implementing intelligent inference through the edge AI inference module reduces the NAS system's dependence on external cloud services, lowering the overall cost of using the NAS system. The intelligent retrieval module can determine the file display interface of the NAS system based on the intelligent index information corresponding to at least one file stored in the data storage module, thereby providing users with intuitive visual information, improving the convenience of manual management of stored files, and further improving the efficiency of file management and file retrieval. Through the collaborative work of the data storage module, the edge AI inference module, and the intelligent retrieval module, the user experience is improved.

[0015] Other features and aspects of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0016] The accompanying drawings, which are included in and form part of this specification, illustrate exemplary embodiments, features, and aspects of this disclosure together with the specification and serve to explain the principles of this disclosure.

[0017] Figure 1 A block diagram of a network-attached storage system according to an embodiment of the present disclosure is shown. Detailed Implementation

[0018] Various exemplary embodiments, features, and aspects of this disclosure will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.

[0019] As used herein, the terms “comprising,” “including,” “having,” or variations thereof are open-ended and include one or more of the stated features, integrals, elements, steps, components, or functions, but do not exclude the presence or addition of one or more other features, integrals, elements, steps, components, functions, or groups thereof.

[0020] When an element is referred to as “connected,” “coupled,” “responding,” or a variation thereof relative to another element, it may be directly connected, coupled, or responding to another element, or there may be an intermediate element present.

[0021] Although the terms first, second, third, etc., may be used herein to describe various elements / operations, these elements / operations should not be limited by these terms. These terms are only used to distinguish one element / operation from another. Therefore, without departing from the teachings of the inventive concept, a first element / operation in some embodiments may be referred to as a second element / operation in other embodiments.

[0022] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.

[0023] Furthermore, to better illustrate this disclosure, numerous specific details are set forth in the following detailed description. Those skilled in the art will understand that this disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art have not been described in detail in order to highlight the main points of this disclosure.

[0024] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant regions.

[0025] According to market analysis reports in the relevant field, the market size of Network Attached Storage (NAS) systems continues to expand at a compound annual growth rate of 16.6%. This market growth is reflected not only in the increasing demand from enterprise users for commercial-grade NAS, but also in the growing demand from home users for consumer-grade NAS due to the increasing use of digital content such as high-definition photos, high-resolution videos, and smart home devices. More and more families and small businesses are adopting NAS private clouds and multi-device sharing / backup centers.

[0026] However, commonly used NAS solutions in the current technology include traditional NAS systems and private cloud systems that combine NAS with cloud AI. Among them, traditional NAS systems lack intelligent management capabilities and can only provide basic file storage, sharing and backup functions. They cannot automatically identify and classify files, nor can they provide intelligent functions such as content indexing, intelligent retrieval and automatic summarization. Users need to manually manage files, which results in low file management efficiency.

[0027] While private cloud systems that combine NAS with cloud-based AI offer intelligent management capabilities, they also pose risks of user privacy and sensitive data leakage because they require uploading stored files to a cloud server for AI analysis and processing. Furthermore, the local NAS device in this solution lacks intelligent management capabilities and relies on cloud server technical support. This creates a high dependence on the cloud server provider and makes the system susceptible to network bandwidth and latency issues. For example, under limited network bandwidth, file uploads can be time-consuming, impacting user experience; and intelligent management is impossible when the local NAS device is offline. On the other hand, the continuous uploading of files from the local NAS device to the cloud server increases network bandwidth pressure and traffic costs for users. In some application scenarios, users may also need to provide computing resources to the cloud server, resulting in a high technical barrier and high operating costs for the entire NAS solution.

[0028] In conclusion, existing NAS solutions are insufficient to meet users' modern data management needs for "efficiency, intelligence, privacy, and security".

[0029] In view of this, embodiments of this disclosure provide a network attached storage system that can utilize an edge AI inference module with a local NPU within the network attached storage system, as well as multiple neural networks deployed in the local NPU, to achieve intelligent management and offline AI services for the network attached storage system. This improves the file management efficiency of the network attached storage system and eliminates the need to upload stored files to an external server, ensuring user privacy and data security. The network attached storage system provided in this disclosure will be described in detail below.

[0030] Figure 1 A block diagram of a network-attached storage system according to an embodiment of the present disclosure is shown. Figure 1 As shown, system 100 includes: data storage module 101, edge AI inference module 102, and intelligent retrieval module 103, wherein edge AI inference module 102 includes local NPU.

[0031] The edge AI inference module 102 is used to perform intelligent inference on any file stored in the data storage module 101 using multiple neural network models deployed in the local NPU to determine the intelligent index information corresponding to the file. The intelligent index information of any file is used to indicate the file content and features of the file.

[0032] The intelligent retrieval module 103 is used to determine the system's file display interface based on the intelligent index information corresponding to at least one file stored in the data storage module 101.

[0033] Specifically, the data storage module 101 can provide basic file storage functions for the system 100 based on a preset file system, and manage the preset storage medium corresponding to the data storage module 101. When the data storage module 101 receives any file uploaded by the user, it can write the file to its corresponding preset storage medium and update the metadata of the preset file system, such as file identifier, path, size, modification time, etc. The file identifier here may include file name, file encoding, file hash value, etc., and can be flexibly set according to actual usage needs. This disclosure does not impose specific limitations on it.

[0034] The specific form of the preset file system can be found in the implementation methods described in relevant technologies. For example, it can be the fourth-generation extended file system (EXT4), B-tree file system (Btrfs), Zettabyte file system (ZFS), or Windows OS. TM The standard file system (New Technology File System, NTFS) and the file system for Mac OS TM and iOS TM The Apple File System (APFS) and other file systems are not specifically limited in this disclosure. The specific form of the preset storage medium can be flexibly set according to actual usage needs. For example, it may include hard disk drives (HDDs), solid state drives (SSDs), solid state hybrid drives (SSHDs), SSD arrays, etc., and this disclosure does not specifically limit it.

[0035] To ensure user privacy and data security, the data storage module 101 can also be used to manage user accounts and operation permissions, allowing users to log in to the system 100 using any pre-registered account and manually manage all stored files according to the operation permissions corresponding to that account. The specific methods for implementing account and operation permission management in the data storage module 101 can be found in related technical implementations, and this disclosure does not impose specific limitations on them.

[0036] The data storage module 101 can also automatically perform data redundancy protection, data backup, and recovery for any stored file to prevent file corruption or data loss due to user misoperation, external network attacks on the system 100, or hardware failure. Specific methods for data redundancy protection, data backup, and recovery can be found in related technical implementations, and this disclosure does not impose specific limitations on them.

[0037] In one example, data redundancy protection can employ Redundant Array of Independent Disks (RAID) technology, which combines multiple independent storage media into a logical unit. This ensures the data integrity of any file stored in the system data storage module 101 even if one or more storage media in the logical unit fail, reducing the possibility of file corruption or data loss.

[0038] In one example, data backup can be performed using any of the following methods: full backup, incremental backup, or differential backup; and in the event that any of the backed-up files is damaged or data is lost, the damaged files can be automatically recovered using the backup files.

[0039] The data storage module 101 can also provide a snapshot function. Before the user performs high-risk management operations on the stored files (e.g., batch deletion of files), and / or before the edge AI inference module 102 performs batch processing on the stored files, the original data state of the data storage module 101 is recorded. This allows the data storage module 101 to be quickly rolled back to the recorded original data state in case of file processing anomalies or hardware failures in the system 100, thereby achieving rapid data recovery. The specific method for implementing the snapshot function can be found in related technical implementations, and this disclosure does not impose specific limitations on it.

[0040] Furthermore, the data storage module 101 can also provide a standardized file access interface for any terminal device corresponding to the system 100 based on a preset network protocol, so as to ensure that users can access the system 100 through any terminal device corresponding to the system 100. The specific form of the preset network protocol can be flexibly set according to actual usage requirements. For example, it can adopt Server Message Block (SMB), Network File System (NFS), File Transfer Protocol (FTP), or Web-based Distributed Authoring and Versioning (WebDAV), etc., and this disclosure does not specifically limit it.

[0041] For any file stored in the data storage module 101, the edge AI inference module 102 can utilize multiple neural network models deployed within the local NPU to perform intelligent inference on the file, determining the corresponding intelligent index information to indicate the file content and characteristics. The specific form of the edge AI inference module 102 can be flexibly configured according to actual usage requirements. For example, in addition to including a local NPU, to accelerate AI inference, the edge AI inference module 102 can also include a graphics processing unit (GPU), a field-programmable gate array (FPGA), or a central processing unit (CPU) with AI instruction set extensions, etc. This disclosure does not specifically limit its scope.

[0042] In one example, the edge AI inference module 102 can use a SoC (System on Chip) chip with an integrated NPU to reduce the overall power consumption of the system 100 while ensuring that the system 100 has a high AI inference capability.

[0043] The specific form of the local NPU can be found in the implementation methods in related technologies, and this disclosure does not impose any specific limitations on it.

[0044] The specific number of neural network models deployed in the local NPU, and the specific form of any neural network, can be flexibly set according to actual usage requirements. For example, it can include image classification models, object detection models, face recognition models, action recognition models, event detection models, keyframe extraction models, feature vector extraction models, semantic segmentation and extraction models for content search and similarity analysis, generative task models, and Large Language Models (LLMs), etc. This disclosure does not impose specific limitations on these. Preferably, any neural network model deployed in the local NPU is a lightweight model. A lightweight model can be defined as a neural network model obtained by reducing the network parameter size by performing model compression, pruning, or quantization on the neural network model while maintaining the accuracy of any neural network model. This reduces the computing power requirement of the system 100 while ensuring the AI ​​performance of the edge AI inference module 102, thereby increasing the overall cost and technical threshold of the edge AI inference module 102 and the system 100.

[0045] The specific content of the intelligent retrieval information corresponding to any file can be flexibly set according to actual usage needs and is related to the data type of the file. This disclosure does not impose specific limitations on this.

[0046] In one example, for any file of image type, the corresponding intelligent retrieval information may include: the file identifier, image classification result, object detection result, face recognition result, and feature vector, etc. The specific content of the image classification result, object detection result, and face recognition result can be flexibly set according to actual usage requirements, and this disclosure does not impose specific limitations on them.

[0047] In one example, for any video file, the corresponding intelligent retrieval information may include: the file identifier, video keyframes, object detection results, face recognition results, action recognition results, event detection results, and feature vectors. The specific content of the video keyframes, object detection results, face recognition results, action recognition results, and event detection results can be flexibly set according to actual usage requirements, and this disclosure does not impose specific limitations on them.

[0048] In one example, for any text-based file, the intelligent retrieval information for that file may include: the file identifier, content summary, content summary result, and feature vector, etc. The specific content of the content summary and content summary result can be flexibly set according to actual usage requirements, and this disclosure does not impose specific limitations on them.

[0049] The intelligent retrieval module 103 can transform the intelligent index information corresponding to at least one file stored in the data storage module 101 into intuitive visual information and generate the system's file display interface. This file display interface can represent a view of all files stored in the data storage module 101 displayed on any terminal device corresponding to the system 100, thereby improving the user's file retrieval and file management efficiency.

[0050] The specific form of the document display interface can be flexibly set according to actual usage needs. For example, it may include a waterfall view, a timeline view, a map view, or a 3D stereoscopic view. This disclosure does not impose any specific limitations on this.

[0051] In one example, if the data storage module 101 stores multiple files of image type, and the intelligent retrieval information corresponding to any file includes the file's shooting time, shooting location information, person recognition result, and object detection result, a timeline view file display interface can be generated based on the shooting time of each file; or, a map view file display interface can be generated based on the shooting location information of each file; or, a categorized album view file display interface can be generated based on the person recognition result and object detection result of each file.

[0052] The specific functions of the edge AI inference module 102 and the intelligent retrieval module 103 will be described in detail later in conjunction with the possible implementation methods of this disclosure, and will not be repeated here.

[0053] This disclosed network-attached storage system includes a data storage module, an edge AI inference module, and an intelligent retrieval module. The data storage module provides basic persistent data storage functionality for the entire NAS system. For any file stored in the data storage module, the edge AI inference module can utilize multiple neural network models deployed within its local NPU to perform intelligent inference on the file, determining the corresponding intelligent index information to indicate the file's content and characteristics. This enables automatic and intelligent file management without uploading files to external devices or servers, directly preventing the leakage of user privacy and sensitive data, improving the speed of intelligent inference and file management, and avoiding limitations imposed by network bandwidth and network traffic costs on the implementation of intelligent inference functions. Furthermore, implementing intelligent inference through the edge AI inference module reduces the NAS system's dependence on external cloud services, lowering the overall cost of using the NAS system. The intelligent retrieval module can determine the file display interface of the NAS system based on the intelligent index information corresponding to at least one file stored in the data storage module, thereby providing users with intuitive visual information, improving the convenience of manual management of stored files, and further improving the efficiency of file management and file retrieval. Through the collaborative work of the data storage module, the edge AI inference module, and the intelligent retrieval module, the user experience is improved.

[0054] In one possible implementation, the edge AI inference module 102 is used to: for any file stored in the data storage module 101, determine at least one target neural network model corresponding to the file from among multiple neural network models based on the data type of the file; and determine the intelligent index information corresponding to the file based on the file and all its corresponding target neural network models.

[0055] For files of different data types, it is typically necessary to invoke different types of neural network models for intelligent inference to ensure the accuracy and reliability of the intelligent index information for each file. Therefore, for any file stored in the data storage module 101, the edge AI inference module 102 can determine at least one target neural network model suitable for processing the file from among multiple neural network models, based on the file's data type.

[0056] The specific method for determining the data type of any file can be flexibly set according to actual usage requirements. For example, the data storage module 101 can directly determine the data type of each file based on its extension; or the edge AI inference module 102 can call a neural network model to identify the data type of each file to avoid inaccurate data type determination by the data storage module 101 based on the file extension. This disclosure does not impose specific limitations on this.

[0057] For any file stored in the data storage module 101, the edge AI inference model 102 can call its corresponding NPU, GPU and CPU computing resources, as well as the storage resources of the system 100, and use at least one target neural network model corresponding to the file to perform intelligent inference on the file. This disclosure does not make any specific limitations in this regard.

[0058] In one possible implementation, system 100 further includes: user interaction module 104; user interaction module 104 is used to display a file display interface using any terminal device corresponding to system 100.

[0059] Based on the above Figure 1 For example, Figure 1 As shown, system 100 also includes a user interaction module 104. The user interaction module 104 is capable of data interaction with the intelligent retrieval module 103.

[0060] The user interaction module 104 provides an application programming interface (API) for any terminal device corresponding to the system 100. This allows the terminal device to call and display the file display interface generated by the intelligent retrieval module 103 in real time, enabling users to perform file management and intelligent retrieval operations on the system 100 through the terminal device, thereby improving the convenience and efficiency of file management in the system 100. The specific form of any terminal device can be flexibly set according to actual usage needs. For example, it can be a user equipment (UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital assistant (PDA), handheld device, computing device, in-vehicle device, wearable device, etc., and this disclosure does not specifically limit it. The specific form of the API can refer to the implementation methods in related technologies, and may include RESTful API, GraphQL interface, or RPC (Remote Procedure Call) interface, etc., and this disclosure does not specifically limit it.

[0061] Furthermore, the user interaction module 104 can also provide the API to any third-party software program, such as a web browser page, desktop client software, mobile app, or dedicated touch screen interface, allowing users to access the system 100 using third-party software, thereby improving the scalability and flexibility of the system 100.

[0062] In one possible implementation, the user interaction module 104 is further configured to determine the user interaction instruction based on the input data of any terminal device corresponding to the system 100; the intelligent retrieval module 103 is configured to determine the processing result of the user interaction instruction by using the data storage module 101 and / or the edge AI inference module 102 in response to the user interaction instruction.

[0063] Furthermore, the user interaction module 104 can determine user interaction commands based on input data from any terminal device corresponding to the system 100. The specific method for determining the input data, as well as the specific format of the input data, are related to the actual situation of the terminal device and can be flexibly set according to actual usage needs. For example, it may include text data, voice data, etc., and this disclosure does not impose specific limitations on this. The specific format and content of the user interaction commands can also be flexibly set according to actual usage needs, and this disclosure does not impose specific limitations on this.

[0064] Upon receiving a user interaction command, the intelligent retrieval module 103 can, based on the specific content of the user interaction command, invoke the data storage module 101 and / or the edge AI inference module 102 to execute the interaction task indicated by the user interaction command and determine the processing result of the user interaction command. The specific content of the processing result of the user interaction command depends on the actual content of the user interaction command and can be flexibly set according to actual usage needs; this disclosure does not impose specific limitations on it.

[0065] In one possible implementation, the user interaction command includes a file retrieval command, and the processing result includes a file retrieval result; the intelligent retrieval module 103 is used to determine the file retrieval result based on the file retrieval command and the intelligent index information corresponding to all files stored in the data storage module 101.

[0066] The specific content of the file retrieval instruction can be flexibly set according to actual usage needs. For example, it can include accurate file information such as the file identifier and data type of the file to be retrieved, or it can include relatively vague reference information such as natural language description information and reference images of the file to be retrieved. This disclosure does not make specific limitations in this regard.

[0067] In one example, if a user needs to retrieve a video file stored in the data storage module 101 but cannot provide accurate file information such as the file identifier and file storage path, the file retrieval instruction may include: natural language description information describing the content of the video file. For example, the natural language description information may include: there is a scene in the video where a girl wearing a red dress is standing on the beach.

[0068] The intelligent retrieval module 103 can automatically perform file retrieval based on the file retrieval command and the intelligent index information corresponding to all files stored in the data storage module 101, and determine the file retrieval results.

[0069] In one example, when the file retrieval instruction includes a file identifier of the file to be retrieved, the intelligent retrieval module 103 can match the file identifier of the file to be retrieved with the file identifiers included in the intelligent index information corresponding to all files stored in the data storage module 101, and determine all files whose file identifiers match as the file retrieval results. The matching method used by the intelligent retrieval module 103 can refer to implementation methods in related technologies, such as the approximate nearest neighbor search algorithm based on Faiss (Facebook AI Similarity Search) or the inverted index technology based on Elasticsearch, etc., and this disclosure does not specifically limit it in this way.

[0070] In one example, when the file retrieval instruction includes natural language description information or reference information such as reference images of the file to be retrieved, the intelligent retrieval module 103 can determine the reference feature vector of the file to be retrieved based on the reference information, and perform similarity analysis on the reference feature vector and the feature vectors included in the intelligent index information corresponding to all files stored in the data storage module 101 to determine the similarity between the reference feature vector and the feature vector of each file, and then determine the files whose similarity meets the preset threshold as the file retrieval results.

[0071] In one possible implementation, the user interaction instructions include index information editing instructions, and the processing result includes the adjusted index information of any file; the intelligent retrieval module 103 is used to adjust the intelligent index information corresponding to at least one file stored in the data storage module 101 according to the index information editing instructions, and determine the adjusted index information of the file.

[0072] Due to factors such as the accuracy of the neural network model deployed in the local NPU and the performance limitations of the edge AI inference module 102 itself, the intelligent index information corresponding to any file automatically inferred by the edge AI inference module 102 may not match the actual content of the file. In this case, the user can adjust the intelligent index information corresponding to the file through the collaborative work of the user interaction module 104 and the intelligent retrieval module 103 to improve the accuracy of the intelligent index information.

[0073] Specifically, user interaction commands may include index information editing commands to instruct the intelligent retrieval module 103 to edit any file. The specific content of the index information editing commands can be flexibly set according to actual usage needs. For example, it may include the file identifier of the file to be adjusted, the content in the intelligent index information that needs to be adjusted, etc. This disclosure does not impose specific limitations on this.

[0074] The intelligent retrieval module 103 can adjust the intelligent index information corresponding to at least one file stored in the data storage module 101 according to the index information editing instructions, and determine the adjusted index information of the file. The specific method by which the intelligent retrieval module 103 adjusts the intelligent index information corresponding to any file can refer to the implementation methods for adjusting file information in related technologies, and this disclosure does not specifically limit it.

[0075] In one possible implementation, the user interaction instructions include AI content generation instructions, and the processing results include AI content generation results; the intelligent retrieval module 103 is used to determine AI content generation requirement information based on the AI ​​content generation instructions; and the edge AI inference module 102 is used to perform intelligent inference based on the AI ​​content generation requirement information and at least one file stored in the data storage module 101 to determine the AI ​​content generation results.

[0076] User interaction instructions may also include AI content generation instructions to perform intelligent inference based on at least one file stored in the data storage module 101, utilizing the intelligent retrieval module 103 and the edge AI inference module 102. The specific content of the AI ​​content generation instructions can be flexibly set according to actual usage requirements; for example, it may include target file information, generation task information, generation parameter information, contextual association information, etc., and this disclosure does not impose specific limitations on it.

[0077] The target file here can refer to at least one file that serves as the basis for AI content generation; the target file information may include file identifier, file storage path, file hash value, etc., and when the target file is a video file, it may also include the time range of the video file, etc., which can be flexibly set according to actual usage needs, and this disclosure does not make specific limitations in this regard.

[0078] The generated task information can be used to indicate the specific type of AI content to be generated. It can be flexibly set according to actual usage needs. For example, it can include text summary generation, video clip generation, image style transfer, intelligent album storyline generation, image / video character and scene reconstruction, and voice content generation, etc. This disclosure does not make specific limitations in this regard.

[0079] The generated parameter information can be used to indicate the technical parameters that need to be generated for AI content. Its specific content is related to the generated task information and can be flexibly set according to actual usage needs. For example, it may include the size parameters of the AI ​​content (number of text characters, video duration, image resolution, etc.), the style of the AI ​​content, content constraints, similarity threshold between the AI ​​content and the target file, etc. This disclosure does not make specific limitations on this.

[0080] Contextual information can be used to indicate the contextual data needed for batch generation of AI content or multiple iterations of AI content generation. For example, it can include the identifier of existing AI content generation results, the user's historical AI content preferences, etc. It can be flexibly set according to actual usage needs, and this disclosure does not make specific limitations on it.

[0081] After receiving the AI ​​content generation instruction, the intelligent retrieval module 103 can perform content parsing and semantic understanding of the instruction to extract pre-structured and standardized AI content generation requirement information. This allows the edge AI inference module 102 to more accurately understand the user's AI content generation needs. The specific form of the AI ​​content generation requirement information can be flexibly set according to actual usage requirements. For example, it may include generation task type encoding, target file identifier list, generation parameter key-value pair set, context feature vector, etc. This disclosure does not impose specific limitations on this.

[0082] The edge AI inference module 102 can read at least one file from the data storage module 101 based on AI content generation requirement information, and use at least one neural network deployed in the local NPU to generate AI content based on the file, and determine the AI ​​content generation result. The specific method for AI content generation is related to the actual situation of the AI ​​content generation requirement information and can be flexibly set according to actual usage needs; this disclosure does not impose specific limitations on it.

[0083] In one example, the AI ​​content generation instruction includes: converting a portrait photo with the file identifier 001.jpg into an oil painting style; the intelligent retrieval module 103 can determine the AI ​​content generation requirement information based on the AI ​​content generation instruction, including: the target file identifier is "001.jpg", the generation task type is "image style conversion", and the style parameter is "oil painting style". The edge AI inference module 102 can read the portrait photo with the file identifier 001.jpg from the data storage module 101 based on the AI ​​content generation requirement information, extract the facial feature mask of 001.jpg using a face recognition model, encode the style descriptor into a text embedding vector, drive the lightweight diffusion model to perform conditional generation, obtain an oil painting style image, compress it into JPEG format, and determine the AI ​​content generation result.

[0084] In one possible implementation, system 100 further includes: a resource management module 105; the resource management module 105 is used to: monitor system 100 in real time and determine the real-time operating data of system 100; and adjust the task priority of the edge AI inference module 102 for intelligent inference of any file stored in data storage module 101 based on the real-time operating data.

[0085] The specific content of real-time operational data can be flexibly set according to actual usage needs. For example, it may include the resource usage of system 100, energy consumption data, temperature data, user interaction status, and real-time interaction response latency, etc. This disclosure does not impose specific limitations on this. Specifically, the specific content of resource usage data can be flexibly set according to actual usage needs. For example, it may include NPU utilization rate, GPU utilization rate, CPU load, remaining storage capacity of data storage module 101, storage medium read / write bandwidth, system memory utilization rate, etc. This disclosure does not impose specific limitations on this. The specific content of energy consumption data can be flexibly set according to actual usage needs. For example, it may include overall system energy consumption, energy consumption of edge AI inference module 102, etc. This disclosure does not impose specific limitations on this. The specific content of temperature data can be flexibly set according to actual usage needs. For example, it may include the SoC chip temperature, NPU temperature, GPU temperature, etc. of system 100, etc. This disclosure does not impose specific limitations on this.

[0086] The resource management module 105 can evaluate the operating status of the edge AI inference module 102 and the entire system 100 based on real-time operating data, and dynamically adjust the task priority of the edge AI inference module 102 for intelligent inference of any file stored in the data storage module 101. The specific settings of the task priorities can be flexibly configured according to actual usage needs. For example, they can include: urgent priority, high priority, normal priority, low priority, and pauseable / suspendable priority, and each priority has different preset inference rates and batch processing strategies, etc. This disclosure does not make specific limitations on this.

[0087] The resource management module 105 adjusts the specific method for the task priority of intelligent reasoning of any file stored in the data storage module 101 by the edge AI inference module 102. This method can be flexibly set according to actual usage requirements, and this disclosure does not impose specific limitations on it.

[0088] In one example, if the resource management module 105 determines that the real-time running data meets preset degradation conditions, the task priority of the edge AI inference module 102 performing intelligent inference on any file stored in the data storage module 101 can be reduced; if the resource management module 105 determines that the real-time running data meets preset upgrade conditions, the task priority of the edge AI inference module 102 performing intelligent inference on any file stored in the data storage module 101 can be increased. The specific content of the preset degradation and upgrade conditions can be flexibly set according to actual usage needs, and this disclosure does not impose specific limitations on them.

[0089] In one example, the preset degradation condition may include an NPU utilization threshold. If the resource management module 105 determines that the real-time NPU utilization of the system 100 is greater than the NPU utilization threshold, all intelligent inference tasks with normal priority in the edge AI inference module 102 can be adjusted to low priority to avoid NPU overload.

[0090] In one example, the preset degradation condition may include an interaction response latency threshold. If the resource management module determines that the user interaction state indicates that the user is interacting with the system 100 in real time, and the real-time interaction response latency is greater than the interaction response latency threshold, all intelligent inference tasks of the edge AI inference module 102 can be adjusted to low priority to ensure the smoothness of the user's real-time interaction with the system 100.

[0091] In one example, the preset upgrade conditions may include: the intelligent retrieval module 103 receiving a new AI content generation instruction. Once the resource management module determines that the intelligent retrieval module 103 has received a new AI content generation instruction, it can adjust the intelligent inference task corresponding to that instruction to an urgent priority to meet the user's needs first.

[0092] Through the above process, the priority of inference tasks can be dynamically optimized to ensure the overall operational stability of system 100 and its responsiveness to user interactions. This also optimizes the energy efficiency of the edge AI inference module 102, improves the utilization of storage and computing resources in system 100, and enables adaptive intelligent scheduling of system 100. Furthermore, by utilizing the resource management module 105 for software-level resource scheduling optimization, the performance requirements of edge AI inference on system 100 and edge AI inference module 102 can be reduced, thereby lowering the technical threshold and cost of edge AI inference.

[0093] In one possible implementation, the resource management module 105 is also used to determine the analysis results of the operating status of the system 100 based on real-time operating data.

[0094] Furthermore, the resource management module 105 can also perform multi-dimensional analysis of the operating status of the system 100, such as energy consumption analysis and fault detection, based on real-time operating data, and determine the operating status analysis results of the system 100. The specific content of the operating status analysis results can be flexibly set according to actual usage needs; for example, it may include energy consumption analysis results, fault detection results, and future operating status prediction results, etc., and this disclosure does not impose specific limitations on this.

[0095] In one possible implementation, system 100 further includes: a power supply module; the power supply module is used to supply power to system 100 based on a preset operating mode.

[0096] Specifically, the power supply module supports both external power input and internal battery power supply modes. Under normal external power input conditions, the power supply module converts the external power input voltage to a preset operating voltage to power system 100 and charges the internal battery. In the event of an external power input interruption or abnormality, the power supply module automatically switches to internal battery power to maintain stable operation of system 100 or ensure that system 100 saves files before stopping operation, thereby reducing the possibility of file corruption or data loss due to power supply abnormalities and improving the safety and reliability of system 100.

[0097] For specific methods of external power input, refer to the implementation methods in related technologies. For example, it may include AC to DC power adapter, USB-PD (Power Delivery) power interface, or Power over Ethernet, etc. This disclosure does not make specific limitations in this regard.

[0098] The specific form of the built-in battery can be flexibly set according to actual usage needs. For example, it may include lithium-ion batteries, lithium polymer batteries, nickel-metal hydride batteries, or supercapacitor energy storage modules, etc. This disclosure does not make specific limitations in this regard.

[0099] The specific content of the preset working mode of the power supply module can be flexibly set according to actual usage needs. For example, it may include high-performance mode, energy-saving mode, and standby mode, etc. This disclosure does not make specific limitations on this. The specific power supply power of any mode can be flexibly set according to actual usage needs, and this disclosure does not make specific limitations on this.

[0100] The power supply module can automatically adjust the preset working mode, or the resource management module 105 can adjust the preset working mode of the power supply module according to the operating status of the system 100 or the intelligent reasoning needs of the edge AI inference module 102 for any file stored in the data storage module 101, thereby flexibly adjusting the power supply to the system 100 and reducing the overall energy consumption of the system 100.

[0101] This disclosed network-attached storage system includes a data storage module, an edge AI inference module, and an intelligent retrieval module. The data storage module provides basic persistent data storage functionality for the entire NAS system. For any file stored in the data storage module, the edge AI inference module can utilize multiple neural network models deployed within its local NPU to perform intelligent inference on the file, determining the corresponding intelligent index information to indicate the file's content and characteristics. This enables automatic and intelligent file management without uploading files to external devices or servers, directly preventing the leakage of user privacy and sensitive data, improving the speed of intelligent inference and file management, and avoiding limitations imposed by network bandwidth and network traffic costs on the implementation of intelligent inference functions. Furthermore, implementing intelligent inference through the edge AI inference module reduces the NAS system's dependence on external cloud services, lowering the overall cost of using the NAS system. The intelligent retrieval module can determine the file display interface of the NAS system based on the intelligent index information corresponding to at least one file stored in the data storage module, thereby providing users with intuitive visual information, improving the convenience of manual management of stored files, and further improving the efficiency of file management and file retrieval. Through the collaborative work of the data storage module, the edge AI inference module, and the intelligent retrieval module, the user experience is improved.

[0102] It should be noted that, although... Figure 1 The above-described network attached storage system provided in this disclosure serves as an example, but those skilled in the art will understand that this disclosure is not limited thereto. In fact, users can flexibly configure the specific structure of the network attached storage system according to their personal preferences and / or actual application scenarios, as long as it can be based on the above principles and processes, utilizing the edge AI inference module with a local NPU in the network attached storage system, and multiple neural networks deployed in the local NPU, to achieve intelligent management and offline AI services for the network attached storage system, improve the file management efficiency of the network attached storage system, and ensure that the network attached storage system does not need to upload stored files to an external server, thus guaranteeing user privacy and data security.

[0103] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A network-attached storage system, characterized in that, The system includes: a data storage module, an edge AI inference module, and an intelligent retrieval module, wherein the edge AI inference module includes a local neural network processor (NPU); The edge AI inference module is used to perform intelligent inference on any file stored in the data storage module using multiple neural network models deployed in the local NPU to determine the intelligent index information corresponding to the file. The intelligent index information of any file is used to indicate the file content and features of the file. The intelligent retrieval module is used to determine the file display interface of the system based on the intelligent index information corresponding to at least one file stored in the data storage module.

2. The system according to claim 1, characterized in that, The edge AI inference module is used for: For any file stored in the data storage module, based on the data type of the file, at least one target neural network model corresponding to the file is determined among the plurality of neural network models; Based on the file and all its corresponding target neural network models, determine the intelligent index information corresponding to the file.

3. The system according to claim 1 or 2, characterized in that, The system also includes: a user interaction module; The user interaction module is used to display the file display interface using any terminal device corresponding to the system.

4. The system according to claim 3, characterized in that, The user interaction module is also used to determine user interaction instructions based on input data from any terminal device corresponding to the system. The intelligent retrieval module is used to respond to the user interaction command by using the data storage module and / or the edge AI inference module to determine the processing result of the user interaction command.

5. The system according to claim 4, characterized in that, The user interaction commands include file retrieval commands, and the processing results include file retrieval results; The intelligent retrieval module is used to determine the file retrieval result based on the file retrieval instruction and the intelligent index information corresponding to all files stored in the data storage module.

6. The system according to claim 4, characterized in that, The user interaction instructions include index information editing instructions, and the processing result includes the adjusted index information of any file; The intelligent retrieval module is used to adjust the intelligent index information corresponding to at least one file stored in the data storage module according to the index information editing instructions, and determine the adjusted index information of the file.

7. The system according to claim 4, characterized in that, The user interaction instructions include AI content generation instructions, and the processing results include AI content generation results; The intelligent retrieval module is used to determine the AI ​​content generation requirement information based on the AI ​​content generation instruction. The edge AI inference module is used to perform intelligent inference based on the AI ​​content generation requirement information and at least one file stored in the data storage module to determine the AI ​​content generation result.

8. The system according to claim 1 or 2, characterized in that, The system also includes: a resource management module; The resource management module is used for: The system is monitored in real time to determine its real-time operating data. Based on the real-time operating data, the task priority of the edge AI inference module for intelligent inference of any file stored in the data storage module is adjusted.

9. The system according to claim 8, characterized in that, The resource management module is also used to determine the system's operating status analysis results based on the real-time operating data.

10. The system according to claim 1 or 2, characterized in that, The system also includes: a power supply module; The power supply module is used to supply power to the system based on a preset working mode.

Images