Trusted distributed system governance device based on large language model

Abstract

The embodiment of the invention provides a trusted distributed system governance device based on a large language model. Mass complex distributed system governance documents and data are learned and understood through a multi-type large language model (LLM), so that distributed system governance experience is obtained, and reasonable, on-basis and timely diagnosis and optimization suggestions are provided for a target distributed system. The confidentiality in a core data cross-domain transmission process and a data calculation processing process is ensured through a trusted execution environment technology, and meanwhile, based on a remote authentication mechanism of a trusted execution environment, the consistency measurement check of data confidentiality and system governance logic is realized. The embodiment of the invention aims to realize a distributed system governance device taking multiple types of large language models as a center, and intelligent, automatic, efficient and credible large distributed system governance and maintenance optimization can be realized.

Description

Technical Field

[0001] The present invention relates to the field of distributed system management and large language models, and in particular to a trusted distributed system management device based on a large language model. Background Art

[0002] Large language models are an important achievement in the field of artificial intelligence in recent years. Large language models based on deep learning, such as GPT and BERT, have powerful natural language processing capabilities and can understand and generate human language. This makes them potential in understanding and solving distributed system governance problems. Currently, large distributed system governance faces the following difficulties and typical deficiencies of existing systems: System complexity: complex components and intricate dependencies make change deployment and troubleshooting difficult.

[0003] High operation and maintenance costs: A large amount of manpower is required to monitor the system, analyze logs, and locate problems, resulting in high operation and maintenance costs.

[0004] Data isolation: Data and logs are isolated between multiple systems and departments, making it difficult to conduct unified analysis.

[0005] Low degree of automation: Manual operations and self-service still account for a high proportion, and the degree of automation and intelligence is limited.

[0006] Audit governance is difficult: operation logs are difficult to record completely, and post-audit and accountability are difficult.

[0007] Configuration management chaos: Different system configurations are scattered everywhere, and manual or semi-automatic management is prone to problems.

[0008] Low governance participation: There are limited participants in organizational governance and operation and maintenance decision-making, information asymmetry, and low decision-making efficiency.

[0009] This requires applying large language models to distributed system governance to solve a variety of problems. First, large language models can help understand and analyze complex data such as logs, instructions, and variable states of distributed systems, so as to better monitor and manage various components of the system. Secondly, large language models can make predictions and optimizations based on historical data, and provide suggestions on system performance, resource utilization, and fault tolerance. In addition, large language models can also handle abnormal situations in distributed systems and provide rapid fault diagnosis and repair suggestions. This will improve operation and maintenance efficiency and governance level, and support the safe and efficient operation of complex large systems. Summary of the invention

[0010] The embodiment of the present invention provides a trusted distributed system management device based on a large language model to solve the defects existing in the prior art.

[0011] A trusted distributed system governance device based on a large language model, characterized in that it includes: a governance text classification vector converter, a governance prompt word generation model, a safe and trusted fault detection large language model, a resource usage detection large language model, a governance instruction decision tree model, an external auxiliary decision module, a governance instruction implementation module, a distributed system monitoring module, and several distributed system modules.

[0012] The governance text classification vector converter is used to classify the massive distributed system governance text corpus of the input device, and divide it into two categories according to the content of the corpus: corpus related to distributed system governance security and trustworthy failures and corpus related to distributed system governance resource usage load anomalies. At the same time, the classified corpus is converted into an input vector format acceptable to the large language model and input into the large language model for model tuning training.

[0013] The governance prompt word generation model collects the distributed system real-time monitoring data returned by the distributed system monitoring module, and generates prompt words acceptable to the safe and reliable fault detection large language model and the resource usage detection large language model according to the collected monitoring data types and values, so that the large language model can generate high-quality governance instruction responses.

[0014] The large language model for safe and reliable fault detection generates a response strategy and optimization suggestion text for dealing with safe and reliable faults of distributed systems according to the prompt words output by the governance prompt word generation model.

[0015] The resource usage detection language model generates a response strategy and optimization suggestion text for dealing with abnormal resource usage and load of the distributed system according to the prompt words output by the governance prompt word generation model.

[0016] The external decision-making assistance module can be connected to an external third-party expert system or knowledge base to assist in decision-making and knowledge retrieval for specific issues in the field of distributed system governance, providing sufficient and scalable information support for the governance instruction decision tree model.

[0017] The governance instruction decision tree model is used to receive the response strategies and optimization suggestion texts output by the safe and reliable fault detection large language model, the resource usage detection large language model, and the external auxiliary decision module to generate governance instructions that can be accepted by the governance instruction implementation module, and send the instructions to the governance instruction implementation module.

[0018] The governance instruction implementation module is used to receive the governance instructions generated by the governance instruction decision tree model, and then perform corresponding governance operations on the relevant modules involved in the distributed system according to the instruction requirements, so as to achieve the stability of the overall operation of the system and the correctness of the governance operation results.

[0019] The distributed system monitoring module is used to monitor the overall operation of the distributed system, mainly collecting information on distributed system security and trustworthiness-related faults and abnormal events, as well as resource consumption and load conditions of the overall operation of the system, and at the same time feeding back the collected fault and abnormal information to the governance prompt word generation model.

[0020] The several distributed system modules refer to various component modules of the distributed system, and different distributed system component modules may be different.

[0021] The trusted distributed system management device based on a large language model is characterized in that the secure and trusted fault detection large language model also requires: The large language model for secure and reliable fault detection based on the Transformer architecture and massive distributed system governance corpus is an innovative technology. The model is based on the deep learning Transformer model and combines specific problem response plans for security, reliability and fault detection in distributed system governance.

[0022] The model has a large scale and strong learning ability, and can learn and understand the meaning and rules of natural language. At the same time, by introducing a safe and reliable mechanism, the reliability of the model's operation and results is ensured. This includes the security of input data, the protection of model parameters, and the verifiability of output results.

[0023] The model has fault detection capabilities and can detect and correct errors and anomalies in input data. It is able to find and correct wrong predictions during inference and provide credible explanations and repair suggestions when needed.

[0024] The large language model for secure and reliable fault detection implemented based on the Transformer architecture combines the technical features of deep learning, security, and fault detection. It has powerful learning capabilities and reliable operation methods, and is suitable for the generation of response strategies and optimization suggestion texts for secure and reliable faults in the distributed system governance process.

[0025] The trusted distributed system governance device based on a large language model is characterized in that the resource usage detection large language model is an innovative model implemented based on the Transformer architecture. The model is constructed using the Transformer model of deep learning combined with a large amount of resource usage detection corpus. The model has large-scale parameters and strong learning ability, and can analyze and understand statements and descriptions about resource usage in natural language. By training the model, it can learn resource usage patterns and rules, and provide predictions and analysis of resource usage. The large language model can be used in multiple fields and scenarios, such as energy management, the Internet of Things, cloud computing, etc. By processing natural language descriptions and instructions, it can help users better understand and manage resource usage. For example, it can predict the energy consumption of equipment, the load of servers, the demand for network bandwidth, etc. In practical applications, the model has high accuracy and reliability, and can provide users with accurate resource usage predictions and suggestions. At the same time, it can also show users the resource usage through explanation and visualization to help users make better decisions.

[0026] The described trusted distributed system governance device based on a large language model is characterized in that the governance instruction implementation module is implemented based on a trusted execution environment. The module adopts the security and protection mechanism provided by the trusted execution environment to ensure the reliability and credibility of the governance instructions. By running in a trusted execution environment, the module can protect the security of sensitive data and key operations. It provides a secure execution environment to ensure the confidentiality, integrity and verifiability of instructions. In this way, trusted governance can be achieved in a distributed system. Trusted governance of distributed systems includes controlling and coordinating the behavior of distributed nodes, monitoring and managing the status of the system, handling anomalies and faults, etc. The governance instruction implementation module can run in a distributed system and pass instructions to each node in a secure and reliable manner to ensure that the governance instructions are correctly executed in the system.

[0027] This governance instruction implementation module based on a trusted execution environment has the following advantages: First, it provides a higher level of security protection to prevent unauthorized access and tampering. Second, through the verifiability mechanism, the execution process of the instruction can be tracked and audited to ensure the correctness and credibility of the instruction. Finally, the module can be seamlessly integrated with other governance components to provide more comprehensive and reliable distributed system governance capabilities.

[0028] Implementing a governance instruction implementation module based on a trusted execution environment is an innovative approach to ensure trusted governance of distributed systems. It provides a secure execution environment by protecting sensitive data and key operations, and implements correct execution and auditing of instructions through a verifiable mechanism. This approach helps improve the credibility and security of distributed systems.

[0029] The described trusted distributed system governance device based on a large language model is characterized in that the governance prompt word generation model, the secure and trusted fault detection large language model, the resource usage detection large language model, and the governance instruction decision tree model are all deployed based on a trusted execution environment. The security hardware and software protection mechanisms provided by the trusted execution environment (TEE) are used to ensure the security of the large language model during execution. Under this deployment method, the operation of the large language model occurs in a trusted execution environment, which provides protection mechanisms such as secure isolation, encrypted storage, and computing protection to prevent sensitive data leakage and unauthorized access. The user's data is encrypted during model training and reasoning, and is only decrypted and processed in the trusted execution environment, ensuring the confidentiality of the data.

[0030] The advantages of this deployment method are data security and privacy protection. Since large language models usually need to process sensitive data, the risk of data leakage can be effectively reduced. At the same time, the trusted execution environment can provide protection measures to prevent malware and attackers from attacking and tampering with the model.

[0031] In addition, deploying large language models based on a trusted execution environment also has the advantage of verifiability. The trusted execution environment provides a verification mechanism that can verify the correctness and integrity of the model and ensure that the model's reasoning results are credible.

[0032] The example of the present invention aims to realize a distributed system management device centered on multiple types of large language models, which can realize intelligent, automatic, efficient and reliable large-scale distributed system management and maintenance optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0034] Figure 1 It is a structural diagram of a system provided by an embodiment of the present invention. Implementation

[0035] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] Figure 1 FIG. 1 is a schematic diagram of a structure of a trusted distributed system governance device based on a large language model provided in an embodiment of the present invention. Figure 1 As shown, the trusted distributed system governance device based on the large language model includes: a governance text classification vector converter, a governance prompt word generation model, a safe and trusted fault detection large language model, a resource usage detection large language model, a governance instruction decision tree model, an external auxiliary decision module, a governance instruction implementation module, a distributed system monitoring module, several distributed system modules and other modules and models.

[0037] In terms of trusted security: (1) The governance prompt word generation model, the large language model for secure and reliable fault detection, the large language model for resource usage detection, the governance instruction decision tree model, and the governance instruction implementation module in the device are deployed based on the trusted execution environment technology, so these modules have consistency measurement checking or remote authentication mechanisms. During the device operation phase, any user who has doubts about the security and reliability of the device can initiate a trusted execution environment remote authentication challenge to the corresponding module to check the consistency of the processing logic of the corresponding module. If the consistency measurement value of the processing logic is found to be different from the expected value, the system is judged to be untrustworthy and a security alarm is triggered.

[0038] (2) When the governance instructions are transmitted, the governance instruction decision tree model and the governance instruction implementation module involved in the system build a secure and trusted instruction transmission channel that can perform consistency measurement. At the same time, symmetric encryption of the transmitted data is performed based on the negotiated key to ensure the confidentiality of the transmitted governance instructions.

[0039] In terms of device processing flow: First, based on the collected distributed system governance corpus, a large amount of text corpus is collected from documents such as white papers and governance solutions for distributed systems. The corpus needs to cover various governance topics. The governance text classification vector converter performs data cleaning and annotation, cleans the corpus, and removes irrelevant content. At the same time, the corpus is classified according to the corpus topic, and texts related to different governance topics are marked. Then, the corpus is vectorized using a word vector training algorithm (Word2Vec, GloVe, etc.). Each word is trained with a fixed-dimensional word vector to represent its semantic information. The word bag model and other algorithms are used to calculate the vector representation of each text based on the word vector. Each paragraph of text will be converted into a dense vector of fixed length.

[0040] Pre-train and fine-tune the large language model for secure and trusted fault detection and the large language model for resource usage detection based on the vector output in the previous step. Pre-training can be performed using a general corpus, or the pre-trained large language model can be used directly. According to the specific application scenario, the secure and trusted fault detection and resource usage detection categories are distinguished, and the pre-trained models are fine-tuned to generate the final model. After training, the secure and trusted fault detection and resource usage detection large language models are deployed in the trusted execution environment.

[0041] During the operation of the distributed system, the distributed system monitoring module continuously detects the resource usage, node load and core logs of several modules in the distributed system, generates corresponding monitoring events and sends the relevant monitoring events to the governance prompt word generation model.

[0042] The governance prompt word generation model automatically generates relevant governance prompt words based on monitoring events. Its working principle is: first, collect various event data generated during the operation of the distributed system, which reflects the real-time status of the system. Then, train a sequence-to-sequence (seq2seq) neural network generation model based on the event data. The model contains an encoder and a decoder, which can learn patterns in event data and generate corresponding text. Finally, when a system event is detected in real time, the event description is input into the model, and the model will output a prompt word, give suggestions or warnings to help governance decisions. By generating personalized governance prompt words, the model can better support distributed operation and maintenance management. After the governance prompt word generation is completed, the governance prompt word is sent to the secure and trusted fault detection large language model and the resource usage detection large language model to generate distributed system governance strategies and optimization suggestion texts.

[0043] The instruction decision tree model receives the response strategies and optimization suggestion texts output by the safe and reliable fault detection large language model, the resource usage detection large language model, and the external auxiliary decision module to generate governance instructions that can be accepted by the governance instruction implementation module, and sends the instructions to the governance instruction implementation module. The governance instruction implementation module performs related distributed system governance operations according to the instruction content to complete a complete operation of the device.

[0044] The embodiment of the present invention provides a trusted distributed system governance device based on a large language model. Through learning and understanding a large amount of complex distributed system governance documents and data through multiple types of large language models (LLM), distributed system governance experience is obtained, and reasonable, well-founded, and timely diagnosis and optimization suggestions are provided for the target distributed system. The confidentiality of the core data cross-domain transmission process and the data calculation and processing process is guaranteed through the trusted execution environment technology, and at the same time, the consistency measurement check of data confidentiality and system governance logic is realized based on the remote authentication mechanism of the trusted execution environment.

[0045] In the embodiments of the present invention, the large language model, trusted execution environment and other technical examples are used to realize a distributed system management device centered on multiple types of large language models, which can realize intelligent, automated, efficient and reliable large-scale distributed system management and maintenance optimization. At the same time, the trusted execution environment technology is used to ensure data security in the management process, prevent malicious attackers from attacking the distributed system management device, and realize the security and reliability of the management process.

[0046] To sum up, a trusted distributed system governance device based on a large language model provided in an embodiment of the present invention gives full play to the characteristics of large language model technology and trusted execution environment technology, supports efficient and secure system governance in large-scale complex distributed system environments, and constructs a governance device that meets the requirements of specifications and can also meet the requirements of data security and privacy protection.

[0047] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0048] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A trusted distributed system management device based on a large language model, characterized in that: include: Governance text classification vector converter, governance prompt word generation model, safe and reliable fault detection large language model, resource usage detection large language model, governance instruction decision tree model, external auxiliary decision module, governance instruction implementation module, distributed system monitoring module, and several distributed system modules; The governance text classification vector converter is used to classify the massive distributed system governance text corpus of the input device, and divide it into two categories according to the content of the corpus: corpus related to distributed system governance security and trustworthy failures and corpus related to distributed system governance resource usage load anomalies. At the same time, the classified corpus is converted into an input vector format acceptable to the large language model and input into the large language model for model tuning training.

2. The governance prompt word generation model collects the distributed system real-time monitoring data returned by the distributed system monitoring module, and generates prompt words acceptable to the safe and reliable fault detection large language model and the resource usage detection large language model according to the collected monitoring data types and values, so that the large language model can generate high-quality governance instruction responses.

3. The large language model for safe and reliable fault detection generates a response strategy and optimization suggestion text for dealing with safe and reliable faults of distributed systems based on the prompt words output by the governance prompt word generation model.

4. The resource usage detection large language model generates a response strategy and optimization suggestion text for dealing with abnormal resource usage and load of the distributed system based on the prompt words output by the governance prompt word generation model.

5. The external decision-making support module can be connected to an external third-party expert system or knowledge base to provide auxiliary decision-making and knowledge retrieval for specific issues in the field of distributed system governance, and provide sufficient and scalable information support for the governance instruction decision tree model.

6. The governance instruction decision tree model is used to receive the response strategies and optimization suggestion texts output by the safe and reliable fault detection large language model, the resource usage detection large language model, and the external auxiliary decision-making module to generate governance instructions that can be accepted by the governance instruction implementation module, and send the instructions to the governance instruction implementation module.

7. The governance instruction implementation module is used to receive the governance instructions generated by the governance instruction decision tree model, and then perform corresponding governance operations on the relevant modules involved in the distributed system according to the instruction requirements, so as to achieve the stability of the overall operation of the system and the correctness of the governance operation results.

8. The distributed system monitoring module is used to monitor the overall operation of the distributed system, mainly collecting information on distributed system security and reliability-related faults and abnormal events, as well as resource consumption and load conditions of the overall system operation, and at the same time feeding back the collected fault and abnormal information to the governance prompt word generation model.

9. The several distributed system modules mentioned above refer to the various component modules of the distributed system, and different distributed system component modules may be different.

10. A trusted distributed system management device based on a large language model according to claim 1, characterized in that: The secure and reliable fault detection large language model also requires: The large language model for secure and reliable fault detection based on the Transformer architecture and massive distributed system governance corpus is an innovative technology. The model is based on the deep learning Transformer model and combines specific problem response plans for security, trustworthiness, and fault detection in distributed system governance.

11. The model has a large scale and strong learning ability, and can learn and understand the meaning and rules of natural language. At the same time, by introducing a safe and reliable mechanism, the reliability of the model's operation and results is ensured. This includes the security of input data, the protection of model parameters, and the verifiability of output results.

12. The model has fault detection capabilities and can detect and correct errors and anomalies in input data. It is able to discover and correct wrong predictions during reasoning and provide credible explanations and repair suggestions when needed.

13. The large language model for secure and reliable fault detection based on the Transformer architecture combines the technical features of deep learning, security, and fault detection. It has strong learning capabilities and reliable operation methods. It is suitable for the generation of response strategies and optimization suggestions for secure and reliable faults in the process of distributed system governance.

14. A trusted distributed system management device based on a large language model according to claim 1, characterized in that: The resource usage detection large language model is an innovative model based on the Transformer architecture. The model is constructed using the deep learning Transformer model combined with a massive resource usage detection corpus. The model has large-scale parameters and strong learning ability, and can analyze and understand statements and descriptions about resource usage in natural language. By training the model, it can learn resource usage patterns and rules, and provide predictions and analysis of resource usage. The large language model can be used in multiple fields and scenarios, such as energy management, the Internet of Things, cloud computing, etc. By processing natural language descriptions and instructions, it can help users better understand and manage resource usage. For example, it can predict the energy consumption of devices, server loads, network bandwidth requirements, etc. In practical applications, the model has high accuracy and reliability, and can provide users with accurate resource usage predictions and suggestions. At the same time, it can also show users the resource usage through explanations and visualizations to help users make better decisions.

15. The trusted distributed system management device based on a large language model according to claim 1, characterized in that: The governance instruction implementation module is implemented based on a trusted execution environment. The module uses the security and protection mechanisms provided by the trusted execution environment to ensure the reliability and credibility of governance instructions. By running in a trusted execution environment, the module can protect the security of sensitive data and key operations. It provides a secure execution environment to ensure the confidentiality, integrity and verifiability of instructions. In this way, trusted governance can be achieved in a distributed system. Trusted governance of distributed systems includes controlling and coordinating the behavior of distributed nodes, monitoring and managing the status of the system, handling anomalies and failures, etc. The governance instruction implementation module can run in a distributed system and pass instructions to each node in a secure and reliable manner to ensure that governance instructions are correctly executed in the system.

16. This governance instruction implementation module based on a trusted execution environment has the following advantages: First, it provides a higher level of security protection to prevent unauthorized access and tampering. Second, through the verifiability mechanism, the execution process of the instruction can be tracked and audited to ensure the correctness and credibility of the instruction. Finally, the module can be seamlessly integrated with other governance components to provide more comprehensive and reliable distributed system governance capabilities.

17. Implementing a governance instruction implementation module based on a trusted execution environment is an innovative approach to ensure trusted governance of distributed systems. It provides a secure execution environment by protecting sensitive data and key operations, and implements correct execution and auditing of instructions through a verifiable mechanism. This approach helps improve the credibility and security of distributed systems.

18. The trusted distributed system management device based on a large language model according to claim 1, characterized in that: The governance prompt word generation model, secure and trusted fault detection large language model, resource usage detection large language model, and governance instruction decision tree model are all deployed based on a trusted execution environment. The secure hardware and software protection mechanisms provided by the trusted execution environment (TEE) are used to ensure the security of the large language model during execution. In this deployment mode, the operation of the large language model occurs in a trusted execution environment, which provides protection mechanisms such as secure isolation, encrypted storage, and computing protection to prevent sensitive data leakage and unauthorized access. The user's data is encrypted during model training and inference, and is only decrypted and processed in the trusted execution environment, ensuring the confidentiality of the data.

19. The advantages of this deployment method are data security and privacy protection. Since large language models usually need to process sensitive data, the risk of data leakage can be effectively reduced. At the same time, the trusted execution environment can provide protection measures to prevent malware and attackers from attacking and tampering with the model.

20. In addition, deploying large language models based on a trusted execution environment also has the advantage of verifiability. The trusted execution environment provides a verification mechanism that can verify the correctness and integrity of the model and ensure that the model's reasoning results are credible.

Images