Storage resource allocation method and device, equipment and medium

Abstract

The invention discloses a storage resource allocation method and device, equipment and a medium, and relates to the technical field of computers, and the method comprises the following steps: obtaining an embedding vector of a current cue word, and searching a target text corresponding to the embedding vector in a preset knowledge base by using a first target pre-training model; original texts and embedded vectors in a preset operation related manual for the persistent volume declaration are stored in the preset knowledge base; and splicing the current prompt word, the target text and a preset executable operation list to obtain a to-be-processed request, identifying the to-be-processed request by using a first target pre-training model, and if an obtained identification result represents that the to-be-processed request is used for triggering calling of a target persistent volume statement management function, executing the target persistent volume statement management function based on the target parameter, obtaining each current persistent volume declaration; and binding each persistent volume with the corresponding current persistent volume declaration. According to the scheme, when the storage resources are allocated, the difficulty of completing the PVC management operation by the user is reduced.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a storage resource allocation method, device, equipment and medium. Background Art

[0002] The container cluster orchestration management system is an open source container orchestration management platform for automating the deployment, expansion and management of containerized applications, such as the Kubernetes system, which can manage container clusters across multiple hosts and provide a set of mechanisms to deploy, maintain and expand applications. The container cluster orchestration management system has strong scalability and flexibility, making it the preferred platform for modern microservice architecture and cloud-native applications. In the artificial intelligence development platform, the container cluster orchestration management system is a basic component of the artificial intelligence development platform, which is used to schedule and allocate inference services among multiple hosts with different specifications to make full use of cluster resources. In the container cluster orchestration management system, PV (PersistentVolumes) is a storage resource within the cluster. Based on the storage class, PV is bound to PVC (Persistent Volume Claims), so as to allocate PV to PVC.

[0003] In order to manage storage resources in a multi-tenant environment, when performing PVC management operations on the AIStation platform, such as querying, deleting, and creating PVCs, users are usually required to receive training from professionals or read the PVC operation manual on their own to complete the PVC management operations. This requires a high level of professionalism from users and is not friendly to users who are new to the platform.

[0004] It can be seen that how to reduce the difficulty of users completing PVC management operations when allocating storage resources is a problem that those skilled in the art need to solve. Summary of the invention

[0005] The purpose of the embodiments of the present invention is to provide a storage resource allocation method, device, equipment and medium, which can reduce the difficulty of users completing PVC management operations when performing storage resource allocation. The specific scheme is as follows:

[0006] In a first aspect, the present invention discloses a storage resource allocation method, comprising:

[0007] Obtain an embedding vector of a current prompt word related to a persistent volume declaration, and use a first target pre-trained model to search a preset knowledge base for a target text corresponding to the embedding vector of the current prompt word; wherein the preset knowledge base stores original texts and embedding vectors of the original texts, and the original texts are various texts in a preset operation-related manual for persistent volume declaration;

[0008] The current prompt word, the target text, and the preset executable operation list are spliced ​​to obtain a pending request, and the pending request is identified by using the first target pre-trained model. If the obtained identification result indicates that the pending request is used to trigger the call of a target persistent volume declaration management function, the target parameters required by the target persistent volume declaration management function are obtained;

[0009] Executing the target persistent volume declaration management function based on the target parameter to complete the persistent volume declaration management operation corresponding to the current prompt word to obtain each current persistent volume declaration;

[0010] Bind each persistent volume in the container cluster orchestration management system to the corresponding current persistent volume claim to allocate the persistent volume for the current persistent volume claim.

[0011] Optionally, before obtaining the embedding vector of the current prompt word related to the persistent volume declaration, the method further includes:

[0012] Segment each text in the manual related to the preset operation of the persistent volume declaration according to the preset segmentation requirements to obtain each original text, and encode each of the original texts using the second target pre-trained model to obtain an embedding vector of the original text;

[0013] Saving the original text and the embedding vector of the original text to a preset knowledge base;

[0014] Accordingly, the step of obtaining the embedding vector of the current prompt word related to the persistent volume declaration includes:

[0015] The second target pre-trained model is used to encode a current prompt word related to the persistent volume declaration to obtain an embedding vector of the current prompt word.

[0016] Optionally, the preset knowledge base also stores source information of each of the original texts;

[0017] The acquiring the embedding vector of the current prompt word related to the persistent volume declaration, and searching the preset knowledge base for the target text corresponding to the embedding vector of the current prompt word by using the first target pre-trained model, includes:

[0018] Get the embedding vector of the current hint word associated with the persistent volume claim and the source lookup requirement;

[0019] Based on the source information of each of the original texts, and using a first target pre-trained model, an initial text corresponding to the source search requirement and an embedding vector of the initial text are screened out from a preset knowledge base;

[0020] The target embedding vectors corresponding to the embedding vector of the current prompt word are searched in the embedding vector of the initial text to determine the initial text corresponding to each of the target embedding vectors as the target text corresponding to the embedding vector of the current prompt word.

[0021] Optionally, the target persistent volume declaration management function is a preset persistent volume declaration creation function and / or a preset persistent volume declaration deletion function, the target parameters required by the target persistent volume declaration management function include any one or more of a namespace, a storage class name, a storage class mode, an access mode, and a persistent volume declaration name, and the persistent volume declaration management operation is a persistent volume declaration creation operation and / or a persistent volume declaration deletion operation.

[0022] Optionally, executing the target persistent volume declaration management function based on the target parameter includes:

[0023] Generate unconfirmed information corresponding to the target persistent volume declaration management function, and receive confirmation information returned by the user based on the unconfirmed information; wherein the unconfirmed information indicates whether to execute the target persistent volume declaration management function;

[0024] If the confirmation information indicates that the target persistent volume declaration management function is confirmed not to be executed, the process of executing the target persistent volume declaration management function is terminated, and a prompt message for terminating execution is generated;

[0025] If the confirmation information indicates confirmation of executing the target persistent volume declaration management function, verifying the target parameters;

[0026] If the target parameter passes the verification, executing the target persistent volume declaration management function based on the target parameter;

[0027] If the target parameter fails to pass the verification and the current number of attempts to obtain the parameter is greater than a preset threshold, an abnormal prompt message is generated;

[0028] If the target parameter fails the verification and the current number of attempts to obtain the parameter is not greater than a preset threshold, a re-enter parameter prompt message is generated to receive a new target parameter returned by the user based on the re-enter parameter prompt message, and jump again to the step of verifying the target parameter.

[0029] Optionally, searching a preset knowledge base for a target text corresponding to the embedding vector of the current prompt word using the first target pre-trained model includes:

[0030] The first target pre-training model is used to determine the similarity between the embedding vector of each original text in a preset knowledge base and the embedding vector of the current prompt word, and the embedding vectors of a preset number of the original texts with the highest similarity are determined as target embedding vectors, so as to determine the original text corresponding to each target embedding vector as the target text corresponding to the embedding vector of the current prompt word.

[0031] Optionally, before binding each persistent volume in the container cluster orchestration management system with the corresponding current persistent volume declaration, the method further includes:

[0032] Creating a storage class based on preset storage definition parameters; wherein the preset storage definition parameters include a network file system service IP, a network file system service path, a recycling mechanism, a volume binding mode, and a total capacity;

[0033] Declare storage resource configuration information corresponding to each of the current persistent volumes and create each persistent volume in the container cluster orchestration management system using the storage class.

[0034] In a second aspect, the present invention discloses a storage resource allocation device, comprising:

[0035] A text acquisition module, used to acquire an embedding vector of a current prompt word related to a persistent volume declaration, and use a first target pre-trained model to search a preset knowledge base for a target text corresponding to the embedding vector of the current prompt word; wherein the preset knowledge base stores original texts and embedding vectors of the original texts, and the original texts are various texts in a preset operation-related manual for persistent volume declaration;

[0036] A parameter acquisition module, used to concatenate the current prompt word, the target text, and a preset executable operation list to obtain a pending request, and use the first target pre-trained model to identify the pending request, and if the obtained recognition result indicates that the pending request is used to trigger the call of a target persistent volume declaration management function, then obtain the target parameters required by the target persistent volume declaration management function;

[0037] A request management module, configured to execute the target persistent volume declaration management function based on the target parameter to complete the persistent volume declaration management operation corresponding to the current prompt word and obtain each current persistent volume declaration;

[0038] The resource allocation module is used to bind each persistent volume in the container cluster orchestration management system to the corresponding current persistent volume declaration to allocate the persistent volume for the current persistent volume declaration.

[0039] In a third aspect, the present invention discloses an electronic device, comprising:

[0040] Memory, used to store computer programs;

[0041] The processor is used to execute the computer program to implement the steps of the storage resource allocation method disclosed above.

[0042] In a fourth aspect, the present invention discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the steps of the aforementioned storage resource allocation method disclosed above are implemented.

[0043] It can be seen that the present invention obtains the embedding vector of the current prompt word related to the persistent volume declaration, and uses the first target pre-trained model to search the target text corresponding to the embedding vector of the current prompt word in the preset knowledge base; wherein the preset knowledge base stores the original text and the embedding vector of the original text, and the original text is each text in the preset operation related manual for the persistent volume declaration; the current prompt word, the target text and the preset executable operation list are spliced ​​to obtain a pending request, and the first target pre-trained model is used to identify the pending request. If the obtained recognition result indicates that the pending request is used to trigger the call of the target persistent volume declaration management function, the target parameters required by the target persistent volume declaration management function are obtained; the target persistent volume declaration management function is executed based on the target parameters to complete the persistent volume declaration management operation corresponding to the current prompt word to obtain each current persistent volume declaration; each persistent volume in the container cluster orchestration management system is bound to the corresponding current persistent volume declaration to allocate the persistent volume for the current persistent volume declaration.

[0044] The beneficial effect is that the preset knowledge base of the present invention stores the original text and the embedding vector of the original text, that is, the various texts in the preset operation-related manual for persistent volume declaration and their corresponding embedding vectors are stored. When the embedding vector of the current prompt word related to the persistent volume declaration issued by the user is obtained, the first target pre-trained model can search for the corresponding target text in the preset knowledge base, that is, the first target pre-trained model can supplement the content in the PVC operation-related manual corresponding to the current prompt word. Then, when the first target pre-trained model identifies the pending request obtained by splicing the current prompt word, the target text, and the preset executable operation list, it can replace the understanding of the persistent volume declaration management operation to be performed by the user, that is, when the recognition result represents that the pending request is used to trigger the call of the target persistent volume declaration management function, the target persistent volume declaration management function is executed based on the target parameters required by the acquired target persistent volume declaration management function. In this way, the user does not need to fully master the content in the preset operation-related manual for persistent volume declaration, which reduces the difficulty of the user to complete the PVC management operation. The subsequent persistent volume is bound to the current persistent volume declaration to complete the allocation of storage resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only 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.

[0046] Figure 1 A flow chart of a storage resource allocation method provided by an embodiment of the present invention;

[0047] Figure 2 A specific storage resource allocation principle diagram provided for an embodiment of the present invention;

[0048] Figure 3 A specific persistent volume declaration operation execution sequence diagram provided by an embodiment of the present invention;

[0049] Figure 4 A flow chart of a specific storage resource allocation method provided by an embodiment of the present invention;

[0050] Figure 5 A specific persistent volume declaration management operation execution schematic diagram provided for an embodiment of the present invention;

[0051] Figure 6 A specific persistent volume declaration management function execution flow chart provided by an embodiment of the present invention;

[0052] Figure 7 A schematic diagram of the structure of a storage resource allocation device provided by an embodiment of the present invention;

[0053] Figure 8 A structural diagram of an electronic device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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 protection scope of the present invention.

[0055] The container cluster orchestration management system is an open source container orchestration management platform for automating the deployment, expansion and management of containerized applications, such as the Kubernetes system, which can manage container clusters across multiple hosts and provide a set of mechanisms to deploy, maintain and expand applications. The container cluster orchestration management system has strong scalability and flexibility, making it the preferred platform for modern microservice architecture and cloud-native applications. In the artificial intelligence development platform, the container cluster orchestration management system is a basic component of the artificial intelligence development platform, which is used to schedule and allocate inference services among multiple hosts with different specifications to make full use of cluster resources. In the container cluster orchestration management system, PV is a storage resource within the cluster. PV is bound to PVC based on the storage class, so as to allocate PV to PVC.

[0056] In order to manage storage resources in a multi-tenant environment, when performing PVC management operations on the AIStation platform, such as querying, deleting, and creating PVCs, users are usually required to receive training from professionals or read the PVC operation manual on their own to complete the PVC management operations. This requires a high level of professionalism from users and is not friendly to users who are new to the platform.

[0057] The terms "including" and "having" in the specification of the present invention and the above-mentioned drawings, as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device including a series of steps or units is not limited to the listed steps or units, but may include steps or units that are not listed.

[0058] In order to enable those skilled in the art to better understand the scheme of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0059] Next, a storage resource allocation solution provided by an embodiment of the present invention is introduced in detail. Figure 1 A storage resource allocation method provided in an embodiment of the present invention includes:

[0060] Step S11: Obtain an embedding vector of a current prompt word related to a persistent volume declaration, and use a first target pre-trained model to search a preset knowledge base for a target text corresponding to the embedding vector of the current prompt word; wherein the preset knowledge base stores original text and the embedding vector of the original text, and the original text is each text in a preset operation-related manual for persistent volume declaration.

[0061] In this embodiment, before obtaining the embedding vector of the current prompt word related to the persistent volume declaration, it also includes: segmenting each text in the preset operation related manual for the persistent volume declaration according to the preset segmentation requirements to obtain each original text, and using the second target pre-trained model to encode each of the original texts to obtain the embedding vector of the original text; saving the original text and the embedding vector of the original text to the preset knowledge base. The preset operation related manual for the persistent volume declaration can specifically be a user manual, product description, installation and deployment manual, interface document, outline design and other manuals of the artificial intelligence development platform. First, according to the preset segmentation requirements, each text in the preset operation related manual for the persistent volume declaration is segmented to obtain each original text (general_text), wherein the preset segmentation requirements include paragraph word count requirements, punctuation segmentation requirements, and paragraph requirements. The paragraph word count requirements refer to the word count requirements for each paragraph of the original text. The punctuation segmentation requirements refer to segmenting based on the preset punctuation marks to obtain each original text, for example, each period is a paragraph, each semicolon is a paragraph, and the paragraph requirements refer to different The content topic is used as the benchmark for segmentation; secondly, the second target pre-trained model is used to encode each of the original texts to obtain the embedding vector (general_vector) of the original text, that is, the second target pre-trained model is a word embedding (Embedding) model; then the original text and the embedding vector of the original text are saved to a preset knowledge base. Therefore, the preset knowledge base contains at least two fields, a "general_text" field of text type, which is used to save the original text of each document, and a "general_vector" field of dense_vector type, which is used to save the embedding vector corresponding to each original text.

[0062] In this embodiment, the method of obtaining the embedding vector of the current prompt word related to the persistent volume declaration includes: using the second target pre-trained model to encode the current prompt word related to the persistent volume declaration to obtain the embedding vector of the current prompt word. Because the preset knowledge base stores the embedding vectors of each original text, after obtaining the current prompt word related to the persistent volume declaration, it is also necessary to use the second target pre-trained model to encode the current prompt word related to the persistent volume declaration to obtain the embedding vector of the current prompt word, so that the first target pre-trained model can be used to search the preset knowledge base for the target text corresponding to the embedding vector of the current prompt word.

[0063] Furthermore, the use of the first target pre-training model to search for the target text corresponding to the embedding vector of the current prompt word in the preset knowledge base includes: using the first target pre-training model to determine the similarity between the embedding vector of each original text in the preset knowledge base and the embedding vector of the current prompt word, and determining the embedding vectors of a preset number of original texts with the highest similarity as the target embedding vector, so as to determine the original text corresponding to each target embedding vector as the target text corresponding to the embedding vector of the current prompt word. The specific process of searching for the target text in the preset knowledge base is: first, using the first target pre-training model to determine the similarity between the embedding vector of each original text in the preset knowledge base and the embedding vector of the current prompt word, that is, calculating general_vector 1 Similarity S with prompt_vector 1 、general_vector 2 Similarity S with prompt_vector 2 , …, general_vector n Similarity S with prompt_vector n Next, the embedding vectors of the preset number of original texts with the highest similarity are determined as the target embedding vectors. For example, the preset number k is 3, where the similarity S 2 , S 7 , S 11 are the three similarities with the highest values, then general_vector 2 、general_vector 7 、general_vector 11 is the target embedding vector, then the corresponding general_text 2 、general_text 7 、general_text 11 It is the target text corresponding to the embedding vector of the current prompt word. Because in natural language processing, the same word may have different meanings in different sentences. By considering multiple similar but not identical embedding vectors, this context sensitivity can be captured more accurately. Therefore, in this embodiment, multiple target texts can be screened out, thereby improving the accuracy of the subsequent target pre-training model in identifying pending requests.

[0064] In this embodiment, the preset knowledge base also stores the source information of each of the original texts; the acquiring the embedding vector of the current prompt word related to the persistent volume declaration, and using the first target pre-trained model to search the preset knowledge base for the target text corresponding to the embedding vector of the current prompt word, includes: acquiring the embedding vector of the current prompt word related to the persistent volume declaration and the source search requirement; based on the source information of each of the original texts, and using the first target pre-trained model to filter out the initial text corresponding to the source search requirement and the embedding vector of the initial text from the preset knowledge base; searching for each target embedding vector corresponding to the embedding vector of the current prompt word in the embedding vector of the initial text, so as to determine the initial text corresponding to each of the target embedding vectors as the target text corresponding to the embedding vector of the current prompt word. The original texts come from different preset operation manuals for persistent volume declarations. There are a large number of original texts. If the corresponding target text is directly searched based on the embedding vector of the current prompt word, it will take a long time. Therefore, the preset knowledge base also stores the source information of each original text. Then, a preliminary screening can be performed according to the source search requirements, that is, based on the source information of each original text, and using the first target pre-trained model, the initial text corresponding to the source search requirements and the embedding vector of the initial text are screened out from the preset knowledge base. The operation of preliminary screening according to the source search requirements is very simple and does not require special processing. The source information and the source search requirements are directly matched to eliminate a large number of original texts that do not meet the conditions. That is, the target embedding vectors corresponding to the embedding vector of the current prompt word are searched in the embedding vector of the initial text to determine the initial text corresponding to each target embedding vector as the target text corresponding to the embedding vector of the current prompt word. Only more accurate embedding vector similarity calculations are performed in a smaller range, which greatly reduces the workload.

[0065] like Figure 2As shown in the figure, a specific storage resource allocation principle diagram is shown. The artificial intelligence development platform includes LLM (Large Language Model)-services, internal-rag-service (internal retrieval enhancement generation service), ElasticSearch (search server), storage management service, and LLM-agen (agent) service. The preset knowledge base is saved in ElasticSearch. The first target pre-trained model is used for natural language processing. Different first target pre-trained models can be pre-integrated in LLM-services. The first target pre-trained model is, for example, Qwen-Chat (chat model), ChatGLM3 (Generative Language Model), etc. The second target pre-trained model is used to encode text into embedded vectors. Different second target pre-trained models are pre-integrated in internal-rag-service. The second target pre-trained model is, for example, BGE (Bilingual General Embedding), GloVe (Global Vectors for Word Representation), ELMo (Embeddingsfrom Language Models), etc.; the LLM-agen service is used to process the current prompt word returned by the front end, and after calling different back-end services for processing, the result is returned to the UI interface (UserInterface) for display to the user. The storage storage management service is used to handle normal persistent volume declarations, storage class (StorageClass) creation, deletion, update, query and other operations and other business logic. It is also responsible for receiving requests and interacting with the container cluster orchestration management system to implement operations on container cluster orchestration management resources such as persistent volume declarations and storage classes.

[0066] Step S12: concatenate the current prompt word, the target text, and a preset executable operation list to obtain a pending request, and use the first target pre-trained model to identify the pending request. If the recognition result obtained indicates that the pending request is used to trigger the call of a target persistent volume declaration management function, then obtain the target parameters required for the target persistent volume declaration management function.

[0067] The first target pre-trained model is used to identify pending requests. Pending requests can be divided into two categories: one is a business consultation request, and the other is a business operation request. Among them, the business consultation request refers to the user asking the first target pre-trained model about business-related processes. For example, the pending request represents "I hope to enable the inference service to use the model weight file on the host machine. How should I do it?" The first target pre-trained model will provide users with professional operation process instructions based on the document content in the knowledge base and its own capabilities, so that users can obtain specific business processes and complete related operations according to the business processes later. Business operation requests refer to users proposing specific operations to the first target pre-trained model and triggering FunctionCalling. The execution of business logic is completed through interaction with the first target pre-trained model. For example, the pending request represents "create PVC", that is, the obtained recognition result represents that the pending request is used to trigger the call of the target persistent volume claim management function, that is, triggering Function Calling. If the target parameters required by the target persistent volume claim management function are missing at this time, the user is required to enter the corresponding target parameters. If the target parameters required by the target persistent volume claim management function are not missing at this time, the target persistent volume claim management function can be directly executed.

[0068] In this embodiment, the target persistent volume declaration management function is a preset persistent volume declaration creation function and / or a preset persistent volume declaration deletion function, and the target parameters required by the target persistent volume declaration management function include any one or more of the namespace, storage class name, storage class mode, access mode, and persistent volume declaration name. In the business operation request, the target persistent volume claim management function is a preset persistent volume claim creation function and / or a preset persistent volume claim deletion function. The target parameters required by the target persistent volume claim management function include namespace, storage class name, storage class mode, access mode, and may also include PVC name (pvc_name), volume capacity, and description. If the business operation request is specifically a PVC creation request, then the namespace can specifically represent the specific namespace corresponding to the PVC created by the user, the storage class name is the storage class used, the PVC name is the name of the created PVC, the access mode is selected by the user from ReadWriteOnce, ReadOnlyMany, and ReadWriteMany, the volume capacity is the size of the storage space allocated to the created PVC, and the description represents that the user can further describe the purpose of the PVC to be created or other related information.

[0069] Step S13: Execute the target persistent volume declaration management function based on the target parameters to complete the persistent volume declaration management operation corresponding to the current prompt word and obtain each current persistent volume declaration.

[0070] In this embodiment, the persistent volume claim management operation is a persistent volume claim creation operation and / or a persistent volume claim deletion operation. It is understandable that if the target persistent volume claim management function is a preset persistent volume claim creation function, then the persistent volume claim management operation is a persistent volume claim creation operation; if the target persistent volume claim management function is a preset persistent volume claim deletion function, then the persistent volume claim management operation is a persistent volume claim deletion operation, for example Figure 3 A specific persistent volume declaration operation execution sequence diagram is shown, which executes the target persistent volume declaration management function based on the target parameters to complete the persistent volume declaration management operation corresponding to the current prompt word, obtains each current persistent volume declaration, and returns each current persistent volume declaration to the LLM-agen service. The LLM-agen service returns the result to the UI interface to display it to the user.

[0071] Step S14: Bind each persistent volume in the container cluster orchestration management system to the corresponding current persistent volume declaration to allocate the persistent volume for the current persistent volume declaration.

[0072] Common storage scenarios for container cluster orchestration management: Database applications usually require persistent storage. Using PVC can ensure that the database data is still available after the Pod (container) is restarted. Some applications need to store logs in persistent volumes for log analysis or backup. Multiple Pods may need to access the same file sharing volume. In this case, shared storage volumes can be used to meet the needs, such as NFS (Network File System). In a cloud environment, the dynamic storage scheduling mechanism of container cluster orchestration management can be used to store applications in the cloud provider's persistent storage service. The storage management of container cluster orchestration management makes it more flexible and efficient to handle persistent storage in a containerized environment, supporting various application scenarios and simplifying complex storage configuration work. Specifically in the artificial intelligence development platform, it is usually used to store the weight files, configuration files, and training data of pre-trained models. The specific storage content depends on the services deployed by the user. If non-inference services are deployed, such as ordinary web (network) services, it may be used to store business data, databases, or user files.

[0073] In this embodiment, before binding each persistent volume in the container cluster orchestration management system with the corresponding current persistent volume declaration, it also includes: creating a storage class based on preset storage definition parameters; wherein the preset storage definition parameters include a network file system service IP, a network file system service path, a recycling mechanism, a volume binding mode, and a total capacity; creating each persistent volume in the container cluster orchestration management system according to the storage resource configuration information corresponding to each current persistent volume declaration and using the storage class. Get the preset storage definition parameters specified by the user, specifically the NFS service IP (Internet Protocol), NFS service path, storage class name, access mode, recycling mechanism, volume binding mode and total capacity to create storage class parameters, and create a storage class (StorageClass) based on the preset storage definition parameters; allocate the created storage class to the user group of the user according to a certain capacity on the system administrator management page, so that ordinary users or group administrator users can log in to the ordinary user interface and allocate the storage class to the scenario where the user joins according to a certain capacity; next, according to the storage resource configuration information corresponding to each current persistent volume declaration, and using the storage class to create each persistent volume in the container cluster orchestration management system, the storage class in the container cluster orchestration management automatically creates PV and binds it to PVC. The storage resource configuration information can specifically be volume type, capacity, performance and other information.

[0074] It can be seen that the present invention obtains the embedding vector of the current prompt word related to the persistent volume declaration, and uses the first target pre-trained model to search the target text corresponding to the embedding vector of the current prompt word in the preset knowledge base; wherein the preset knowledge base stores the original text and the embedding vector of the original text, and the original text is each text in the preset operation related manual for the persistent volume declaration; the current prompt word, the target text and the preset executable operation list are spliced ​​to obtain a pending request, and the first target pre-trained model is used to identify the pending request. If the obtained recognition result indicates that the pending request is used to trigger the call of the target persistent volume declaration management function, the target parameters required by the target persistent volume declaration management function are obtained; the target persistent volume declaration management function is executed based on the target parameters to complete the persistent volume declaration management operation corresponding to the current prompt word to obtain each current persistent volume declaration; each persistent volume in the container cluster orchestration management system is bound to the corresponding current persistent volume declaration to allocate the persistent volume for the current persistent volume declaration.

[0075] The beneficial effect is as follows: the preset knowledge base of the present invention stores the original text and the embedding vector of the original text, that is, the various texts in the preset operation-related manual for persistent volume declaration and their corresponding embedding vectors are stored. When the embedding vector of the current prompt word related to the persistent volume declaration issued by the user is obtained, the first target pre-trained model can search for the corresponding target text in the preset knowledge base, that is, the first target pre-trained model can supplement the content in the PVC operation-related manual corresponding to the current prompt word. Then, when the first target pre-trained model identifies the pending request obtained by splicing the current prompt word, the target text and the preset executable operation list, it can replace the understanding of the persistent volume declaration management operation to be performed by the user, that is, when the recognition result represents that the pending request is used to trigger the call of the target persistent volume declaration management function, the target persistent volume declaration management function is executed based on the target parameters required by the acquired target persistent volume declaration management function. In this way, the user does not need to fully master the content in the preset operation-related manual for persistent volume declaration, which reduces the difficulty of the user to complete the PVC management operation. The subsequent persistent volume is bound to the current persistent volume declaration to complete the allocation of storage resources.

[0076] See also Figure 4 The embodiment of the present invention discloses a specific storage resource allocation method. Compared with the previous embodiment, this embodiment further illustrates and optimizes the technical solution. It includes:

[0077] Step S21: Obtain an embedding vector of a current prompt word related to a persistent volume declaration, and use a first target pre-trained model to search a preset knowledge base for a target text corresponding to the embedding vector of the current prompt word; wherein the preset knowledge base stores original text and the embedding vector of the original text, and the original text is each text in a preset operation-related manual for persistent volume declaration.

[0078] In the process of obtaining the embedding vector of the current prompt word related to the persistent volume declaration, it is necessary to first obtain the current prompt word related to the persistent volume declaration input by the user. The user can enter the current prompt word in text form in the text input box, and can also enter the current prompt word in voice form in the voice input box. Voice input provides an interactive method that does not rely on vision, and text input can express the current prompt word more accurately. Supporting multiple input methods helps to improve the inclusiveness of the system, allowing users with different abilities and different needs to easily use the system, thereby expanding the user group.

[0079] Step S22: concatenate the current prompt word, the target text, and a preset executable operation list to obtain a pending request, and use the first target pre-trained model to identify the pending request. If the recognition result obtained indicates that the pending request is used to trigger a call to a target persistent volume declaration management function, then obtain the target parameters required for the target persistent volume declaration management function.

[0080] For example Figure 5 The diagram shows a specific persistent volume declaration management operation execution diagram. The first target pre-trained model supports users to enter prompt words and select the tools workflow, that is, the current prompt word, target text and preset executable operation list are spliced ​​to obtain pending requests. The first target pre-trained model recognizes the spliced ​​pending requests. Users can also not enter prompt words and not use the tools workflow. Instead, users independently complete the call of the target persistent volume declaration management function, but this method requires a high level of user expertise. The tools parameter (i.e., the preset executable operation list) is a list. Each tool in the list represents an executable operation. When the persistent volume declaration management operation is the "Create PVC" operation, the preset executable operation list can be specifically as follows: [

[0082] {

[0083] "type": "function",

[0084] "function": {

[0085] "name": "create_pvc",

[0086] "description": "To create a persistent volume claim (PVC), you need to specify the scenario (namespace), storage class (storage class), PVC name, access mode, volume capacity and description.",

[0087] "parameters": {

[0088] "type": "object",

[0089] "properties": {

[0090] "namespace": {

[0091] "type": "string",

[0092] "description": "The namespace where the persistent volume claim resides"

[0093] },

[0094] "storageClass": {

[0095] "type": "string",

[0096] "description": "Specified storage class"

[0097] },

[0098] "pvcName": {

[0099] "type": "string",

[0100] "description": "The name of the persistent volume claim"

[0101] },

[0102] "accessMode": {

[0103] "type": "string",

[0104] "enum": ["ReadWriteOnce", "ReadOnlyMany", "ReadWriteMany"],

[0105] "description": "Specified access mode"

[0106] },

[0107] "capacity": {

[0108] "type": "string",

[0109] "description": "Specified volume capacity, e.g. '10Gi'"

[0110] },

[0111] "description": {

[0112] "type": "string",

[0113] "description": "Description of the persistent volume claim"

[0114] }

[0115] }

[0116] }

[0117] }

[0118] } ]

[0120] The name of the target persistent volume declaration management function corresponding to the "create PVC" operation is "create_pvc", and the type is "function", indicating that this is an executable function. "parameters" are the parameters required by the function, and the type of the parameter is "object", that is, an object. The member variables of this parameter object are in "properties", including namespace, StorageClass, pvcName, accessMode, capacity, and description in this example.

[0121] Step S23: Generate confirmation information corresponding to the target persistent volume declaration management function, and receive confirmation information returned by the user based on the confirmation information; wherein the confirmation information indicates whether to execute the target persistent volume declaration management function.

[0122] After obtaining the target parameters required by the target persistent volume declaration management function, the pending confirmation information corresponding to the target persistent volume declaration management function is generated. The pending confirmation information includes the name of the target persistent volume declaration management function and its corresponding operation, so that the user can confirm whether to start executing the target persistent volume declaration management function to prevent the first target pre-trained model from recognizing errors and then starting to execute the wrong function, resulting in wrong results. Whether to execute the function needs to be determined based on the confirmation information returned by the receiving user based on the pending confirmation information.

[0123] Step S24: If the confirmation information indicates confirmation that the target persistent volume declaration management function will not be executed, the process of executing the target persistent volume declaration management function is terminated, and a termination prompt message is generated; if the confirmation information indicates confirmation that the target persistent volume declaration management function is executed, the target parameters are verified.

[0124] It can be understood that if the confirmation information indicates that a specific persistent volume claim management function is not to be executed, the execution process of the function will be interrupted and a prompt message will be displayed to indicate that the execution has been terminated; on the contrary, if the confirmation information indicates that the persistent volume claim management function needs to be executed, the target parameters required by the function will be verified next.

[0125] S25: If the target parameters pass the verification, the target persistent volume declaration management function is executed based on the target parameters to complete the persistent volume declaration management operation corresponding to the current prompt word and obtain each current persistent volume declaration.

[0126] For example Figure 6A specific persistent volume declaration management function execution flow chart is shown. If the target parameters pass the verification process, the target persistent volume declaration management function will be executed according to these parameters to complete the persistent volume declaration management task associated with the current prompt word and generate all related current persistent volume declarations.

[0127] Step S26: If the target parameter fails to pass the verification and the current number of attempts to obtain the parameter is greater than the preset threshold, an abnormal prompt message is generated; if the target parameter fails to pass the verification and the current number of attempts to obtain the parameter is not greater than the preset threshold, a re-enter parameter prompt message is generated to receive the new target parameter returned by the user based on the re-enter parameter prompt message, and jump again to the step of verifying the target parameter.

[0128] like Figure 6 As shown, if the target parameter fails to pass the verification and the user has tried many times, that is, the current number of attempts to obtain the parameter is greater than the preset threshold, an abnormal prompt message will be generated; however, if the parameter fails to pass the verification but the number of user attempts is still within the preset threshold, the user will be prompted to re-enter the parameter, and the user will be prompted with the wrong parameter and the cause of the error, and the new parameters returned by the user based on this prompt will be received, and the system will then perform the verification process for these new parameters again.

[0129] Verifying the target parameters can ensure data accuracy. That is, the verification process can verify whether the target parameters meet the expected format, range, and logical requirements. This helps avoid operation failures or data corruption due to parameter errors, thereby ensuring that the configuration information of storage resources is accurate and improving stability. By verifying the target parameters, potential errors can be discovered and corrected in advance, reducing system anomalies or crashes caused by parameter errors, and ensuring the smooth progress of persistent volume declaration management operations.

[0130] When the user re-enters the target parameters based on the prompt information for re-entering the parameters, computing resources need to be used to re-verify the new target parameters. However, in order to prevent excessive consumption of system resources, it is prohibited to allow users to re-try to enter the target parameters without restriction. Therefore, it can avoid affecting the overall performance and stability, and ensure that system resources are used reasonably. In some cases, frequent incorrect attempts may be regarded as potential security threats. Setting thresholds can be used as a simple security measure to prevent malicious users from attacking the system through brute force cracking and other methods by limiting the number of attempts.

[0131] Step S27: Bind each persistent volume in the container cluster orchestration management system to the corresponding current persistent volume declaration to allocate the persistent volume for the current persistent volume declaration.

[0132] It can be seen that when the user needs to execute the persistent volume declaration management function, he only needs to enter the prompt word. The first target pre-trained model can supplement the prompt word according to the various texts in the preset operation-related manual for persistent volume declaration in the preset knowledge base, thereby obtaining a pending request and determining whether the pending request is used to trigger the call of the target persistent volume declaration management function. If so, the target parameters are obtained to generate pending confirmation information corresponding to the target persistent volume declaration management function. When the user confirms the execution of the target persistent volume declaration management function, the persistent volume declaration management function can be executed to complete the persistent volume declaration management operation. The professional requirements for users are greatly reduced, and the learning burden of users is reduced. Users can quickly start using the platform without guidance, and understand the operation and resource concepts and usage methods of the artificial intelligence development platform during the interaction, which also reduces the burden on technical support personnel.

[0133] Figure 7 A schematic diagram of the structure of a storage resource allocation device provided in an embodiment of the present invention includes:

[0134] A text acquisition module 11 is used to obtain an embedding vector of a current prompt word related to a persistent volume declaration, and use a first target pre-trained model to search a preset knowledge base for a target text corresponding to the embedding vector of the current prompt word; wherein the preset knowledge base stores original texts and embedding vectors of the original texts, and the original texts are various texts in a preset operation-related manual for persistent volume declaration;

[0135] A parameter acquisition module 12 is used to concatenate the current prompt word, the target text, and a preset executable operation list to obtain a pending request, and use the first target pre-trained model to identify the pending request, and if the obtained recognition result indicates that the pending request is used to trigger the call of a target persistent volume declaration management function, then obtain the target parameters required by the target persistent volume declaration management function;

[0136] A request management module 13, configured to execute the target persistent volume declaration management function based on the target parameter to complete the persistent volume declaration management operation corresponding to the current prompt word and obtain each current persistent volume declaration;

[0137] The resource allocation module 14 is used to bind each persistent volume in the container cluster orchestration management system to the corresponding current persistent volume declaration, so as to allocate the persistent volume according to the current persistent volume declaration.

[0138] It can be seen that the present invention obtains the embedding vector of the current prompt word related to the persistent volume declaration, and uses the first target pre-trained model to search the target text corresponding to the embedding vector of the current prompt word in the preset knowledge base; wherein the preset knowledge base stores the original text and the embedding vector of the original text, and the original text is each text in the preset operation related manual for the persistent volume declaration; the current prompt word, the target text and the preset executable operation list are spliced ​​to obtain a pending request, and the first target pre-trained model is used to identify the pending request. If the obtained recognition result indicates that the pending request is used to trigger the call of the target persistent volume declaration management function, the target parameters required by the target persistent volume declaration management function are obtained; the target persistent volume declaration management function is executed based on the target parameters to complete the persistent volume declaration management operation corresponding to the current prompt word to obtain each current persistent volume declaration; each persistent volume in the container cluster orchestration management system is bound to the corresponding current persistent volume declaration to allocate the persistent volume for the current persistent volume declaration.

[0139] The beneficial effect is as follows: the preset knowledge base of the present invention stores the original text and the embedding vector of the original text, that is, the various texts in the preset operation-related manual for persistent volume declaration and their corresponding embedding vectors are stored. When the embedding vector of the current prompt word related to the persistent volume declaration issued by the user is obtained, the first target pre-trained model can search for the corresponding target text in the preset knowledge base, that is, the first target pre-trained model can supplement the content in the PVC operation-related manual corresponding to the current prompt word. Then, when the first target pre-trained model identifies the pending request obtained by splicing the current prompt word, the target text and the preset executable operation list, it can replace the understanding of the persistent volume declaration management operation to be performed by the user, that is, when the recognition result represents that the pending request is used to trigger the call of the target persistent volume declaration management function, the target persistent volume declaration management function is executed based on the target parameters required by the acquired target persistent volume declaration management function. In this way, the user does not need to fully master the content in the preset operation-related manual for persistent volume declaration, which reduces the difficulty of the user to complete the PVC management operation. The subsequent persistent volume is bound to the current persistent volume declaration to complete the allocation of storage resources.

[0140] Furthermore, the present application also discloses an electronic device. Figure 8This is a structural diagram of an electronic device according to an exemplary embodiment, and the content in the diagram cannot be regarded as any limitation on the scope of use of this application. The electronic device may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the storage resource allocation method disclosed in any of the aforementioned embodiments. In addition, the electronic device in this embodiment may specifically be an electronic computer.

[0141] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device; the communication interface 24 can create a data transmission channel between the electronic device and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0142] In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0143] The operating system 221 is used to manage and control various hardware devices on the electronic device and the computer program 222, which can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the storage resource allocation method performed by the electronic device disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks.

[0144] Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein the computer program, when executed by a processor, implements the storage resource allocation method disclosed above. The specific steps of the method can refer to the corresponding contents disclosed in the above embodiments, and will not be repeated here.

[0145] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0146] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0147] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0148] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0149] The technical solution provided by the present application is introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for general technicians in this field, according to the idea of ​​the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A storage resource allocation method, It is characterized in that include: Obtain an embedding vector of a current prompt word related to a persistent volume declaration, and use a first target pre-trained model to search a preset knowledge base for a target text corresponding to the embedding vector of the current prompt word; wherein the preset knowledge base stores original texts and embedding vectors of the original texts, and the original texts are various texts in a preset operation-related manual for persistent volume declaration; The current prompt word, the target text, and the preset executable operation list are spliced ​​to obtain a pending request, and the pending request is identified by using the first target pre-trained model. If the obtained identification result indicates that the pending request is used to trigger the call of a target persistent volume declaration management function, the target parameters required by the target persistent volume declaration management function are obtained; Executing the target persistent volume declaration management function based on the target parameter to complete the persistent volume declaration management operation corresponding to the current prompt word to obtain each current persistent volume declaration; Bind each persistent volume in the container cluster orchestration management system to the corresponding current persistent volume claim to allocate the persistent volume for the current persistent volume claim.

2. The storage resource allocation method according to claim 1, It is characterized in that Before acquiring the embedding vector of the current prompt word related to the persistent volume declaration, the method further includes: Segment each text in the manual related to the preset operation of the persistent volume declaration according to the preset segmentation requirements to obtain each original text, and encode each of the original texts using the second target pre-trained model to obtain an embedding vector of the original text; Saving the original text and the embedding vector of the original text to a preset knowledge base; Accordingly, the step of obtaining the embedding vector of the current prompt word related to the persistent volume declaration includes: The second target pre-trained model is used to encode a current prompt word related to the persistent volume declaration to obtain an embedding vector of the current prompt word.

3. The storage resource allocation method according to claim 1, It is characterized in that The preset knowledge base also stores the source information of each original text; The acquiring the embedding vector of the current prompt word related to the persistent volume declaration, and searching the preset knowledge base for the target text corresponding to the embedding vector of the current prompt word by using the first target pre-trained model, includes: Get the embedding vector of the current hint word associated with the persistent volume claim and the source lookup requirement; Based on the source information of each of the original texts, and using a first target pre-trained model, an initial text corresponding to the source search requirement and an embedding vector of the initial text are screened out from a preset knowledge base; The target embedding vectors corresponding to the embedding vector of the current prompt word are searched in the embedding vector of the initial text to determine the initial text corresponding to each of the target embedding vectors as the target text corresponding to the embedding vector of the current prompt word.

4. The storage resource allocation method according to claim 1, It is characterized in that The target persistent volume declaration management function is a preset persistent volume declaration creation function and / or a preset persistent volume declaration deletion function. The target parameters required by the target persistent volume declaration management function include any one or more of the namespace, storage class name, storage class mode, access mode, and persistent volume declaration name. The persistent volume declaration management operation is a persistent volume declaration creation operation and / or a persistent volume declaration deletion operation.

5. The storage resource allocation method according to claim 1, It is characterized in that The executing the target persistent volume declaration management function based on the target parameter includes: Generate unconfirmed information corresponding to the target persistent volume declaration management function, and receive confirmation information returned by the user based on the unconfirmed information; wherein the unconfirmed information indicates whether to execute the target persistent volume declaration management function; If the confirmation information indicates that the target persistent volume declaration management function is confirmed not to be executed, the process of executing the target persistent volume declaration management function is terminated, and a prompt message for terminating execution is generated; If the confirmation information indicates confirmation of executing the target persistent volume declaration management function, verifying the target parameters; If the target parameter passes the verification, executing the target persistent volume declaration management function based on the target parameter; If the target parameter fails to pass the verification and the current number of attempts to obtain the parameter is greater than a preset threshold, an abnormal prompt message is generated; If the target parameter fails the verification and the current number of attempts to obtain the parameter is not greater than a preset threshold, a re-enter parameter prompt message is generated to receive a new target parameter returned by the user based on the re-enter parameter prompt message, and jump again to the step of verifying the target parameter.

6. The storage resource allocation method according to claim 1, It is characterized in that The step of searching a preset knowledge base for a target text corresponding to the embedding vector of the current prompt word using the first target pre-trained model includes: The first target pre-training model is used to determine the similarity between the embedding vector of each original text in a preset knowledge base and the embedding vector of the current prompt word, and the embedding vectors of a preset number of the original texts with the highest similarity are determined as target embedding vectors, so as to determine the original text corresponding to each target embedding vector as the target text corresponding to the embedding vector of the current prompt word.

7. The storage resource allocation method according to any one of claims 1 to 6, It is characterized in that Before binding each persistent volume in the container cluster orchestration management system with the corresponding current persistent volume declaration, the method further includes: Creating a storage class based on preset storage definition parameters; wherein the preset storage definition parameters include a network file system service IP, a network file system service path, a recycling mechanism, a volume binding mode, and a total capacity; Declare storage resource configuration information corresponding to each of the current persistent volumes and create each persistent volume in the container cluster orchestration management system using the storage class.

8. A storage resource allocation device, It is characterized in that include: A text acquisition module, used to acquire an embedding vector of a current prompt word related to a persistent volume declaration, and use a first target pre-trained model to search a preset knowledge base for a target text corresponding to the embedding vector of the current prompt word; wherein the preset knowledge base stores original texts and embedding vectors of the original texts, and the original texts are various texts in a preset operation-related manual for persistent volume declaration; A parameter acquisition module, used to concatenate the current prompt word, the target text, and a preset executable operation list to obtain a pending request, and use the first target pre-trained model to identify the pending request, and if the obtained recognition result indicates that the pending request is used to trigger the call of a target persistent volume declaration management function, then obtain the target parameters required by the target persistent volume declaration management function; A request management module, configured to execute the target persistent volume declaration management function based on the target parameter to complete the persistent volume declaration management operation corresponding to the current prompt word and obtain each current persistent volume declaration; The resource allocation module is used to bind each persistent volume in the container cluster orchestration management system to the corresponding current persistent volume declaration to allocate the persistent volume for the current persistent volume declaration.

9. An electronic device, It is characterized in that include: Memory for storing computer programs; A processor, configured to execute the computer program to implement the steps of the storage resource allocation method according to any one of claims 1 to 7.

10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the storage resource allocation method according to any one of claims 1 to 7 are implemented.

Images