Large model hierarchical cache loading method for inclusive computing power intelligent computing center

Abstract

The present invention provides a large-model layered cache loading method for a universal computing power intelligent computing center, which relates to the technical fields of intelligent computing centers, intelligent computing centers, and computing power infrastructure. The method includes: step S1: counting the number of times a model layer stored in a remote database is used, where the model layer is obtained by disassembling a large model; step S2: when the number of times a model layer is used exceeds a first preset threshold, copying the model layer from the remote database to a local first storage device; step S3: when the number of times a model layer is used exceeds a second preset threshold, determining whether the model layer is a public model layer or a private model layer; if the model layer is a public model layer, moving the model layer from the first storage device to a local second storage device, where the second preset threshold is greater than the first preset threshold, and the access speed of the second storage device is higher than the access speed of the first storage device. The present invention can save network traffic used for loading large models and improve loading speed.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent computing centers, smart computing centers and computing power infrastructure, and specifically to a large-model hierarchical cache loading method for inclusive computing power intelligent computing centers. Background Art

[0002] With the rapid development of artificial intelligence technology, "intelligent computing centers" and "intelligent computing centers" have emerged.

[0003] An "Intelligent Computing Center" is a facility that utilizes large-scale heterogeneous computing resources, including general-purpose and intelligent computing power, to provide the computing power, data, and algorithms required for AI applications (such as AI deep learning model development, model training, and model inference). The Intelligent Computing Center encompasses facilities, hardware, and software, providing a full stack of capabilities, from bottom-level computing power to top-level application enablement.

[0004] “Intelligent Computing Center” includes but is not limited to “Smart Computing Center”.

[0005] "Intelligent Computing Center" refers to an artificial intelligence computing center. It is a type of computing power infrastructure that is based on artificial intelligence theory, adopts artificial intelligence computing architecture, and provides computing power services, data services, and algorithm services required for artificial intelligence applications.

[0006] "Computing power" is the core of "intelligent computing center" and "intelligent computing center". It is the ability of computer equipment or computing / data center to process information. It is the ability of computer hardware and software to work together to perform certain computing needs. It is the computing power to achieve target result output by processing information data. It is a new type of productivity that integrates information computing power, network carrying capacity, and data storage capacity. It mainly provides services to society through computing power infrastructure.

[0007] The intelligent computing center can provide data storage capabilities for storing large models of users of tenant services. Currently, large models are usually stored in remote databases. When the large model needs to be read, it needs to be loaded from the remote database to the local computer. The large model can be as small as a few GB or as large as hundreds of GB. Each loading from the remote database will bring a large network traffic overhead. In addition, the loading speed will also be affected by factors such as network speed and computing power resource storage performance. That is, the existing large model loading method has a high economic cost and low loading efficiency, making it difficult to achieve widespread application of inclusive computing power. Summary of the Invention

[0008] The present invention provides a large-model hierarchical cache loading method for inclusive computing power intelligent computing centers, which is used to solve the problem that intelligent computing centers need to load large models from remote databases, which has high economic costs, low loading efficiency, and difficulty in achieving widespread application of inclusive computing power.

[0009] In order to solve the above-mentioned technical problems, the present invention is achieved as follows:

[0010] In a first aspect, the present invention provides a large model hierarchical cache loading method for an inclusive computing power intelligent computing center, comprising:

[0011] Step S1: Counting the number of times a model layer stored in a remote database is used in real time, wherein the model layer is obtained by disassembling a large model, and each large model stored in the remote database is disassembled into multiple model layers for storage;

[0012] Step S2: When the number of times the model layer is used exceeds a first preset threshold, the model layer is copied from the remote database to the first storage device of the local computing power node;

[0013] Step S3: When the number of times the model layer is used exceeds a second preset threshold, determine whether the model layer is a public model layer or a private model layer. If the model layer is a public model layer, move the model layer from the first storage device to the second storage device of the local computing power node, wherein the second preset threshold is greater than the first preset threshold, and the access speed of the second storage device is higher than the access speed of the first storage device. The public model layer refers to a model layer shared by multiple tenants, and the private model layer refers to a model layer that belongs to only one of the tenants.

[0014] Optionally, step S2 includes:

[0015] Step S21: When the number of times the model layer is used exceeds a first preset threshold, the model layer is copied from the remote database to the first storage device of all local computing nodes of the intelligent computing center; wherein, for the local computing node where the user who currently needs to read the model layer is located, the model layer is copied from the remote database to the first storage device of the local computing node; for other local computing nodes of the intelligent computing center, the usage ratio of the network bandwidth between the other local computing nodes and the remote database is monitored, and when the usage ratio of the network bandwidth is less than a preset bandwidth usage ratio threshold, the model layer is copied from the remote database to the first storage device of the other local computing nodes;

[0016] and / or

[0017] The step S3 comprises:

[0018] Step S31: When the number of times the model layer is used exceeds a second preset threshold, determine whether the model layer is a public model layer or a private model layer. If the model layer is a public model layer, for each local computing power node of the intelligent computing center, move the model layer from the first storage device of the local computing power node to the second storage device of the local computing power node.

[0019] Optionally, if the large model is a model obtained by fine-tuning the basic model, the large model includes two model layers, and the two model layers include: a basic model parameter layer and a fine-tuning parameter layer;

[0020] and / or

[0021] The large model includes multiple model layers based on model structure decomposition.

[0022] Optionally, the method further includes:

[0023] Step S4: Monitor the storage capacity of the second storage device in real time. If the storage capacity of the second storage device is higher than the first preset capacity ratio, compare the number of times each of the public model layers stored in the second storage device is used, and move the public model layer with the lowest number of times to the first storage device for storage.

[0024] Optionally, the method further includes:

[0025] Step S5: Monitor the storage capacity of the first storage device in real time. If the storage capacity of the first storage device is higher than the second preset capacity ratio, determine whether the private model layer is stored in the first storage device. If the private model layer is stored in the first storage device, compare the number of uses of each of the private model layers stored in the first storage device, and delete the private model layer with the lowest number of uses. If the private model layer is not stored in the first storage device, compare the number of uses of each of the public model layers stored in the first storage device, and delete the public model layer with the lowest number of uses.

[0026] Optionally, the method further includes:

[0027] Step S6: in response to the instruction of reading the large model triggered by the user, obtaining information of all model layers of the large model;

[0028] Step S7: judging whether the model layer is a public model layer or a private model layer according to the information of the model layer;

[0029] Step S8: If the model layer is a public model layer, query whether the model layer is stored in the second storage device; if the model layer is stored in the second storage device, load the model layer into the memory; if the model layer is not stored in the second storage device, query whether the model layer is stored in the first storage device; if the model layer is stored in the first storage device, load the model layer into the memory; if the model layer is not stored in the first storage device, pull the model layer from the remote database and save it to the first storage device, and load the model layer into the memory;

[0030] Step S9: If the model layer is a private model layer, query whether the model layer is stored in the first storage device; if the model layer is stored in the first storage device, load the model layer into the memory; if the model layer is not stored in the first storage device, pull the model layer from the remote database and save it to the first storage device, and load the model layer into the memory;

[0031] Step S10: All the models of the large model loaded into the memory are layered and merged to obtain the large model.

[0032] Optionally, the method further includes:

[0033] Step S11: Each time the model layer is used by a user of the tenant, the number of times the model layer is used is increased by 1.

[0034] Optionally, before step S1, the method further includes:

[0035] Step S12: in response to a user-triggered instruction to save the large model, the large model of the user is disassembled to obtain multiple model layers;

[0036] Step S13: Calculate the hash value of the content of each model layer;

[0037] Step S14: searching the index library for an index record corresponding to the hash value of the model layer; if so, not saving the model layer; if not, saving the model layer to the remote database, and generating an index record based on the hash value of the model layer and the storage path of the model layer and storing it in the index library;

[0038] Step S15: Record the index values ​​of all model layers of the large model.

[0039] Optionally, step S14 includes:

[0040] Step S141: using the hash value of the model layer as an index value, generating an index record according to the index value of the model layer and the storage path of the model layer, and storing the generated index record in the index library.

[0041] Optionally, step S14 further includes:

[0042] Step S142: If the index record corresponding to the hash value of the model layer does not exist in the index library, the model layer is saved as a private model layer in the private database of the remote database, and an index record corresponding to the hash value of the model layer is generated and stored in the index library, where the tenant to which the user belongs is recorded in the index record;

[0043] and / or

[0044] Step S143: If an index record corresponding to the hash value of the model layer exists in the index library, determine whether the tenant to which the user belongs is recorded in the index record;

[0045] Step S144: If the tenant to which the user belongs is not recorded in the index record, the number of tenant sharing times in the index record is increased by 1, and the tenant to which the user belongs is recorded in the index record; determining whether the model layer is stored in the private database; if the model layer is stored in the private database, moving the model layer as a public model layer to the public database of the remote database;

[0046] Step S145: If the tenant to which the user belongs has been recorded in the index record, there is no need to increase the number of tenant sharing times in the index record by 1.

[0047] In a second aspect, the present invention provides a large model hierarchical cache loading device for an inclusive computing power intelligent computing center, comprising:

[0048] a statistics module for counting in real time the number of times a model layer stored in a remote database is used, wherein the model layer is obtained by disassembling a large model, and each large model stored in the remote database is disassembled into multiple model layers for storage;

[0049] A first cache module is configured to copy the model layer from the remote database to a first storage device of a local computing power node when the number of times the model layer is used exceeds a first preset threshold;

[0050] The second cache module is used to determine whether the model layer is a public model layer or a private model layer when the number of times the model layer is used exceeds a second preset threshold; if the model layer is a public model layer, move the model layer from the first storage device to the second storage device of the local computing power node, wherein the second preset threshold is greater than the first preset threshold, the access speed of the second storage device is higher than the access speed of the first storage device, the public model layer refers to a model layer shared by multiple tenants, and the private model layer refers to a model layer belonging to only one of the tenants.

[0051] In a third aspect, the present invention provides a computing power device, comprising: a processor, a memory, and a program stored in the memory and runnable on the processor. When the program is executed by the processor, the steps of the large-model hierarchical cache loading method for the inclusive computing power intelligent computing center as described in the first aspect above are implemented.

[0052] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the large-model hierarchical cache loading method for the inclusive computing power intelligent computing center as described in the first aspect above are implemented.

[0053] In a fifth aspect, the present invention provides a computer program product comprising computer instructions, which, when executed by a processor, implement the steps of the large-model hierarchical cache loading method for the inclusive computing power intelligent computing center as described in the first aspect above.

[0054] In the present invention, the user's large model is disassembled to obtain multiple model layers, and the layers are stored in a remote database. The number of times the model layers stored in the remote database are used is counted in real time. According to the number of times the model layers are used, the model layers are cached in the storage device of the local computing power node. Therefore, when the model layers need to be loaded, if the model layers are stored in the storage device of the local computing power node, they can be directly loaded from the storage device of the local computing power node to the memory. This can effectively save the network traffic used for loading large models, and effectively improve the loading speed of large models, greatly reduce economic costs, and promote the widespread application of inclusive computing power. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0056] Figure 1This is a flow chart of a large-model hierarchical cache loading method for a universal computing power intelligent computing center according to the present invention;

[0057] Figure 2 This is a structural diagram of a large-model hierarchical cache loading device for a universal computing power intelligent computing center according to the present invention;

[0058] Figure 3 It is a structural diagram of the computing power device of the present invention. DETAILED DESCRIPTION

[0059] The following will clearly and completely describe the technical solutions of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0060] First, the technical terms involved in the present invention are briefly explained below.

[0061] The "computing power" mentioned in the present invention refers to: the ability of computer equipment or computing / data centers to process information, the ability of computer hardware and software to work together to execute certain computing requirements, and the computing power to achieve target result output by processing information data. It is a new type of productivity that integrates information computing power, network carrying capacity, and data storage capacity, and mainly provides services to society through computing power infrastructure.

[0062] The "computing power" (CP) mentioned in this invention refers to the ability of a data center server to process data and output results. It is a comprehensive indicator for measuring the computing power of a data center, including general computing power, supercomputing power, and intelligent computing power. The commonly used unit of measurement is the number of floating-point operations per second (FLOPS, 1EFLOPS=10^18 FLOPS). The larger the value, the stronger the comprehensive computing power. According to calculations, 1 EFLOPS is approximately the computing power output of 5 Tianhe-2A units, or 500,000 mainstream server CPUs, or 2 million mainstream laptops. The calculation formula is: CP=CP 通用 +CP 智能 +CP 超级 .

[0063] The "Network Power" (NP) mentioned in this invention refers to: it is a manifestation of the data transmission capability of computing power facilities, including comprehensive capabilities such as network architecture, network bandwidth, transmission latency, intelligent management and scheduling, etc. It involves network transmission within and between data centers, and is a comprehensive indicator for measuring network transmission scheduling capabilities.

[0064] "Storage Power" (SP) as used in this document refers to the comprehensive capabilities of a data center in four areas: data storage capacity, performance, security and reliability, and environmental friendliness. It serves as a comprehensive indicator for measuring a data center's data storage capabilities, encompassing both external storage devices like storage arrays and internal server storage. Storage capacity is commonly measured in exabytes (EB, 1EB = 2^60 bytes), while performance is commonly measured in IOPS / TB (Input / Output Operations Per Second). Disaster recovery ratio is a key indicator of security and reliability.

[0065] The "computing power infrastructure" mentioned in the present invention refers to a new type of information infrastructure that integrates information computing power, network carrying capacity, and data storage capacity, and can realize the centralized calculation, storage, transmission and application of information.

[0066] The "new information infrastructure" mentioned in the present invention refers to: mainly including network infrastructure such as 5G networks, fiber-optic broadband networks, backbone networks, international communication networks, satellite Internet, computing power infrastructure such as data centers, general computing power centers, intelligent computing centers, supercomputing centers, and new technology facilities such as artificial intelligence, blockchain, and quantum computing.

[0067] The "computing power" mentioned in the present invention includes: general computing power, intelligent computing power and super computing power.

[0068] The "general computing power" mentioned in the present invention refers to the computing power provided by servers based on CPU (Central Processing Unit) chips, which is used to support basic general computing such as cloud computing and edge computing.

[0069] The "intelligent computing power" mentioned in this invention refers to: a computing platform based on specialized chips such as GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array), and ASIC (Application Specific Integrated Circuit) for various innovative artificial intelligence applications, such as natural language processing and machine vision.

[0070] The "supercomputing power" mentioned in the present invention refers to the computing power provided by high-performance computing clusters such as supercomputers. It utilizes the centralized computing resources of multiple computer systems working in parallel and uses a dedicated operating system to handle extremely complex or data-intensive problems. It is mainly used for calculations in cutting-edge scientific fields, such as planetary simulation, drug molecule design, genetic analysis, etc.

[0071] The term "intelligent computing center" as used in this document refers to a facility that utilizes large-scale heterogeneous computing resources, including general-purpose computing power (CPUs) and intelligent computing power (GPUs, FPGAs, ASICs, etc.), primarily to provide the computing power, data, and algorithms required for AI applications (such as AI deep learning model development, model training, and model inference). An intelligent computing center encompasses facilities, hardware, and software, providing a full stack of capabilities, from bottom-level computing power to top-level application enablement.

[0072] The "intelligent computing center" mentioned in the present invention includes but is not limited to the "intelligent computing center".

[0073] The "intelligent computing center" mentioned in the present invention is an artificial intelligence computing center, which is a type of computing power infrastructure based on artificial intelligence theory, adopts artificial intelligence computing architecture, and provides computing power services, data services and algorithm services required for artificial intelligence applications.

[0074] The "computing power center" mentioned in the present invention refers to: a facility that is mainly composed of infrastructure such as wind, fire, water, electricity, and IT hardware and software equipment, and has computing power, transportation capacity, and storage capacity, including general data centers, intelligent computing centers, supercomputing centers, etc.

[0075] The "supercomputing center" mentioned in the present invention refers to: a supercomputing data center, which is a data center based on a supercomputer or a large-scale computing cluster, which can provide large-scale computing, storage and network services and other functions, and is widely used in application scenarios such as aerospace, national defense, oil exploration, climate modeling and genome sequencing.

[0076] The "computing resources" mentioned in the present invention refer to: technologies and facilities with information computing, transmission, storage and application capabilities required for the development of a digital society, including but not limited to computing resources such as CPUs and GPUs, network resources such as switches and routers, storage resources such as storage arrays and distributed storage, security resources such as firewalls and intrusion detection systems, and supporting and guarantee resources such as wind, fire, water and electricity.

[0077] The "big model" mentioned in the present invention includes but is not limited to a "big language model" and a "big multimodal model".

[0078] The "large language model" mentioned in this invention refers to a large language model (LLM), which is a language model with a large parameter scale. It is designed to understand and generate human language. It is trained with a large amount of text data and can perform a wide range of tasks including text summarization, translation, sentiment analysis, etc.

[0079] The "Multimodal Large Models" mentioned in this invention refer to models that combine multimodal information such as text, images, video, and audio for training, including but not limited to multimodal large language models.

[0080] The "inclusive computing power" mentioned in this invention refers to providing appropriate and effective computing power services at an affordable cost to all social classes and groups that have computing power service needs based on the requirements of equal opportunity and the principle of commercial sustainability.

[0081] To solve the problem that existing intelligent computing centers need to load large models from remote databases, which is very costly, has low loading efficiency, and is difficult to achieve widespread application of inclusive computing power, please refer to Figure 1 The present invention provides a large model hierarchical cache loading method for a universal computing power intelligent computing center, the method comprising:

[0082] Step S1: Counting the number of times a model layer stored in a remote database is used in real time, wherein the model layer is obtained by disassembling a large model, and each large model stored in the remote database is disassembled into multiple model layers for storage;

[0083] Step S2: When the number of times the model layer is used exceeds a first preset threshold, the model layer is copied from the remote database to the first storage device of the local computing power node;

[0084] Step S3: When the number of times the model layer is used exceeds a second preset threshold, determine whether the model layer is a public model layer or a private model layer. If the model layer is a public model layer, move the model layer from the first storage device to the second storage device of the local computing power node, wherein the second preset threshold is greater than the first preset threshold, and the access speed of the second storage device is higher than the access speed of the first storage device. The public model layer refers to a model layer shared by multiple tenants, and the private model layer refers to a model layer that belongs to only one of the tenants.

[0085] It should be noted that the tenant in the present invention refers to the tenant who rents the computing power resource storage space of the intelligent computing center. There can be multiple tenants, and the user is the user served by the tenant. The user can store the large model in the computing power resource storage space of the tenant serving him, and one tenant can serve multiple users.

[0086] In the present invention, the user's large model is disassembled to obtain multiple model layers, and the layers are stored in a remote database. The number of times the model layers stored in the remote database are used is counted in real time. According to the number of times the model layers are used, the model layers are cached in the storage device of the local computing power node. Therefore, when the model layers need to be loaded, if the model layers are stored in the storage device of the local computing power node, they can be directly loaded from the storage device of the local computing power node to the memory. This can effectively save the network traffic used for loading large models, and effectively improve the loading speed of large models, greatly reduce economic costs, and promote the widespread application of inclusive computing power.

[0087] Please refer to Table 1, which shows the cache rules of the model layer of the present invention.

[0088] Table 1

[0089] serial number Model Layering Hardware computing resources Whether to store locally 1 Public high-frequency use layer High-speed storage (nvme) yes 2 Public low-frequency use layer Ordinary storage (disk) yes 3 Private high-frequency usage layer Ordinary storage (disk) yes 4 Private low-frequency use layer Remote Database no

[0090] In Table 1, the public high-frequency usage layer refers to a public model layer whose usage count exceeds the second preset threshold. This type of model layer can be cached to the second storage device of the local computing power node. The second storage device is a local high-speed storage device, such as a solid-state drive (NVMe).

[0091] The public low-frequency usage layer refers to the public model layer whose usage times exceed the first preset threshold, which can be stored in the first storage device of the local computing power node. The first storage device is a local ordinary storage device, such as a disk.

[0092] The private high-frequency usage layer refers to a private model layer whose usage exceeds a first preset threshold, which can be stored in the first storage device of the local computing power node. The first storage device is a local ordinary storage device, such as a disk.

[0093] Private low-frequency usage refers to private model layers whose usage times are lower than a first preset threshold, and is stored in a remote database.

[0094] In addition, the public model layers whose usage times are lower than the first preset threshold are also stored in the remote database.

[0095] In the present invention, optionally, step S2 includes:

[0096] Step S21: When the number of times the model layer is used exceeds a first preset threshold, the model layer is copied from the remote database to the first storage device of all local computing nodes of the intelligent computing center; wherein, for the local computing node where the user who currently needs to read the model layer is located, the model layer is copied from the remote database to the first storage device of the local computing node; for other local computing nodes of the intelligent computing center, the usage ratio of the network bandwidth between the other local computing nodes and the remote database is monitored, and when the usage ratio of the network bandwidth is less than the preset bandwidth usage ratio threshold, the model layer is copied from the remote database to the first storage device of the other local computing nodes.

[0097] That is to say, for the local computing power node where the user who currently needs to read the model layer is located, since the model layer needs to be used immediately, the model layer needs to be copied from the remote database to the first storage device of the local computing power node immediately. For other local computing power nodes, the model layer can be copied from the remote database to the first storage device of the other local computing power nodes when the network is idle.

[0098] The preset bandwidth usage ratio threshold, such as 30%, can be flexibly set as needed.

[0099] In the present invention, optionally, step S3 includes:

[0100] Step S31: When the number of times the model layer is used exceeds a second preset threshold, determine whether the model layer is a public model layer or a private model layer. If the model layer is a public model layer, for each local computing power node of the intelligent computing center, move the model layer from the first storage device of the local computing power node to the second storage device of the local computing power node.

[0101] In the present invention, since the model layers are cached on all local computing nodes, when the user needs to read the model layers, they can load them from any local computing node, thereby effectively improving the loading speed of large models.

[0102] In the present invention, optionally, if the large model is a model obtained by fine-tuning the basic model, the large model includes two model layers, and the two model layers include: a basic model parameter layer and a fine-tuning parameter layer.

[0103] The basic model refers to a pre-trained large model, such as GPT or BERT.

[0104] Fine-tuning a large model refers to the technical process of performing secondary training on a pre-trained large model using data from specific fields or tasks to adapt the model to specific application scenarios.

[0105] In the present invention, the large model can be a large model using LoRA fine-tuning (Low-Rank Adaptation). LoRA fine-tuning is an efficient model fine-tuning technology. Its core idea is to reduce the number of parameters required for fine-tuning by introducing a low-rank matrix based on the pre-trained model, thereby improving training efficiency and avoiding overfitting.

[0106] In the present invention, optionally, the large model includes multiple model layers based on model structure decomposition.

[0107] For example, assuming the large model is a Transformer model architecture, the large model can be divided into two model layers: the encoding component and the decoding component. Or, more finely, it can be divided into the self-attention layer and feedforward network of the encoding component, and the self-attention layer, attention layer and feedforward network of the decoding component, for a total of five model layers.

[0108] In the present invention, optionally, the large model hierarchical cache loading method for the inclusive computing power intelligent computing center further includes:

[0109] Step S4: Monitor the storage capacity of the second storage device in real time. If the storage capacity of the second storage device is higher than the first preset capacity ratio, compare the number of times each of the public model layers stored in the second storage device is used, and move the public model layer with the lowest number of times to the first storage device for storage.

[0110] The first preset capacity ratio may be, for example, 70% or 80%, etc., and may be flexibly set according to specific circumstances.

[0111] In the present invention, optionally, the large model hierarchical cache loading method for the inclusive computing power intelligent computing center further includes:

[0112] Step S5: Monitor the storage capacity of the first storage device in real time. If the storage capacity of the first storage device is higher than the second preset capacity ratio, determine whether the private model layer is stored in the first storage device. If the private model layer is stored in the first storage device, compare the number of uses of each of the private model layers stored in the first storage device, and delete the private model layer with the lowest number of uses. If the private model layer is not stored in the first storage device, compare the number of uses of each of the public model layers stored in the first storage device, and delete the public model layer with the lowest number of uses.

[0113] The second preset capacity ratio may be, for example, 70% or 80%, etc., and may be flexibly set according to specific circumstances.

[0114] In the present invention, by real-time monitoring of the storage capacity of the local storage device, the local storage device can be cleaned up in time, thereby improving the utilization rate of the local storage device.

[0115] In the present invention, optionally, the large model hierarchical cache loading method for the inclusive computing power intelligent computing center further includes:

[0116] Step S6: in response to the instruction of reading the large model triggered by the user, obtaining information of all model layers of the large model;

[0117] The model layer information may include, for example, information that can uniquely identify the model layer, such as an index value or a name of the model layer. The model layer information may also include the storage address or type of the model layer. Based on the storage address or type of the model layer, it can be determined whether the model layer is a public model layer or a private model layer.

[0118] Step S7: judging whether the model layer is a public model layer or a private model layer according to the information of the model layer;

[0119] Step S8: If the model layer is a public model layer, query whether the model layer is stored in the second storage device; if the model layer is stored in the second storage device, load the model layer into the memory; if the model layer is not stored in the second storage device, query whether the model layer is stored in the first storage device; if the model layer is stored in the first storage device, load the model layer into the memory; if the model layer is not stored in the first storage device, pull the model layer from the remote database and save it to the first storage device, and load the model layer into the memory;

[0120] Step S9: If the model layer is a private model layer, query whether the model layer is stored in the first storage device; if the model layer is stored in the first storage device, load the model layer into the memory; if the model layer is not stored in the first storage device, pull the model layer from the remote database and save it to the first storage device, and load the model layer into the memory;

[0121] Step S10: All the models of the large model loaded into the memory are layered and merged to obtain the large model.

[0122] In the present invention, optionally, the large model hierarchical cache loading method for the inclusive computing power intelligent computing center further includes:

[0123] Step S11: Each time the model layer is used by a user of the tenant, the number of times the model layer is used is increased by 1.

[0124] In the present invention, optionally, the large model hierarchical cache loading method for the inclusive computing power intelligent computing center further includes:

[0125] Step S12: in response to a user-triggered instruction to save the large model, the large model of the user is disassembled to obtain multiple model layers;

[0126] Step S13: Calculate the hash value of the content of each model layer;

[0127] In the present invention, a hash algorithm is used to calculate the hash value of the content of the model layer. A hash algorithm is a function that maps input data of any length to an output of a fixed length. The hash algorithm has the following characteristics:

[0128] Determinism: The same input will always produce the same output, ensuring that the results are verifiable.

[0129] Irreversibility: The original data cannot be derived from the hash value, ensuring security.

[0130] Collision resistance: The probability of different inputs producing the same output is extremely low, reducing the risk of data tampering.

[0131] Avalanche effect: Small changes in input lead to significant differences in output, increasing sensitivity.

[0132] Efficiency: Fast computing, suitable for large-scale data or real-time scenarios.

[0133] Step S14: searching the index library for an index record corresponding to the hash value of the model layer; if so, not saving the model layer; if not, saving the model layer to the remote database, and generating an index record based on the hash value of the model layer and the storage path of the model layer and storing it in the index library;

[0134] In the present invention, the index library can store multiple index records, and each index record includes a hash value of a model layer and a storage path of the model layer.

[0135] For example, the index records stored in the index library can be as follows:

[0136] index1={Hash1, path1};

[0137] index2={Hash2, path2};

[0138] …

[0139] Among them, index1={Hash1, path1} is an index record, and index1 is the index value of the index record.

[0140] In the present invention, optionally, the hash value may be used as the index value, and the index records stored in the index library may be as follows:

[0141] index1={path1};

[0142] index2={path2};

[0143] …

[0144] In the present invention, it should be noted that when the hash value is not used as the index value but the index value is set separately, each hash value uniquely corresponds to one index value.

[0145] In the present invention, optionally, when saving the model layer, it can be stored in a file format. Optionally, the index record can also include the file size of the model layer.

[0146] In the present invention, optionally, the index record may further include the tenant sharing count (or tenant reference count) of the model layer, where the tenant sharing count is used to indicate how many tenants share the model layer.

[0147] For example, the index records stored in the index library can be as follows:

[0148] index1={path1,size1,count1};

[0149] index2={path2,size2,count2};

[0150] …

[0151] In the present invention, optionally, the index record may also include the number of times the model layer is used (also called the number of times the model layer is used). The number of times the model layer is used is used to indicate how many users use the model layer. The number of times the model layer is used in the present invention may include two types of times. The first is how many users store it, and the second is how many users read it. The number of times mentioned in the above embodiment refers to how many users read it.

[0152] In the present invention, when the index record includes file size, tenant sharing times and / or usage times, an index record is generated based on the hash value of the model layer, the storage path of the model layer, and the file size of the model layer, tenant sharing times and / or usage times, and the generated index record is stored in the index library.

[0153] Step S15: Record the index values ​​of all model layers of the large model.

[0154] Optional, the index value of the model layer is the hash value of the model layer.

[0155] For example, a large model can be recorded as follows: Model1: {index1, index2}, which means that Model1 is split into two model layers, where the index values ​​of the two model layers are index1 and index2 respectively.

[0156] Optionally, while recording the index values ​​of all model layers of the large model, the storage paths of the model layers and / or the file sizes of the model layers may also be recorded.

[0157] For example, a large model might record the following:

[0158] Model 1

[0159] { index1= / c: / publish / model_level1,50G

[0160] index2= / c: / private / model_level2,1G

[0161] }

[0162] Among them, / c: / publish / model_level1 is the storage path of the model layer, and 50G is the file size of the model layer.

[0163] In the present invention, when a user needs to store a large model in an intelligent computing center, the user's large model is disassembled to obtain multiple model layers, the hash value of the content of each model layer is calculated, and the index library is searched to see whether there is an index record corresponding to the hash value of the model layer. If so, it means that the content of the model layer has been stored before, and there is no need to store it repeatedly. If not, it means that the content of the model layer has not been stored before, and the model layer is stored, and an index record is generated according to the hash value of the model layer and the storage path of the model layer and stored in the index library. When the content of the model layer of the large model has been stored before, there is no need to store it repeatedly, that is, there is no need to store the large model in full, thereby effectively saving storage space, reducing the storage and network traffic waste of computing resources, optimizing the computing resources of the intelligent computing center, greatly reducing the waste of computing resources, greatly reducing economic costs, and promoting the widespread application of inclusive computing power.

[0164] In the present invention, optionally, step S14 includes:

[0165] Step S141: using the hash value of the model layer as an index value, generating an index record according to the index value of the model layer and the storage path of the model layer, and storing the generated index record in the index library.

[0166] By directly using the hash value of the model layer as the index value, there is no need to introduce the index value separately, and the retrieval of the index library is more convenient.

[0167] In the present invention, optionally, step S14 further includes:

[0168] Step S142: If the index record corresponding to the hash value of the model layer does not exist in the index library, the model layer is saved as a private model layer in the private database of the remote database, and an index record corresponding to the hash value of the model layer is generated and stored in the index library, where the tenant to which the user belongs is recorded in the index record;

[0169] and / or

[0170] Step S143: If an index record corresponding to the hash value of the model layer exists in the index library, determine whether the tenant to which the user belongs is recorded in the index record;

[0171] Step S144: If the tenant to which the user belongs is not recorded in the index record, the number of tenant sharing times in the index record is increased by 1, and the tenant to which the user belongs is recorded in the index record; determining whether the model layer is stored in the private database; if the model layer is stored in the private database, moving the model layer as a public model layer to the public database of the remote database;

[0172] Step S145: If the tenant to which the user belongs has been recorded in the index record, there is no need to increase the number of tenant sharing times in the index record by 1.

[0173] In the present invention, in the above determination process, if it is determined that the tenant record to which the user belongs is in the index record, there is no need to increase the tenant sharing times in the index record by 1.

[0174] In the present invention, in the above-mentioned determination process, if it is determined that the model is hierarchically stored in a public database, no further storage processing is required.

[0175] In the present invention, optionally, each tenant has its own private database, and the public database is shared by multiple tenants.

[0176] In the present invention, the intelligent computing center provides a private database and a public database. If there is no index record corresponding to the hash value of a certain model layer in the index library, it means that the model layer has not been stored before, and the model layer can be stored in the private database.

[0177] If there is an index record corresponding to the hash value of a model layer in the index library, it means that the model layer has been stored before. At this time, it is also necessary to determine whether the tenant who currently needs to store the model layer is recorded in the index record. If the tenant has been recorded in the index record, there is no need to increase the number of tenant sharing times of the index record.

[0178] If the tenant is not recorded in the index record, the tenant needs to be recorded in the index record and the number of tenant sharing times of the index record is increased by 1. Then, it is necessary to determine whether the model layer is currently saved in the private database. If so, the model layer needs to be moved to the public database. If the model layer is currently saved in the public database, it does not need to be saved again.

[0179] In the present invention, optionally, the large model hierarchical cache loading method for the inclusive computing power intelligent computing center further includes:

[0180] Step S16: determining the storage fee for each tenant based on the model type, tenant sharing times and / or file size of the stored model layer;

[0181] Wherein, for the model layer shared by multiple tenants, the storage fee of the model layer is evenly shared by the multiple tenants according to the number of times the model layer is shared by the tenants; or, for the model layer shared by multiple tenants, if the model layer is a base model, the tenants are not charged the storage fee of the base model;

[0182] and / or

[0183] For the tenant-private model layer, the storage fee of the model layer is calculated according to the file size of the model layer.

[0184] In the present invention, optionally, the tenant-private model layer refers to a model layer stored in a private database.

[0185] In the present invention, the large model storage platform provided by the intelligent computing center can have multiple tenants, each tenant serves multiple users, and the storage fee of the large model stored by each tenant can be calculated.

[0186] In some cases, it is not excluded to calculate the storage fee for each user.

[0187] In the present invention, optionally, step S15 includes:

[0188] Step S151: Using the hash value of the model layer as an index value, the index values ​​of all model layers of the large model are recorded.

[0189] In the present invention, each large model may have a unique identifier, such as Model1, and the intelligent computing center may separately provide an identifier library for recording index values ​​of all model layers of the large model.

[0190] For example, the records corresponding to a large model are as follows:

[0191] Model 1

[0192] {index1= / c: / publish / model_level1,50G

[0193] index2= / c: / private / model_level2,1G

[0194] }

[0195] Please refer to Figure 2 The present invention also provides a large model hierarchical cache loading device 100 for a universal computing power intelligent computing center, comprising:

[0196] A statistics module 101 is used to count the number of times a model layer stored in a remote database is used in real time, wherein the model layer is obtained by decomposing a large model, and each large model stored in the remote database is decomposed into multiple model layers for storage;

[0197] A first cache module 102 is configured to copy the model layer from the remote database to a first storage device of a local computing node when the number of times the model layer is used exceeds a first preset threshold;

[0198] The second cache module 103 is used to determine whether the model layer is a public model layer or a private model layer when the number of times the model layer is used exceeds a second preset threshold. If the model layer is a public model layer, the model layer is moved from the first storage device to the second storage device of the local computing power node, wherein the second preset threshold is greater than the first preset threshold, the access speed of the second storage device is higher than the access speed of the first storage device, the public model layer refers to a model layer shared by multiple tenants, and the private model layer refers to a model layer that belongs to only one of the tenants.

[0199] In the present invention, the user's large model is disassembled to obtain multiple model layers, and the layers are stored in a remote database. The number of times the model layers stored in the remote database are used is counted in real time. According to the number of times the model layers are used, the model layers are cached in the storage device of the local computing power node. Therefore, when the model layers need to be loaded, if the model layers are stored in the storage device of the local computing power node, they can be directly loaded from the storage device of the local computing power node to the memory. This can effectively save the network traffic used for loading large models, and effectively improve the loading speed of large models, greatly reduce economic costs, and promote the widespread application of inclusive computing power.

[0200] In the present invention, optionally, the first cache module 102 is used to copy the model layer from the remote database to the first storage device of all local computing nodes of the intelligent computing center when the number of times the model layer is used exceeds a first preset threshold, wherein, for the local computing node where the user who currently needs to read the model layer is located, the model layer is copied from the remote database to the first storage device of the local computing node; for other local computing nodes of the intelligent computing center, the network bandwidth usage ratio between the other local computing nodes and the remote database is monitored, and when the network bandwidth usage ratio is less than the preset bandwidth usage ratio threshold, the model layer is copied from the remote database to the first storage device of the other local computing nodes.

[0201] In the present invention, optionally, the second cache module 103 is used to determine whether the model layer is a public model layer or a private model layer when the number of times the model layer is used exceeds a second preset threshold. If the model layer is a public model layer, for each local computing power node of the intelligent computing center, the model layer is moved from the first storage device of the local computing power node to the second storage device of the local computing power node.

[0202] In the present invention, optionally, if the large model is a model obtained by fine-tuning the basic model, the large model includes two model layers, and the two model layers include: a basic model parameter layer and a fine-tuning parameter layer.

[0203] In the present invention, optionally, the large model includes multiple model layers based on model structure decomposition.

[0204] In the present invention, optionally, the large model hierarchical cache loading device 100 for the inclusive computing power intelligent computing center further includes:

[0205] The first monitoring module is used to monitor the storage capacity of the second storage device in real time. If the storage capacity of the second storage device is higher than the first preset capacity ratio, the number of times each of the public model layers stored in the second storage device is used is compared, and the public model layer with the lowest number of times is moved to the first storage device for storage.

[0206] In the present invention, optionally, the large model hierarchical cache loading device 100 for the inclusive computing power intelligent computing center further includes:

[0207] The second monitoring module is used to monitor the storage capacity of the first storage device in real time. If the storage capacity of the first storage device is higher than the second preset capacity ratio, it is determined whether the private model layer is stored in the first storage device. If the private model layer is stored in the first storage device, the number of times each of the private model layers stored in the first storage device is used is compared, and the private model layer with the lowest number of uses is deleted. If the private model layer is not stored in the first storage device, the number of times each of the public model layers stored in the first storage device is compared, and the public model layer with the lowest number of uses is deleted.

[0208] In the present invention, optionally, the large model hierarchical cache loading device 100 for the inclusive computing power intelligent computing center further includes:

[0209] an acquisition module, configured to acquire information of all model layers of the large model in response to a user-triggered instruction to read the large model;

[0210] A judgment module, configured to judge whether the model layer is a public model layer or a private model layer according to the information of the model layer;

[0211] a first query module configured to, if the model layer is a public model layer, query whether the model layer is stored in the second storage device; if the model layer is stored in the second storage device, load the model layer into the memory; if the model layer is not stored in the second storage device, query whether the model layer is stored in the first storage device; if the model layer is stored in the first storage device, load the model layer into the memory; if the model layer is not stored in the first storage device, pull the model layer from the remote database, save it to the first storage device, and load the model layer into the memory;

[0212] a second query module configured to, if the model layer is a private model layer, query whether the model layer is stored in the first storage device; if the model layer is stored in the first storage device, load the model layer into the memory; if the model layer is not stored in the first storage device, pull the model layer from the remote database and save it to the first storage device, and load the model layer into the memory;

[0213] A merging module is used to merge all the models of the large model loaded into the memory in layers to obtain the large model.

[0214] In the present invention, optionally, the large model hierarchical cache loading device 100 for the inclusive computing power intelligent computing center further includes:

[0215] The counting module is configured to increase the number of times the model layer is used by 1 each time the model layer is used by a user of the tenant.

[0216] In the present invention, optionally, the large model hierarchical cache loading device 100 for the inclusive computing power intelligent computing center further includes:

[0217] a disassembly module, configured to disassemble the user's large model in response to a user-triggered instruction to save the large model, to obtain a plurality of model layers;

[0218] A hash value calculation module, used to calculate the hash value of the content of each model layer;

[0219] A storage module is used to search the index library for an index record corresponding to the hash value of the model layer, and if so, not save the model layer; if not, save the model layer to the remote database, and generate an index record based on the hash value of the model layer and the storage path of the model layer and store it in the index library;

[0220] The recording module is used to record the index values ​​of all model layers of the large model.

[0221] Optionally, the storage module is configured to use the hash value of the model layer as an index value, generate an index record according to the index value of the model layer and the storage path of the model layer, and store the generated index record in the index library.

[0222] Optionally, the storage module is used to save the model layer as a private model layer in the private database of the remote database if the index record corresponding to the hash value of the model layer does not exist in the index library, generate an index record corresponding to the hash value of the model layer and store it in the index library, and the index record records the tenant to which the user belongs.

[0223] Optionally, the storage module is used to determine whether the tenant to which the user belongs is recorded in the index record if there is an index record corresponding to the hash value of the model layer in the index library; if the tenant to which the user belongs is not recorded in the index record, increase the number of tenant sharing times in the index record by 1, and record the tenant to which the user belongs in the index record; determine whether the model layer is saved in a private database, and if the model layer is saved in a private database, move the model layer as a public model layer to the public database of the remote database; if the tenant to which the user belongs is already recorded in the index record, there is no need to increase the number of tenant sharing times in the index record by 1.

[0224] Please refer to Figure 3 The present invention also provides a computing power device 200, including a processor 201, a memory 202, and a computer program stored in the memory 202 and executable on the processor 201. When the computer program is executed by the processor 201, the various processes of the above-mentioned large-model hierarchical cache loading method embodiment for the inclusive computing power intelligent computing center are implemented, and the same technical effects can be achieved. To avoid repetition, they will not be described here.

[0225] The present invention also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the various processes of the aforementioned embodiment of the large-model hierarchical cache loading method for the Inclusive Computing Power Intelligent Computing Center, achieving the same technical effects. To avoid repetition, the description is omitted here. The computer-readable storage medium may be, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0226] The present invention also provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the above Figure 1The various processes of the embodiment of the large-model hierarchical cache loading method for the inclusive computing power intelligent computing center are shown, and can achieve the same technical effect. To avoid repetition, they will not be repeated here.

[0227] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0228] Through the above description of the embodiments, those skilled in the art will clearly understand that the methods of the above embodiments can be implemented using software plus the necessary general-purpose hardware platform. Of course, hardware can also be used, but in many cases the former is the more preferred implementation method. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, or optical disk) and includes a number of instructions for enabling a terminal (such as a mobile phone, computer, server, air conditioner, or network device) to execute the methods described in the various embodiments of the present invention.

[0229] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are protected by the present invention.

Claims

1. A large model hierarchical cache loading method for the universal computing power intelligent computing center, characterized in that: include: Step S1: Counting the number of times a model layer stored in a remote database is used in real time, wherein the model layer is obtained by disassembling a large model, and each large model stored in the remote database is disassembled into multiple model layers for storage; Step S2: When the number of times the model layer is used exceeds a first preset threshold, the model layer is copied from the remote database to the first storage device of the local computing power node; Step S3: When the number of times the model layer is used exceeds a second preset threshold, determine whether the model layer is a public model layer or a private model layer. If the model layer is a public model layer, move the model layer from the first storage device to the second storage device of the local computing power node, wherein the second preset threshold is greater than the first preset threshold, and the access speed of the second storage device is higher than the access speed of the first storage device. The public model layer refers to a model layer shared by multiple tenants, and the private model layer refers to a model layer that belongs to only one of the tenants.

2. The method according to claim 1, wherein: The step S2 comprises: Step S21: When the number of times the model layer is used exceeds a first preset threshold, the model layer is copied from the remote database to the first storage device of all local computing nodes of the intelligent computing center; wherein, for the local computing node where the user who currently needs to read the model layer is located, the model layer is copied from the remote database to the first storage device of the local computing node; for other local computing nodes of the intelligent computing center, the usage ratio of the network bandwidth between the other local computing nodes and the remote database is monitored, and when the usage ratio of the network bandwidth is less than a preset bandwidth usage ratio threshold, the model layer is copied from the remote database to the first storage device of the other local computing nodes; And / or, step S3 includes: Step S31: When the number of times the model layer is used exceeds a second preset threshold, determine whether the model layer is a public model layer or a private model layer. If the model layer is a public model layer, for each local computing power node of the intelligent computing center, move the model layer from the first storage device of the local computing power node to the second storage device of the local computing power node.

3. The method according to claim 1, characterized in that If the large model is a model obtained by fine-tuning the basic model, the large model includes two model layers, and the two model layers include: a basic model parameter layer and a fine-tuning parameter layer; And / or, the large model includes multiple model layers based on model structure decomposition.

4. The method according to claim 1, wherein Also includes: Step S4: Monitor the storage capacity of the second storage device in real time. If the storage capacity of the second storage device is higher than the first preset capacity ratio, compare the number of times each of the public model layers stored in the second storage device is used, and move the public model layer with the lowest number of times to the first storage device for storage.

5. The method according to claim 1, wherein Also includes: Step S5: Monitor the storage capacity of the first storage device in real time. If the storage capacity of the first storage device is higher than the second preset capacity ratio, determine whether the private model layer is stored in the first storage device. If the private model layer is stored in the first storage device, compare the number of uses of each of the private model layers stored in the first storage device, and delete the private model layer with the lowest number of uses. If the private model layer is not stored in the first storage device, compare the number of uses of each of the public model layers stored in the first storage device, and delete the public model layer with the lowest number of uses.

6. The method according to claim 1, characterized in that Also includes: Step S6: in response to the instruction of reading the large model triggered by the user, obtaining information of all model layers of the large model; Step S7: judging whether the model layer is a public model layer or a private model layer according to the information of the model layer; Step S8: If the model layer is a public model layer, query whether the model layer is stored in the second storage device; if the model layer is stored in the second storage device, load the model layer into the memory; if the model layer is not stored in the second storage device, query whether the model layer is stored in the first storage device; If the model layer is stored in the first storage device, the model layer is loaded into the memory; if the model layer is not stored in the first storage device, the model layer is pulled from the remote database and saved to the first storage device, and the model layer is loaded into the memory; Step S9: If the model layer is a private model layer, query whether the model layer is stored in the first storage device; If the model layer is stored in the first storage device, the model layer is loaded into the memory; if the model layer is not stored in the first storage device, the model layer is pulled from the remote database and saved to the first storage device, and the model layer is loaded into the memory; Step S10: All the models of the large model loaded into the memory are layered and merged to obtain the large model.

7. The method according to claim 6, characterized in that Also includes: Step S11: Each time the model layer is used by a user of the tenant, the number of times the model layer is used is increased by 1.

8. The method according to claim 1, characterized in that Before step S1, the method further includes: Step S12: in response to a user-triggered instruction to save the large model, the large model of the user is disassembled to obtain multiple model layers; Step S13: Calculate the hash value of the content of each model layer; Step S14: searching the index library for an index record corresponding to the hash value of the model layer; if so, not saving the model layer; if not, saving the model layer to the remote database, and generating an index record based on the hash value of the model layer and the storage path of the model layer and storing it in the index library; Step S15: Record the index values ​​of all model layers of the large model.

9. The method according to claim 8, characterized in that The step S14 includes: Step S141: using the hash value of the model layer as an index value, generating an index record according to the index value of the model layer and the storage path of the model layer, and storing the generated index record in the index library.

10. The method according to claim 8, characterized in that The step S14 further includes: Step S142: If the index record corresponding to the hash value of the model layer does not exist in the index library, the model layer is saved as a private model layer in the private database of the remote database, and an index record corresponding to the hash value of the model layer is generated and stored in the index library, where the tenant to which the user belongs is recorded in the index record; and / or Step S143: If an index record corresponding to the hash value of the model layer exists in the index library, determine whether the tenant to which the user belongs is recorded in the index record; Step S144: If the tenant to which the user belongs is not recorded in the index record, the number of tenant sharing times in the index record is increased by 1, and the tenant to which the user belongs is recorded in the index record; determining whether the model layer is stored in the private database; if the model layer is stored in the private database, moving the model layer as a public model layer to the public database of the remote database; Step S145: If the tenant to which the user belongs has been recorded in the index record, there is no need to increase the number of tenant sharing times in the index record by 1.

Images