Cloud-edge collaborative AI computing power scheduling method and device for broadcast multi-layer station, and medium

Abstract

The application discloses a cloud-edge collaborative AI computing power scheduling method for a multi-layer station of a radio and television, equipment and a medium, and the method comprises the following steps: constructing an AI capability hub platform cooperated by a central cloud and multi-layer edge nodes; collecting cloud-edge computing power resource information, uniformly abstracting and pooling the information to form a standardized computing power pool; establishing a three-dimensional multi-tenant topology structure of stations, businesses and applications, implementing three-layer isolation of networks, storages and processes, and configuring tenant resource strategies; receiving a business-side AI model calling request, and completing protocol normalization, authentication, flow limiting, routing and monitoring by a unified model proxy gateway; collecting cloud-edge real-time load information, and combining the load, tenant topology and business priority to perform dynamic allocation and elastic scheduling on the computing power pool. The scheme improves the utilization rate of computing power, guarantees the isolation safety of multi-tenants, is compatible with AI models of multiple manufacturers, realizes elastic scheduling of computing power, and meets the requirements of radio and television security broadcasting and signal creation compliance.

Description

Technical Field

[0001] This invention relates to the fields of smart broadcasting platform technology and broadcasting security technology, and in particular to a cloud-edge collaborative AI computing power scheduling method, equipment and medium for multi-level broadcasting stations. Background Technology

[0002] Currently, radio and television stations generally adopt a decentralized AI capability deployment model, which has many technical pain points and seriously restricts the digital and intelligent upgrading process of the radio and television system.

[0003] First, computing resources are scattered, with provincial, municipal, and county-level stations managing their resources independently without a unified abstraction and pooling mechanism. This results in low utilization of computing resources, with some nodes idle and others overloaded, making efficient collaborative scheduling of computing power across the entire network impossible. Second, the multi-tenant management mechanism is incomplete, failing to form a three-dimensional collaborative tenant topology structure involving stations, services, and applications. Furthermore, it lacks strong isolation measures at the network, storage, and process levels, making it prone to resource contention and data leakage among multiple tenants, which is difficult to meet the rigid control requirements for safe broadcasting. Third, AI models from different vendors use custom proprietary communication protocols, making model access processes complex. The lack of a unified proxy gateway for protocol normalization and centralized management leads to high system integration costs and operational difficulties, making it impossible to achieve compatible access and efficient invocation of AI models from multiple vendors.

[0004] Furthermore, the lack of intelligence and dynamism in computing power scheduling makes it impossible to perceive load changes in the central cloud and edge nodes in real time, making it difficult to achieve precise allocation and elastic scaling of computing power in conjunction with business priorities, thus failing to guarantee the stable operation of core businesses. At the same time, existing technologies do not fully consider the limited bandwidth of broadcast network private networks, the need for data not to leave the domain, and the requirements for domestic adaptation. The model operation has high computing power consumption and high transmission pressure, and lacks a full-process security encryption and compliance audit mechanism, which cannot meet the standard requirements for broadcast network innovation construction and safe broadcasting. Summary of the Invention

[0005] This invention provides a cloud-edge collaborative AI computing power scheduling method, equipment, and medium for multi-level broadcasting stations, in order to achieve centralized management of computing power across the entire network, improve computing power utilization, ensure multi-tenant security isolation, be compatible with AI models from multiple vendors, and realize dynamic and elastic scheduling of computing power, thereby meeting the requirements of broadcasting security and information technology innovation standards.

[0006] According to one aspect of the present invention, a cloud-edge collaborative AI computing power scheduling method for multi-layer broadcasting stations is provided, comprising: Build a central AI capability hub platform that coordinates central cloud and multi-layer edge nodes; Collect computing power resource information from the central cloud and the multi-layer edge nodes, and perform unified abstraction and pooling processing on the computing power resource information to form a standardized computing power pool; Establish a three-dimensional multi-tenant topology structure corresponding to stations, services, and applications; perform network, storage, and process three-layer isolation on the three-dimensional multi-tenant topology structure; and configure tenant resource usage policies. Receive AI model call requests from the business side, and perform protocol normalization, unified authentication, rate limiting, routing and monitoring on the AI ​​model call requests through a unified model proxy gateway; The system collects real-time load information of the central cloud and the multi-layer edge nodes. Based on the real-time load information, the three-dimensional multi-tenant topology, and business priorities, it performs dynamic allocation and elastic scheduling of computing power in the standardized computing power pool.

[0007] According to another aspect of the present invention, a cloud-edge collaborative AI computing power scheduling device for multi-level broadcasting stations is provided, comprising: The platform building module is used to build an AI capability hub platform that coordinates the central cloud and multi-layer edge nodes; The computing power pooling module is used to collect computing power resource information from the central cloud and the multi-layer edge nodes, and to perform unified abstraction and pooling processing on the computing power resource information to form a standardized computing power pool. The tenant management module is used to establish a three-dimensional multi-tenant topology for stations, services, and applications, perform network, storage, and process isolation on the three-dimensional multi-tenant topology, and configure tenant resource usage policies. The request control module is used to receive AI model call requests from the business side and perform protocol normalization, unified authentication, rate limiting, routing and monitoring on the AI ​​model call requests through a unified model proxy gateway. The computing power scheduling module is used to collect real-time load information of the central cloud and the multi-layer edge nodes, and to perform dynamic allocation and elastic scheduling of computing power for the standardized computing power pool based on the real-time load information, the three-dimensional multi-tenant topology and business priorities.

[0008] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores a computer program that can be executed by the at least one processor, which is then executed by the at least one processor to enable the at least one processor to execute the cloud-edge collaborative AI computing power scheduling method for multi-layer broadcasting stations as described in any embodiment of the present invention.

[0009] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions, the computer instructions being configured to cause a processor to execute and implement the cloud-edge collaborative AI computing power scheduling method for multi-layer broadcasting stations as described in any embodiment of the present invention.

[0010] The technical solution of this invention constructs an AI capability hub platform that coordinates a central cloud and multiple edge nodes; collects computing resource information from the central cloud and multiple edge nodes, performs unified abstraction and pooling processing on the computing resource information to form a standardized computing pool; establishes a three-dimensional multi-tenant topology corresponding to stations, services, and applications, performs network, storage, and process three-layer isolation on the three-dimensional multi-tenant topology, and configures tenant resource usage policies; receives AI model call requests from the service side, and performs protocol normalization, unified authentication, rate limiting, routing, and monitoring on the AI ​​model call requests through a unified model proxy gateway; and collects data from the central cloud and multiple edge nodes. Based on real-time load information, a three-dimensional multi-tenant topology, and business priorities, the system dynamically allocates and elastically schedules computing power in a standardized computing pool. Further optimization measures, such as lightweight model adaptation, hierarchical computing power management, end-to-end security encryption, and compliance auditing, address existing technical issues in broadcasting stations, including dispersed computing power, low utilization, insufficient multi-tenant isolation, difficulty in accessing multi-vendor models, unintelligent computing power scheduling, and poor security compliance. This achieves centralized management of network-wide computing power, improves computing power utilization, ensures secure multi-tenant isolation, is compatible with multi-vendor AI models, enables dynamic and elastic computing power scheduling, and meets broadcasting security and information technology innovation standards.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart of a cloud-edge collaborative AI computing power scheduling method for multi-level broadcasting stations provided in this embodiment of the invention; Figure 2 A flowchart of another cloud-edge collaborative AI computing power scheduling method for multi-layered broadcasting stations provided in an embodiment of the present invention; Figure 3This is a schematic diagram of the structure of a cloud-edge collaborative AI computing power scheduling device for multi-level broadcasting stations provided in an embodiment of the present invention; Figure 4 A schematic diagram of the structure of an electronic device for implementing the cloud-edge collaborative AI computing power scheduling method for multi-level broadcasting stations according to an embodiment of the present invention. Detailed Implementation

[0014] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0015] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0016] Figure 1 This is a flowchart illustrating a cloud-edge collaborative AI computing power scheduling method for multi-level broadcasting stations, provided by an embodiment of the present invention. This embodiment is applicable to situations involving unified access to AI capabilities, elastic allocation of computing resources, and multi-tenant security isolation for multi-level broadcasting stations. The method can be executed by a cloud-edge collaborative AI computing power scheduling device for multi-level broadcasting stations. This device can be implemented in hardware and / or software and can be configured in an electronic device. Figure 1 As shown, the method specifically includes the following steps: S110, Constructing an AI capability hub platform that coordinates central cloud and multi-layer edge nodes.

[0017] The central cloud refers to a centralized computing power cluster deployed at provincial-level broadcasting nodes, used for overall coordination and scheduling. Multi-layered edge nodes can be understood as distributed execution nodes deployed at city, county, and grassroots stations. The AI ​​capability hub platform is the unified carrier for cloud-edge collaboration, used to integrate computing power, manage models, and support business operations.

[0018] Specifically, based on the provincial-municipal-county three-tiered broadcasting architecture, a central cloud is built at the provincial node, regional edge nodes are deployed at the municipal node, and lightweight edge nodes are deployed at the county-level stations. A private network communication link is established between the central cloud and the edge nodes to form a unified and collaborative AI capability hub platform.

[0019] S120. Collect computing power resource information of the central cloud and the multi-layer edge nodes, and perform unified abstraction and pooling processing on the computing power resource information to form a standardized computing power pool.

[0020] Here, computing resource information refers to the schedulable resource information of a node, such as CPU, memory, GPU, and inference accelerator cards. Unified abstraction is the process of converting hardware resources from different manufacturers and with different specifications into a unified scheduling unit. A standardized computing pool can be understood as a unified resource set formed by integrating the computing power of the entire network.

[0021] Specifically, the central cloud and edge nodes can report their computing power capacity, load status, and other information in real time. The platform shields the differences in underlying hardware, abstracting all types of resources into allocable computing power units, thereby integrating the distributed computing power across the entire network into a standardized, elastically scalable computing power pool. For example, a provincial-level central cloud has large-scale inference computing power, a city-level edge node has medium-scale computing power, and a county-level node has lightweight computing power. The platform abstracts these three types of resources and incorporates them into the same computing power pool for management. This eliminates resource fragmentation and improves computing power utilization and scheduling flexibility.

[0022] In some possible implementations, the step of collecting computing resource information from the central cloud and the multi-layer edge nodes, and performing unified abstraction and pooling processing on the computing resource information to form a standardized computing pool includes: reporting node load and resource capacity information through the edge proxy module; and performing unified management of the entire network's computing resources through the central cloud's global computing scheduling module to form a standardized computing pool that can be elastically scaled.

[0023] The edge proxy module is deployed on edge nodes and is used to collect and report resource information. The global computing power scheduling module is the module in the central cloud responsible for the overall coordination of network resources.

[0024] Specifically, the edge proxy module periodically collects the load, capacity, and health status of local nodes and reports them to the central cloud via a dedicated network. The global computing power scheduling module receives the information reported from the entire network, performs unified management of all nodes, and maintains the hierarchical computing power relationship between the central, branch, and grassroots stations.

[0025] In some possible implementations, the unified management of the entire network's computing resources through the central cloud global computing power scheduling module includes: maintaining the hierarchical computing power association between the center, branch centers, and grassroots stations; allocating resource quotas to the standardized computing power pool based on the tenant resource usage strategy; and performing computing power allocation and recycling on edge nodes in response to scheduling instructions.

[0026] The hierarchical computing power relationship refers to the hierarchical correspondence between the central cloud, branch centers, and grassroots stations, which enables hierarchical control and resource linkage. Resource quotas can be understood as the amount of usable computing power allocated from the standardized computing power pool for different tenants and different businesses. Computing power allocation and recycling can be the platform's regulatory actions of distributing computing power resources to edge nodes or reclaiming idle resources based on changes in business load.

[0027] Specifically, the global computing power scheduling module can organize the affiliation and scheduling links between provincial centers, municipal sub-centers, and county-level grassroots stations, forming a fixed hierarchical computing power relationship. Thus, based on the configured tenant resource usage strategies, independent resource quotas are allocated within the overall standardized computing power pool according to business importance and station affiliation. After receiving scheduling instructions from the upper layer, the platform issues computing power expansion allocation instructions to the target edge nodes, and automatically performs surplus computing power recovery operations when business load decreases and resources are idle.

[0028] For example, the provincial center can be set as the highest dispatch level, and the municipal sub-center can manage multiple county-level stations. Dedicated computing power quotas can be allocated according to two types of services: safe broadcasting and intelligent duty. During peak business hours, additional computing power can be allocated to county-level edge nodes, and idle computing power can be recovered during off-peak business hours.

[0029] S130. Establish a three-dimensional multi-tenant topology structure corresponding to stations, services, and applications, perform network, storage, and process three-layer isolation on the three-dimensional multi-tenant topology structure, and configure tenant resource usage policies.

[0030] In some embodiments, the establishment of a three-dimensional multi-tenant topology corresponding to stations, services, and applications, the implementation of network, storage, and process three-layer isolation on the three-dimensional multi-tenant topology, and the configuration of tenant resource usage policies include: dividing tenant units according to station level, service type, and application scenario; allocating independent network domains, independent storage volumes, and independent operating environments to the tenant units; setting tenant computing power guarantee priorities based on service level, and restricting cross-tenant resource preemption.

[0031] The three-dimensional multi-tenant topology refers to the tenant organization structure built according to three dimensions: station level, service type, and application scenario. The three layers of isolation refer to network isolation, storage isolation, and process isolation, respectively.

[0032] Specifically, the platform can be hierarchically divided by provincial, municipal, and county-level broadcasting stations; its business types can be categorized according to secure broadcasting, intelligent monitoring, and operation and maintenance knowledge services; and its application scenarios can be divided according to inspection and identification, voice broadcasting, and fault response, thereby constructing a three-dimensional tenant structure. The platform allocates independent network domains, independent storage volumes, and independent process runtime environments to different tenants, achieving mutual isolation. This ensures that tenants do not interfere with each other, data security, and meets the requirements for secure broadcasting.

[0033] S140. Receive AI model call requests from the business side, and perform protocol normalization, unified authentication, rate limiting, routing, and monitoring on the AI ​​model call requests through a unified model proxy gateway.

[0034] Among these, AI model invocation requests are requests initiated by business systems to utilize AI capabilities. A unified model proxy gateway can be understood as an intermediary service that unifies access to and manages models from multiple vendors. Protocol normalization is the process of converting different vendors' proprietary protocols into a unified standard format.

[0035] Specifically, the platform can receive call requests from business sides such as secure broadcasting, intelligent monitoring, and knowledge services. Through a unified model proxy gateway, the platform performs protocol normalization, authentication, rate limiting, routing, and monitoring of the requests to ensure that the requests are legal, orderly, and traceable.

[0036] S150. Collect real-time load information of the central cloud and the multi-layer edge nodes. Based on the real-time load information, the three-dimensional multi-tenant topology and business priority, perform dynamic allocation and elastic scheduling of computing power in the standardized computing power pool.

[0037] Among them, real-time load information is the node's current resource usage and operating load data.

[0038] Specifically, the platform can continuously collect real-time load data from the central cloud and edge nodes, and dynamically allocate computing power pools based on tenant topology and business priorities. It increases computing power supply when the load increases and reclaims idle computing power when the load decreases, ensuring the supply of resources for high-priority businesses and improving computing power utilization efficiency.

[0039] The technical solution of this invention constructs an AI capability hub platform that coordinates a central cloud and multiple edge nodes; collects computing resource information from the central cloud and multiple edge nodes, performs unified abstraction and pooling processing on the computing resource information to form a standardized computing pool; establishes a three-dimensional multi-tenant topology corresponding to stations, services, and applications, performs network, storage, and process three-layer isolation on the three-dimensional multi-tenant topology, and configures tenant resource usage policies; receives AI model call requests from the service side, and performs protocol normalization, unified authentication, rate limiting, routing, and monitoring on the AI ​​model call requests through a unified model proxy gateway; and collects data from the central cloud and multiple edge nodes. Based on real-time load information, a three-dimensional multi-tenant topology, and business priorities, the system dynamically allocates and elastically schedules computing power in a standardized computing pool. Further optimization measures, such as lightweight model adaptation, hierarchical computing power management, end-to-end security encryption, and compliance auditing, address existing technical issues in broadcasting stations, including dispersed computing power, low utilization, insufficient multi-tenant isolation, difficulty in accessing multi-vendor models, unintelligent computing power scheduling, and poor security compliance. This achieves centralized management of network-wide computing power, improves computing power utilization, ensures secure multi-tenant isolation, is compatible with multi-vendor AI models, enables dynamic and elastic computing power scheduling, and meets broadcasting security and information technology innovation standards.

[0040] Figure 2 This is a flowchart illustrating another cloud-edge collaborative AI computing power scheduling method for multi-layered broadcast television stations, provided as an embodiment of the present invention. This embodiment is an optimization of the above embodiment. Technical terms that are the same as or corresponding to those in the above embodiment will not be repeated here. Figure 2 As shown, the method specifically includes the following steps: S210: Construct an AI capability hub platform that coordinates central cloud and multi-layer edge nodes.

[0041] S220. Collect the computing power resource information of the central cloud and the multi-layer edge nodes, and perform unified abstraction and pooling processing on the computing power resource information to form a standardized computing power pool.

[0042] S230. Establish a three-dimensional multi-tenant topology structure corresponding to stations, services, and applications, perform network, storage, and process three-layer isolation on the three-dimensional multi-tenant topology structure, and configure tenant resource usage policies.

[0043] S240. Receive AI model call requests from the business side, and perform protocol normalization, unified authentication, rate limiting, routing, and monitoring on the AI ​​model call requests through a unified model proxy gateway.

[0044] In some possible implementations, the step of performing protocol normalization on the AI ​​model call request through a unified model proxy gateway includes: converting AI model call requests with multiple vendor proprietary protocol formats into a standard communication protocol format; and performing unified format encapsulation on the AI ​​model call request and response results.

[0045] Private protocol formats refer to communication formats defined by different AI vendors. Standard communication protocols, on the other hand, can be universal interaction formats adopted by the system.

[0046] Specifically, the platform can receive call requests from different vendor models, unify the conversion of various protocols such as REST, WebSocket, and private gRPC into the standard gRPC protocol, and perform unified format encapsulation of request parameters and return results. This allows upper-layer businesses to be free from concern about the differences in underlying protocols, significantly reducing the cost of accessing multiple vendor models and improving system compatibility.

[0047] For example, vendor A's ASR model uses a proprietary HTTP protocol, and vendor B's CV model uses a proprietary WebSocket protocol. The platform uniformly converts them to the standard gRPC format before providing services to the outside world.

[0048] In some possible implementations, the method further includes: performing lightweight adaptation processing on the model file supporting the AI ​​model call request; deploying the model backbone network in the central cloud and deploying the model lightweight head network in the edge nodes.

[0049] Lightweight adaptation is an optimization process that reduces model size and computational power consumption. The backbone network can be understood as the main part of the model responsible for core feature calculation and inference.

[0050] Specifically, the platform performs INT8 quantization and structured pruning on the model to compress its size and reduce computational overhead. The platform deploys the bulky, computationally intensive backbone network in the central cloud and the lightweight head network on edge nodes. Edge nodes perform only lightweight inference, and the original data remains within their domain. For example, the backbone feature network of a video analytics model is deployed in the provincial central cloud, while edge nodes only run the detection head, enabling rapid local inference. This approach adapts to scenarios with limited bandwidth on broadcast networks, reducing transmission pressure and ensuring data security.

[0051] S250. Collect real-time load information of the central cloud and the multi-layer edge nodes, perform load prediction based on the historical operating data of the nodes, and generate a computing power scheduling plan.

[0052] Load forecasting refers to predicting load trends over a future period based on historical operating indicators.

[0053] Specifically, the platform collects historical time-series data such as node CPU, memory, and load rate, calculates the load trend for future periods through a predictive model, generates computing power scheduling plans in advance, and completes expansion preparations before the load increases, thereby realizing the transformation from passive scheduling to proactive pre-scheduling and improving system response speed and stability.

[0054] S260. If the node load exceeds the set range, perform computing power expansion or computing power reduction processing.

[0055] Specifically, when the node load continues to increase, the platform automatically expands the computing power for the corresponding business; when the load continues to decrease, the platform automatically reclaims idle computing power, so that the system always maintains a reasonable load level.

[0056] In some possible implementations, the method further includes: performing transmission encryption and storage encryption on the dynamic allocation of computing power, the elastic scheduling, and the processing of AI model invocation requests; and performing log recording and compliance auditing on related operational behaviors.

[0057] Among these, transmission encryption and storage encryption refer to the use of national cryptographic algorithms to protect data. Compliance auditing refers to the complete recording of operations, calls, and scheduling behaviors.

[0058] Specifically, the platform can use national cryptographic algorithms SM2, SM3, and SM4 to encrypt and protect model calls, computing power scheduling, data transmission, and storage. Furthermore, it logs all operational behaviors, scheduling records, and call records, supporting security auditing and accountability, meeting the requirements of broadcasting innovation and secure broadcasting. It ensures data remains within the domain and is fully auditable, meeting industry compliance and self-control requirements.

[0059] The technical solution of this invention constructs an AI capability hub platform that coordinates a central cloud and multiple edge nodes; collects computing resource information from the central cloud and multiple edge nodes, performs unified abstraction and pooling processing on the computing resource information to form a standardized computing pool; establishes a three-dimensional multi-tenant topology corresponding to stations, services, and applications, performs network, storage, and process three-layer isolation on the three-dimensional multi-tenant topology, and configures tenant resource usage policies; receives AI model call requests from the service side, and performs protocol normalization, unified authentication, rate limiting, routing, and monitoring on the AI ​​model call requests through a unified model proxy gateway; and collects data from the central cloud and multiple edge nodes. Based on real-time load information, a three-dimensional multi-tenant topology, and business priorities, the system dynamically allocates and elastically schedules computing power in a standardized computing pool. Further optimization measures, such as lightweight model adaptation, hierarchical computing power management, end-to-end security encryption, and compliance auditing, address existing technical issues in broadcasting stations, including dispersed computing power, low utilization, insufficient multi-tenant isolation, difficulty in accessing multi-vendor models, unintelligent computing power scheduling, and poor security compliance. This achieves centralized management of network-wide computing power, improves computing power utilization, ensures secure multi-tenant isolation, is compatible with multi-vendor AI models, enables dynamic and elastic computing power scheduling, and meets broadcasting security and information technology innovation standards.

[0060] Figure 3 This is a schematic diagram of a cloud-edge collaborative AI computing power scheduling device for multi-level broadcasting stations, provided as an embodiment of the present invention. Figure 3 As shown, the device includes: Platform building module 310 is used to build an AI capability hub platform that coordinates the central cloud and multi-layer edge nodes; The computing power pooling module 320 is used to collect computing power resource information of the central cloud and the multi-layer edge nodes, and to perform unified abstraction and pooling processing on the computing power resource information to form a standardized computing power pool. The tenant management module 330 is used to establish a three-dimensional multi-tenant topology structure corresponding to stations, services, and applications, perform network, storage, and process three-layer isolation on the three-dimensional multi-tenant topology structure, and configure tenant resource usage policies. The request control module 340 is used to receive AI model call requests from the business side and perform protocol normalization, unified authentication, rate limiting, routing and monitoring on the AI ​​model call requests through the unified model proxy gateway. The computing power scheduling module 350 is used to collect real-time load information of the central cloud and the multi-layer edge nodes, and to perform dynamic allocation and elastic scheduling of computing power for the standardized computing power pool based on the real-time load information, the three-dimensional multi-tenant topology and business priority.

[0061] In some possible implementations, the computing power pooling module 320 includes: The information reporting submodule is used to report node load and resource capacity information through the edge proxy module; The global management submodule is used to perform unified management of the entire network's computing resources through the central cloud's global computing power scheduling module, forming a standardized computing power pool that can be elastically scaled.

[0062] In some possible implementations, the global sizing submodule includes: The hierarchical association unit is used to maintain the hierarchical computing power association relationship between the center, branch centers, and grassroots stations; A quota allocation unit is used to allocate resource quotas to the standardized computing power pool based on the tenant resource usage strategy. The scheduling execution unit is used to respond to scheduling instructions to perform computing power allocation and computing power recovery on edge nodes.

[0063] In some possible implementations, the tenant management module 330 includes: The tenant segmentation submodule is used to divide tenant units according to station level, business type, and application scenario. The resource isolation submodule is used to allocate independent network domains, independent storage volumes, and independent operating environments to the tenant units. The priority configuration submodule is used to set the priority of tenant computing power protection based on business level and restrict cross-tenant resource preemption.

[0064] In some possible implementations, the request control module 340 includes: The protocol conversion submodule is used to convert AI model call requests in multiple vendor proprietary protocol formats into standard communication protocol formats; The format encapsulation submodule is used to perform unified format encapsulation on the AI ​​model call requests and response results.

[0065] In some possible implementations, the request control module 340 further includes: The model adaptation submodule is used to perform lightweight adaptation processing on the model files that support the AI ​​model call request; The deployment and execution submodule is used to deploy the model backbone network in the central cloud and the model lightweight head network in the edge nodes.

[0066] In some possible implementations, the computing power scheduling module 350 includes: The load prediction submodule is used to perform load prediction based on the historical operating data of nodes and generate computing power scheduling plans; The elastic scaling submodule is used to perform computing power scaling up or down when the node load exceeds a set range.

[0067] In some possible implementations, the device further includes: The security audit module is used to perform transmission encryption and storage encryption on the dynamic allocation of computing power, the elastic scheduling, and the processing of AI model call requests, and to perform log recording and compliance auditing on related operations. The cloud-edge collaborative AI computing power scheduling device for multi-level broadcasting stations provided in this embodiment of the invention can execute the cloud-edge collaborative AI computing power scheduling method for multi-level broadcasting stations provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0068] Figure 4 This is a schematic diagram of the structure of an electronic device for implementing the cloud-edge collaborative AI computing power scheduling method for multi-layer broadcasting stations according to embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptops, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0069] like Figure 3 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0070] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0071] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 executes the various methods and processes described above, such as the cloud-edge collaborative AI computing power scheduling method for multi-layer broadcasting stations.

[0072] In some embodiments, the cloud-edge collaborative AI computing power scheduling method for multi-layered broadcasting stations can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the cloud-edge collaborative AI computing power scheduling method for multi-layered broadcasting stations described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to execute the cloud-edge collaborative AI computing power scheduling method for multi-layered broadcasting stations by any other suitable means (e.g., by means of firmware).

[0073] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific integrated circuits (ASICs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0074] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0075] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0076] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a cathode ray tube (CRT) or liquid crystal display (LCD) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0077] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0078] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and Virtual Private Server (VPS) services, such as high management difficulty and weak business scalability.

[0079] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0080] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A cloud-edge collaborative AI computing power scheduling method for multi-layered broadcasting stations, characterized in that, include: Build a central AI capability hub platform that coordinates central cloud and multi-layer edge nodes; Collect computing power resource information from the central cloud and the multi-layer edge nodes, and perform unified abstraction and pooling processing on the computing power resource information to form a standardized computing power pool; Establish a three-dimensional multi-tenant topology structure corresponding to stations, services, and applications; perform network, storage, and process three-layer isolation on the three-dimensional multi-tenant topology structure; and configure tenant resource usage policies. Receive AI model call requests from the business side, and perform protocol normalization, unified authentication, rate limiting, routing and monitoring on the AI ​​model call requests through a unified model proxy gateway; The system collects real-time load information of the central cloud and the multi-layer edge nodes. Based on the real-time load information, the three-dimensional multi-tenant topology, and business priorities, it performs dynamic allocation and elastic scheduling of computing power in the standardized computing power pool.

2. The method according to claim 1, characterized in that, The process of collecting computing resource information from the central cloud and the multi-layer edge nodes, and then performing unified abstraction and pooling of this computing resource information to form a standardized computing pool includes: The edge proxy module reports node load and resource capacity information. The central cloud global computing power scheduling module performs unified management of the entire network's computing power resources, forming a standardized computing power pool that can be elastically scaled.

3. The method according to claim 2, characterized in that, The unified management of network computing resources through the central cloud global computing power scheduling module includes: Maintain the hierarchical computing power relationship between the central station, branch stations, and grassroots stations; Resource quotas are allocated to the standardized computing power pool based on the tenant resource usage strategy. Responding to scheduling commands, perform computing power allocation and recycling on edge nodes.

4. The method according to claim 1, characterized in that, The establishment of a three-dimensional multi-tenant topology corresponding to stations, services, and applications, the implementation of network, storage, and process three-layer isolation on the three-dimensional multi-tenant topology, and the configuration of tenant resource usage policies include: Tenant units are divided according to station level, business type, and application scenario; Each tenant unit is allocated an independent network domain, an independent storage volume, and an independent operating environment. Tenant computing power protection priorities are set based on business level to restrict cross-tenant resource preemption.

5. The method according to claim 1, characterized in that, The step of normalizing the AI ​​model call request through a unified model proxy gateway includes: Convert AI model call requests using multiple vendors' proprietary protocol formats into standard communication protocol formats; The AI ​​model call requests and response results are encapsulated in a unified format.

6. The method according to claim 5, characterized in that, The method further includes: Lightweight adaptation processing is performed on the model files that support the AI ​​model call requests; The model backbone network is deployed in the central cloud, and the model lightweight head network is deployed on the edge nodes.

7. The method according to claim 1, characterized in that, The process of dynamically allocating and elastically scheduling computing power to the standardized computing power pool based on the real-time load information, the three-dimensional multi-tenant topology, and service priorities includes: Based on the historical operating data of nodes, perform load prediction and generate computing power scheduling plans; If the node load exceeds the set range, perform computing power expansion or reduction processing.

8. The method according to claim 1, characterized in that, The method further includes: The processes of dynamic allocation of computing power, elastic scheduling, and processing of AI model invocation requests are subject to transmission encryption and storage encryption. Log and compliance audit of relevant operations.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores a computer program that can be executed by the at least one processor, which is then executed by the at least one processor to enable the at least one processor to execute the cloud-edge collaborative AI computing power scheduling method for multi-layer broadcasting stations as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are used to cause a processor to execute the cloud-edge collaborative AI computing power scheduling method for multi-level broadcasting stations as described in any one of claims 1-8.

Images