AI control system based on deterministic network and calculation task calculation method

Abstract

The embodiment of the invention provides an AI control system based on a deterministic network and a calculation task calculation method, the system comprises a calculation network controller, an AI controller and a network controller, and the calculation network controller is in communication connection with the AI controller and the network controller; the network controller is used for providing deterministic network information and intelligent service routing information for the computing network controller, and the AI controller is used for transmitting a network scheduling forwarding model obtained through AI analysis to the computing network controller. And the computing network controller is used for determining an AI network path and an intelligent computing center which are matched with the computing task based on the deterministic network information, the intelligent service routing information, the network scheduling forwarding model and the service information of the computing task, and transmitting the computing task to the optimal intelligent computing center for computing through the optimal AI network path. According to the embodiment of the invention, two technical systems of network control and AI control are organically fused, and efficient calculation and output of various AI tasks can be completed.

Description

Technical Field

[0001] The present application relates to the field of deterministic computing network technology, and in particular to an AI control system and computing task calculation method based on a deterministic network. Background Art

[0002] With the rapid development of technologies such as cloud computing and edge computing, artificial intelligence (AI) has moved from theoretical research to widespread practical applications. The popularization of AI technology has not only profoundly changed people's lifestyles, but also promoted the digital transformation of various industries. In this context, the deep integration and network application of AI technology has become a new development trend. By deeply integrating AI technology with network architecture, more efficient and intelligent data processing, transmission and decision-making capabilities can be achieved, thus providing unprecedented innovation opportunities for all walks of life.

[0003] However, in the traditional technical system, the network management and control system and the AI ​​control system are independent of each other and have no integration, which cannot meet the requirements of efficient calculation and output of various AI tasks. Therefore, how to achieve the organic integration of the two technical systems of network control and AI control, and realize the synergy of AI and network from the user's perspective, so as to complete the efficient calculation and output of various AI tasks, is an urgent problem to be solved. Summary of the invention

[0004] The embodiments of the present application provide an AI control system and a computing task calculation method based on a deterministic network, which can realize the organic integration of the two technical systems of network control and AI control, realize the coordination of AI and network from the user's perspective, and thus complete the efficient calculation and output of various AI tasks.

[0005] The embodiment of the present application provides an AI control system based on a deterministic network, including a computing network controller, an AI controller and a deterministic network;

[0006] The deterministic network includes a network controller, and the computing network controller is communicatively connected with the AI ​​controller and the network controller respectively;

[0007] The network controller is used to provide deterministic network information and intelligent business routing information to the computing network controller, and the AI ​​controller is used to pass the network scheduling and forwarding model obtained by AI analysis to the computing network controller through an API interface. The computing network controller is used to determine the AI ​​network path and intelligent computing center that are suitable for the computing task based on the deterministic network information, intelligent business routing information, network scheduling and forwarding model, and business information of the computing task, and transmit the computing task to the optimal intelligent computing center through the optimal AI network path for calculation.

[0008] In the AI ​​control system based on the deterministic network described in the embodiment of the present application, a data collection and analysis system is also included, and the data collection and analysis system is communicatively connected with the AI ​​controller;

[0009] The data collection and analysis system is used to collect deterministic network topology information, whole network performance information, whole network IP routing information and network equipment power consumption information in real time, and use machine learning to analyze the collected information in real time to identify the business characteristics of various types of data, and send the identified business characteristics of various types of data to the AI ​​controller through an API interface for further analysis;

[0010] The AI ​​controller generates the network scheduling forwarding model by analyzing the service characteristics through AI.

[0011] In the AI ​​control system based on a deterministic network described in the embodiment of the present application, the data collection and analysis system is communicatively connected to the computing network controller and the network controller respectively;

[0012] The computing network controller and the network controller provide the data collection and analysis system with network status information of the intelligent computing center, computing power resource status information of the intelligent computing center, deterministic network performance status information, and deterministic network path status information through an API interface;

[0013] The data collection and analysis system is used to perform real-time analysis on the information provided by the computing network controller and the network controller by means of machine learning to identify the business characteristics of various types of data, and send the identified business characteristics of various types of data to the AI ​​controller through an API interface for further analysis;

[0014] The AI ​​controller generates the network scheduling forwarding model by analyzing the service characteristics through AI.

[0015] In the AI ​​control system based on deterministic network described in the embodiment of the present application, the AI ​​controller is also used to provide AI identification information, AI model information, AI version information, AI training time information, AI training maturity information and network security information of AI automatically evaluated to the data collection and analysis system through an API interface, and the data collection and analysis system is also used to provide characteristic data flow information with AI services in the network data flow and known AI application information to the AI ​​controller through an API interface.

[0016] In the deterministic network-based AI control system described in the embodiment of the present application, the data collection and analysis system collects the deterministic network topology information, the entire network performance information, the entire network IP routing information and the network equipment power consumption information in real time through BGP-LS or SNMP or Telemetry or NetFlow or port mirroring.

[0017] In the AI ​​control system based on the deterministic network described in the embodiment of the present application, the deterministic network information, intelligent service routing information and the network scheduling forwarding model are stored in a pre-set computing network control scheduling database, and the computing network controller can obtain them from the computing network control scheduling database.

[0018] In the deterministic network-based AI control system described in an embodiment of the present application, the service information includes service type and service SLA.

[0019] In the AI ​​control system based on the deterministic network described in the embodiment of the present application, the network controller and the network elements in the deterministic network exchange the whole network topology and network performance information through SNMP or Telemetry or BGP_LS or iFIT flow detection technology.

[0020] The embodiment of the present application further provides a computing task calculation method, which is used in the AI ​​control system based on the deterministic network described in any of the above embodiments, and the method includes:

[0021] Receive deterministic network information and intelligent service routing information provided by the network controller, as well as the network scheduling and forwarding model delivered by the AI ​​controller through the API interface;

[0022] Determine an AI network path and intelligent computing center adapted to the computing task based on the deterministic network information, intelligent business routing information, network scheduling and forwarding model, and business information of the computing task;

[0023] The computing task is transmitted to the optimal intelligent computing center through the optimal AI network path for calculation.

[0024] In the computing task calculation method described in the embodiment of the present application, if there are multiple AI network paths and intelligent computing centers adapted to the computing task, before transmitting the computing task through the optimal AI network path to the optimal intelligent computing center for calculation, it also includes:

[0025] Determine the optimal AI network path and the optimal intelligent computing center from multiple adapted AI network paths and intelligent computing centers.

[0026] The AI ​​control system based on a deterministic network provided in the embodiment of the present application includes a computing network controller, an AI controller and a deterministic network; the deterministic network includes a network controller, and the computing network controller is communicated with the AI ​​controller and the network controller respectively; the network controller is used to provide deterministic network information and intelligent business routing information to the computing network controller, and the AI ​​controller is used to pass the network scheduling and forwarding model obtained by AI analysis to the computing network controller through an API interface, and the computing network controller is used to determine the AI ​​network path and intelligent computing center adapted to the computing task based on the deterministic network information, intelligent business routing information, network scheduling and forwarding model, and business information of the computing task, and transmit the computing task to the optimal intelligent computing center through the optimal AI network path for computing, thereby realizing the organic integration of the two technical systems of network control and AI control, realizing the coordination of AI and network from the user's perspective, and completing the efficient computing and output of various AI tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application, and those skilled in the art can obtain other drawings based on these drawings without creative work.

[0028] Figure 1 A schematic diagram of the structure of an AI control system based on a deterministic network provided in an embodiment of the present application.

[0029] Figure 2 Schematic diagram of the AI ​​routing publishing and propagation mechanism provided in an embodiment of the present application.

[0030] Figure 3 Another structural schematic diagram of an AI control system based on a deterministic network provided in an embodiment of the present application.

[0031] Figure 4 A flowchart of a computing task calculation method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0033] The organic integration of the two technical systems of network control and AI control can solve the following problems: 1. Computing resource requirements. The training of large models of AI-generated content requires a large amount of computing resources. The parameter scale of the model reaches billions to trillions. Training these ultra-large-scale models requires ultra-high computing power and a large amount of video memory. For example, training a large model may require hundreds of GPU nodes, and having a large number of GPUs alone is not enough to effectively train large models. Three parallel methods of data, pipeline and tensor parallelism are also needed to coordinate these resources. 2. Bandwidth requirements. In the process of large model training, bandwidth plays a vital role in data transmission. High-bandwidth networks can significantly increase data transmission rates and reduce bottlenecks, thereby accelerating model training, reducing latency and improving overall high-performance computing performance. This requires the use of high-speed interconnect technology, advanced data compression algorithms and data caching mechanisms. 3. Network delay. Network delay in data communication includes static delay and dynamic delay. Static delay is composed of data sequence delay, device forwarding delay and optoelectronic transmission delay, while dynamic delay is mainly composed of internal queuing delay and retransmission delay caused by network congestion and packet loss. Network jitter has a significant impact on training efficiency, so it is necessary to optimize the network to reduce these delays and jitter. 4. Network stability. In large-scale model training, the stability of the network system is crucial. Network failures can seriously affect the connectivity of multiple computing nodes, thereby compromising the overall integrity of the system. Maintaining a stable and efficient network system is a major challenge in large-scale model training, which requires timely fault-tolerant replacement and elastic expansion mechanisms. 5. Data availability and cost. The training process of AI models requires a large amount of data, and data availability and management become a major issue.

[0034] The present application embodiment provides an AI control system based on a deterministic network, referring to Figure 1 , the AI ​​control system based on deterministic network includes a computing network controller, an AI controller and a deterministic network.

[0035] The deterministic network includes a network controller, and the computing network controller is communicatively connected with the AI ​​controller and the network controller respectively.

[0036] The network controller is used to provide deterministic network information and intelligent business routing information to the computing network controller, and the AI ​​controller is used to pass the network scheduling and forwarding model obtained by AI analysis to the computing network controller through an API interface. The computing network controller is used to determine the AI ​​network path and intelligent computing center that are suitable for the computing task based on the deterministic network information, intelligent business routing information, network scheduling and forwarding model, and business information of the computing task, and transmit the computing task to the optimal intelligent computing center through the optimal AI network path for calculation.

[0037] Among them, deterministic networks also include AI routing publishing and propagation mechanisms, such as Figure 2 As shown in the figure, the mechanism is as follows: ① Add an AI programmable chip to the existing Ethernet card, plan the AI ​​header field in the Ethernet frame data segment, and then AIPayload. ② The application in the server operating system modifies the Ethernet frame data according to the business application type. If it is an AI business, write AIHeader and AIPayload, AI user level. ③ The deterministic network routing protocol defines a new routing field that can accommodate new routing information, labels or specific AI optimization parameters. Such as AI-generated path weights, AI-predicted network delays, AI computing power capabilities, etc. ④ The deterministic network router generates a new TLV and distributes the AI ​​routing of the intelligent computing center to other nodes and network controllers in the deterministic network through the network routing protocol LSP (Link State PDU). ⑤ The deterministic network routing protocol periodically verifies the validity of AI routing information, and updates or revokes invalid TLVs when necessary to prevent network performance degradation caused by errors in the prediction of the network scheduling forwarding model analyzed by the AI ​​controller. ⑥ The AI ​​controller system predicts future traffic peaks and network delays based on AI analysis, and synchronizes them to the computing network controller through the API interface to make forwarding path selection in advance and reduce network congestion. ⑦ The AI ​​controller, computing network controller, and network controller work together to achieve dynamic path optimization. The computing network control automatically sends adjustment messages to the network controller based on the results of real-time AI controller analysis and the requirements of intelligent computing tasks. The network controller implements dynamic network path adjustment to achieve optimal bandwidth allocation and personalized computing resource allocation.

[0038] The computing network controller and the network controller can exchange the performance of the entire network through the API interface. The AI ​​controller is one of the core components of the AI ​​control system based on deterministic networks.

[0039] Among them, the computing tasks are uploaded and submitted by the customers.

[0040] In some embodiments, the service information includes service type and service SLA.

[0041] Among them, the computing network controller determines the AI ​​network path and intelligent computing center that are suitable for the computing task based on information such as business type, business SLA, network performance, AI routing, AI prediction, network security, and network scheduling and forwarding model, and transmits the computing task to the optimal intelligent computing center through the optimal AI network path for calculation.

[0042] In some embodiments, the network controller and the network elements in the deterministic network exchange the whole network topology and network performance information through SNMP or Telemetry or BGP_LS or iFIT flow detection technology.

[0043] In some embodiments, the deterministic network information, intelligent service routing information and the network scheduling forwarding model are stored in a pre-set computing network control scheduling database, and the computing network controller can obtain them from the computing network control scheduling database.

[0044] Among them, a computing network control scheduling database is constructed in advance. The computing network control scheduling database records the network scheduling forwarding model, deterministic network information and other information analyzed by AI in the AI ​​controller. The design format corresponding to the computing network control scheduling database is as follows: {AI Model, Route Path, Cost, Network Performance}, where AI Model is the network scheduling forwarding model automatically analyzed by AI, Route Path is the deterministic network forwarding path, Cost is the forwarding path overhead, and Network performance is the deterministic network performance information. The information is synchronized to the computing network controller by the AI ​​controller and the network controller using the API interface to update regularly, and the computing network controller updates the database according to the updated version number.

[0045] In some embodiments, Figure 3 As shown, the AI ​​control system based on the deterministic network also includes a data collection and analysis system, which is communicatively connected to the AI ​​controller; the data collection and analysis system is used to collect deterministic network topology information, whole network performance information, whole network IP routing information and network equipment power consumption information in real time, and use machine learning to perform real-time analysis on the collected information to identify the business characteristics of various types of data, and send the identified business characteristics of various types of data to the AI ​​controller through an API interface for further analysis; the AI ​​controller generates the network scheduling and forwarding model by analyzing the business characteristics through AI.

[0046] Among them, the data collection and analysis system uses machine learning to conduct in-depth analysis of various types of collected data packets, identify various types of data business characteristics, and then pass them to the AI ​​controller through the API interface for further AI analysis.

[0047] In some embodiments, the data collection and analysis system is communicatively connected to the computing network controller and the network controller respectively; the computing network controller and the network controller provide the data collection and analysis system with network status information of the intelligent computing center, computing power resource status information of the intelligent computing center, deterministic network performance status information and deterministic network path status information through an API interface; the data collection and analysis system is used to use machine learning to perform real-time analysis on the information provided by the computing network controller and the network controller to identify the business characteristics of various types of data, and send the identified business characteristics of various types of data to the AI ​​controller through an API interface for further analysis; the AI ​​controller generates the network scheduling and forwarding model through AI analysis of the business characteristics.

[0048] In some embodiments, the AI ​​controller is also used to provide AI identification information, AI model information, AI version information, AI training time information, AI training maturity information, and network security information of AI automatically evaluated to the data collection and analysis system through an API interface, and the data collection and analysis system is also used to provide characteristic data flow information with AI services in network data flows and known AI application information to the AI ​​controller through an API interface.

[0049] Among them, the AI ​​controller provides the intelligent computing center and AI computing and application-related information to the data collection and analysis system through the API interface, such as AI identification, AI model, model version, model training time, model accuracy or maturity, as well as network security information automatically evaluated by AI.

[0050] Among them, the AI ​​controller database is artificially constructed to record data information such as various AI applications, AI regional distribution, AI industry, model, version, model maturity, and AI requirements for network performance. The design format corresponding to the AI ​​controller database is as follows: {Type, Module, Version, Area, Industry, Maturity level, CPR}, where Type is the network data type, Area is the AI ​​application area, Industry corresponds to the industry where AI is used, such as unmanned driving and smart factories, and CPR is the AI ​​network performance requirement information. There are two ways to collect data information in the AI ​​controller database. The first is to collect data through the northbound interface of the network device AI interface, analyze and model the collected data, and identify various AI business characteristics. The second is that the AI ​​controller and the data collection and analysis system are synchronized regularly through the API interface to extract business characteristics from the synchronized data information.

[0051] In some embodiments, the data collection and analysis system collects the deterministic network topology information, the whole network performance information, the whole network IP routing information and the network equipment power consumption information in real time through BGP-LS or SNMP or Telemetry or NetFlow or port mirroring.

[0052] Among them, by collecting the flow data of the ports of deterministic network equipment, the flow characteristics are extracted, such as data packet size, data flow direction, communication port, transmission rate, etc. In addition, the data collection system is connected to the computer room dynamic environment system through the API interface to obtain the power environment data of the computer room, such as the power supply and temperature of the computer room, and provide the AI ​​control system with AI energy-saving model calculation data information.

[0053] Among them, a network data collection database is built in advance to record network data traffic characteristics, traffic size, protocol type, port, session, network performance information and other data information. The design format corresponding to the network data collection database is as follows: {Traffic Feature, Traffic Size, Protocol Type, Port, Session, NetworkPerformance}. There are two ways to collect data information in the network database. The first is to collect through deterministic network device port splitting, mirroring, SNMP, Telemetry, NetFlow and other methods. The second is that the network data collection and analysis system and the computing network controller and AI controller are synchronized regularly through the API interface.

[0054] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present application, which will not be described one by one here.

[0055] In specific implementation, the present application is not limited by the execution order of the various steps described. If no conflict occurs, some steps can be performed in other orders or simultaneously.

[0056] From the above, it can be seen that the AI ​​control system based on the deterministic network provided in the embodiment of the present application includes a computing network controller, an AI controller and a deterministic network; the deterministic network includes a network controller, and the computing network controller is communicated with the AI ​​controller and the network controller respectively; the network controller is used to provide deterministic network information and intelligent business routing information to the computing network controller, and the AI ​​controller is used to pass the network scheduling and forwarding model obtained by AI analysis to the computing network controller through an API interface, and the computing network controller is used to determine the AI ​​network path and intelligent computing center adapted to the computing task based on the deterministic network information, intelligent business routing information, network scheduling and forwarding model, and business information of the computing task, and transmit the computing task to the optimal intelligent computing center through the optimal AI network path for calculation, thereby realizing the organic integration of the two technical systems of network control and AI control, realizing the collaboration of AI and network from the user's perspective, and completing the efficient calculation and output of various AI tasks.

[0057] The present application also provides a computing task calculation method, which is applied to the AI ​​control system based on the deterministic network described in any of the above embodiments, with reference to Figure 4 , Figure 4 A flowchart of a computing task calculation method provided in an embodiment of the present application, the method comprising:

[0058] Step 101, receiving deterministic network information and intelligent service routing information provided by the network controller, and the network scheduling forwarding model transmitted by the AI ​​controller through the API interface.

[0059] Among them, deterministic network information, intelligent service routing information and network scheduling forwarding model are stored in a pre-set computing network control scheduling database.

[0060] Step 102: Determine an AI network path and intelligent computing center that are compatible with the computing task based on the deterministic network information, intelligent business routing information, network scheduling and forwarding model, and business information of the computing task.

[0061] The computing tasks are uploaded and submitted by the customers, and the business information includes the business type and business SLA.

[0062] Among them, the AI ​​network path and intelligent computing center adapted to the computing task can be one or more.

[0063] Step 103: transmit the computing task through the optimal AI network path to the optimal intelligent computing center for computing.

[0064] Among them, if it is determined that there is only one AI network path and intelligent computing center that are suitable for the computing task, the optimal AI network path and the optimal intelligent computing center are the ones that are suitable. If it is determined that there are multiple AI network paths and intelligent computing centers that are suitable for the computing task, the optimal AI network path is one of them, and the optimal intelligent computing center is one of them.

[0065] In some embodiments, if there are multiple AI network paths and intelligent computing centers adapted to the computing task, before transmitting the computing task to the optimal intelligent computing center through the optimal AI network path for computing, it also includes:

[0066] Determine the optimal AI network path and the optimal intelligent computing center from multiple adapted AI network paths and intelligent computing centers.

[0067] From the above, it can be seen that the computing task calculation method provided in the embodiment of the present application receives the deterministic network information and intelligent business routing information provided by the network controller, and the network scheduling and forwarding model transmitted by the AI ​​controller through the API interface, and then determines the AI ​​network path and intelligent computing center adapted to the computing task based on the deterministic network information, intelligent business routing information, network scheduling and forwarding model, and business information of the computing task, and finally transmits the computing task to the optimal intelligent computing center through the optimal AI network path for calculation, thereby realizing the organic integration of the two technical systems of network control and AI control, realizing the collaboration of AI and network from the user's perspective, and completing the efficient calculation and output of various AI tasks.

[0068] The above is a detailed introduction to the AI ​​control system based on deterministic network and the computing task calculation method provided by the embodiment of the present application. This article uses specific examples to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application; at the same time, for technicians in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. An AI control system based on a deterministic network, characterized in that: Including computing network controller, AI controller and deterministic network; The deterministic network includes a network controller, and the computing network controller is communicatively connected with the AI ​​controller and the network controller respectively; The network controller is used to provide deterministic network information and intelligent business routing information to the computing network controller, and the AI ​​controller is used to pass the network scheduling and forwarding model obtained by AI analysis to the computing network controller through an API interface. The computing network controller is used to determine the AI ​​network path and intelligent computing center that are suitable for the computing task based on the deterministic network information, intelligent business routing information, network scheduling and forwarding model, and business information of the computing task, and transmit the computing task to the optimal intelligent computing center through the optimal AI network path for calculation.

2. The AI ​​control system based on a deterministic network as claimed in claim 1, characterized in that: Also included is a data collection and analysis system, the data collection and analysis system is communicatively connected to the AI ​​controller; The data collection and analysis system is used to collect deterministic network topology information, whole network performance information, whole network IP routing information and network equipment power consumption information in real time, and use machine learning to analyze the collected information in real time to identify the business characteristics of various types of data, and send the identified business characteristics of various types of data to the AI ​​controller through an API interface for further analysis; The AI ​​controller generates the network scheduling forwarding model by analyzing the service characteristics through AI.

3. The AI ​​control system based on a deterministic network as claimed in claim 2, characterized in that: The data collection and analysis system is respectively connected to the computing network controller and the network controller for communication; The computing network controller and the network controller provide the data collection and analysis system with network status information of the intelligent computing center, computing power resource status information of the intelligent computing center, deterministic network performance status information, and deterministic network path status information through an API interface; The data collection and analysis system is used to perform real-time analysis on the information provided by the computing network controller and the network controller by means of machine learning to identify the business characteristics of various types of data, and send the identified business characteristics of various types of data to the AI ​​controller through an API interface for further analysis; The AI ​​controller generates the network scheduling forwarding model by analyzing the service characteristics through AI.

4. The AI ​​control system based on a deterministic network as claimed in claim 2, characterized in that: The AI ​​controller is also used to provide AI identification information, AI model information, AI version information, AI training time information, AI training maturity information and network security information of AI automatically evaluated to the data collection and analysis system through an API interface. The data collection and analysis system is also used to provide characteristic data flow information with AI services in network data flows and known AI application information to the AI ​​controller through an API interface.

5. The AI ​​control system based on a deterministic network as claimed in claim 2, characterized in that: The data collection and analysis system collects the deterministic network topology information, the whole network performance information, the whole network IP routing information and the network equipment power consumption information in real time through BGP-LS or SNMP or Telemetry or NetFlow or port mirroring.

6. The AI ​​control system based on a deterministic network as claimed in claim 1, characterized in that: The deterministic network information, intelligent service routing information and the network scheduling forwarding model are stored in a pre-set computing network control scheduling database, and the computing network controller can obtain them from the computing network control scheduling database.

7. The AI ​​control system based on a deterministic network as claimed in claim 1, characterized in that: The service information includes service type and service SLA.

8. The AI ​​control system based on a deterministic network as claimed in claim 1, characterized in that: The network controller and the network elements in the deterministic network exchange the whole network topology and network performance information through SNMP or Telemetry or BGP_LS or iFIT flow detection technology.

9. A computing task calculation method, applied to the AI ​​control system based on a deterministic network as claimed in any one of claims 1 to 8, characterized in that: The method comprises: Receive deterministic network information and intelligent service routing information provided by the network controller, as well as the network scheduling and forwarding model delivered by the AI ​​controller through the API interface; Determine an AI network path and intelligent computing center adapted to the computing task based on the deterministic network information, intelligent business routing information, network scheduling and forwarding model, and business information of the computing task; The computing task is transmitted to the optimal intelligent computing center through the optimal AI network path for calculation.

10. The computing task calculation method according to claim 9, characterized in that: If there are multiple AI network paths and intelligent computing centers that are compatible with the computing task, before transmitting the computing task to the optimal intelligent computing center through the optimal AI network path for computing, the method further includes: Determine the optimal AI network path and the optimal intelligent computing center from multiple adapted AI network paths and intelligent computing centers.

Images