A method, apparatus, equipment, medium, and program product for collaborative scheduling of computing networks.
By parsing business requests and decomposing global constraints in the intelligent agent Internet, distinguishing between stateless and stateful service requests, and adopting different scheduling mechanisms, the problems of system versatility and low resource utilization are solved, achieving efficient resource scheduling and deterministic response.
Patent Information
- Application Number
- CN202611140265.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-29
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies in the Internet of Intelligent Agents suffer from poor system versatility and scalability, as well as low resource utilization and execution efficiency, due to the need to customize processing strategies according to different scenarios and tasks.
By parsing business requests through the orchestration and scheduling module, global constraints are broken down into node and link quantification constraints. Stateless and stateful service requests are distinguished and scheduled autonomously by the network controller or uniformly planned by the orchestration and scheduling module, respectively. This decouples the processing flow and improves system adaptability and resource utilization.
It achieves complete decoupling of the link from user needs to scheduling instructions, improves the system's versatility, resource utilization efficiency and scheduling determinism in multiple scenarios, and reduces processing latency and signaling overhead.
Smart Images

Figure CN123123349A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of communication technology, and in particular to a computing network collaborative scheduling method, apparatus, equipment, medium, and program product. Background Technology
[0002] With the convergence of artificial intelligence, distributed computing, and network communication technologies, the Internet of Intelligent Agents has become a new type of infrastructure supporting multi-agent, multi-task collaboration. Autonomous intelligent agents complete complex business tasks through dynamic collaboration, and user needs are characterized by multiple scenarios, multiple types, strong collaboration, and high dynamism, placing higher demands on collaboration mechanisms, orchestration capabilities, and resource scheduling efficiency.
[0003] Currently, related technologies receive stateless service requests and stateful service requests, and customize parsing rules, orchestration logic, and scheduling strategies according to different scenarios and tasks to complete parsing, decomposition, selection, matching, and scheduling execution. Ultimately, multiple agents collaborate to complete the task. In this process, the various links are deeply coupled, and global constraints such as end-to-end latency and service quality are closely related to deployment location, link status, and computing power.
[0004] However, existing technologies suffer from poor system versatility and scalability, as they require customized processing strategies based on different scenarios and tasks, resulting in low resource utilization and execution efficiency. Summary of the Invention
[0005] This application provides a computer network collaborative scheduling method, device, equipment, medium, and program product to improve the system's versatility, resource utilization efficiency, and scheduling determinism in multiple scenarios.
[0006] In a first aspect, embodiments of this application provide a computer network collaborative scheduling method, the method comprising:
[0007] The orchestration and scheduling module determines the request type and structured requirement information based on the business request information. The structured requirement information includes node information, connection information between nodes, and global constraint information. The request type includes stateless service requests or stateful service requests.
[0008] The global constraint information is decomposed by the orchestration and scheduling module to obtain constraint decomposition information;
[0009] When the request type is the stateless service request, the node information, the connection information and the constraint decomposition information are transmitted to the network controller through the orchestration and scheduling module, so that the network controller performs network scheduling according to the structured requirement information.
[0010] When the request type is a stateful service request, the orchestration and scheduling module selects nodes and network paths based on the node information, connection information, and constraint decomposition information, generates a computing network scheduling instruction containing the node selection results and network path selection results, and sends the computing network scheduling instruction to the computing power controller and the network controller, so that the computing power controller and the network controller allocate computing network resources according to the computing network scheduling instruction.
[0011] The technical solution provided in this application brings at least the following beneficial effects:
[0012] The method in this embodiment first determines the request type and structured requirement information based on the service request information. Then, the orchestration and scheduling module decomposes the global constraint information to obtain structured constraint decomposition information, transforming global constraints such as end-to-end latency into quantifiable and verifiable node-level and connection-level constraints. Based on this, for stateless service requests, node information, connection information, and constraint decomposition information are transparently transmitted to the network controller, enabling it to autonomously complete scheduling based on real-time network conditions. This avoids redundant intervention by the orchestration and scheduling module for simple requests, reducing processing latency and signaling overhead. For stateful service requests, the orchestration and scheduling module centrally selects computing nodes and plans network paths based on node information, connection information, and constraint decomposition information, generating a computing network scheduling instruction containing resource allocation results and synchronously sending it to the computing power controller and network controller. This ensures that computing network resources are allocated collaboratively according to a unified strategy, avoiding global constraint failure or scheduling conflicts. Therefore, this method decouples the request type-driven processing flow at the architectural level and opens up the complete link from user needs to executable scheduling instructions through the structured decomposition of global constraints, effectively improving the system's versatility, resource utilization efficiency and scheduling determinism in multiple scenarios.
[0013] One possible implementation is that the constraint decomposition information includes linkage rule information and delay quota information;
[0014] The step of decomposing the global constraint information to obtain constraint decomposition information includes:
[0015] The orchestration and scheduling module decomposes the global constraint information to obtain the linkage rule information and the delay quota information.
[0016] The linkage rule information includes processing delay linkage rules between nodes, bandwidth linkage rules between different segments, and global dynamic constraint linkage rules; the delay quota information includes the total node processing delay quota and the total network transmission delay quota.
[0017] Another possible implementation, wherein the step of transparently transmitting the node information, the connection information, and the constraint decomposition information to the network controller, includes:
[0018] Network-related information is extracted from the structured requirements information, including network location-related attributes in the node information and network transmission-related attributes in the connection information;
[0019] The network-related information, the linkage rule information, and the latency quota information are transmitted as transparent information. A transparent identifier is added to the transparent information and then sent to the network controller. The transparent identifier is used to instruct the network controller to autonomously select a service instance based on the actual network status and to verify it based on the linkage rule information and the latency quota information.
[0020] Another possible implementation includes, before selecting nodes and network paths based on the node information, the connection information, and the constraint decomposition information:
[0021] Based on the service path represented by the connection information, the total network transmission delay quota is allocated to each connection segment, generating transmission delay constraints for each connection segment, and using these constraints as filtering conditions for network path selection; and
[0022] The total node processing latency quota is used as the cumulative upper limit constraint for node processing latency when selecting a node.
[0023] Another possible implementation is that the computing network scheduling instructions include computing power scheduling instructions and network scheduling instructions;
[0024] The step of selecting nodes and network paths based on the node information, the connection information, and the constraint decomposition information, and generating a network scheduling instruction containing node selection results and network path selection results, includes:
[0025] From the candidate node set corresponding to the node information, determine the target node that satisfies the transmission delay constraint and the cumulative upper limit constraint, and allocate computing resources to the target node according to the node information;
[0026] Select a target network path based on the location of the target node, and configure network parameters for the target network path based on the connection information;
[0027] Associate the linkage rule information with the target node and the target network path;
[0028] Based on the target node, the allocated computing resources, the target network path, the configured network parameters, and the associated linkage rule information, computing power scheduling instructions and network scheduling instructions are generated.
[0029] Another possible implementation, after selecting nodes and network paths based on the node information, the connection information, and the constraint decomposition information, and generating a network scheduling instruction containing the node selection results and network path selection results, further includes:
[0030] Based on the actual computing power of each target node, the total processing latency quota of the node is allocated to each target node, and the executable computing latency parameters of each target node are generated.
[0031] Based on the actual transmission delay of each segment of the target network path, the total network transmission delay quota is allocated to each segment of the connection, and executable network delay parameters for each segment of the connection are generated.
[0032] The executable computation latency parameter and the executable network latency parameter are respectively carried in the computing power scheduling instruction and the network scheduling instruction.
[0033] Another possible implementation includes:
[0034] Receive a feedback signal generated by the network controller or the computing power controller based on the computing network scheduling instruction. The feedback signal is used to characterize that the actual network status or the actual computing power resource status cannot meet the linkage rule information or the latency quota information.
[0035] In response to the feedback signal, the splitting ratio of the global constraint information is adjusted or a scaling up / down operation is performed. Based on the adjusted splitting ratio, node selection and network path selection are re-executed, and a new computing network scheduling instruction is generated and issued.
[0036] Secondly, embodiments of this application provide a computer network collaborative scheduling device, comprising:
[0037] An orchestration and scheduling module is used to execute the method described in the first aspect above.
[0038] Thirdly, this application provides an electronic device comprising: a processor and a memory; the memory stores a program or instructions executable on the processor, wherein the program or instructions, when executed by the processor, implement the method of the first aspect described above.
[0039] Fourthly, this application provides a readable storage medium on which a program or instructions are stored, which, when executed by a computer, implement the method of the first aspect described above.
[0040] Fifthly, this application provides a computer program product stored in a storage medium, which, when executed by a computer, implements the method described in the first aspect.
[0041] In a sixth aspect, embodiments of this application provide a chip including a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the method described in the first aspect.
[0042] The beneficial effects of the second to sixth aspects mentioned above are described in the corresponding description of the first aspect and will not be repeated here. Attached Figure Description
[0043] Figure 1 This application provides a network architecture for a computer network collaborative scheduling method.
[0044] Figure 2 This is a flowchart illustrating a computer network collaborative scheduling method provided in an embodiment of this application.
[0045] Figure 3 This is a schematic diagram of the transparent transmission processing flow of an orchestration and scheduling module under a stateless service request, provided in an embodiment of this application.
[0046] Figure 4 This is a flowchart illustrating a method for generating global constraint quota decomposition and scheduling screening conditions, provided in an embodiment of this application.
[0047] Figure 5 This is a schematic flowchart illustrating a method for generating network scheduling instructions, provided in an embodiment of this application.
[0048] Figure 6 This is a schematic diagram illustrating the process of converting a global constraint quota into an executable latency parameter, as provided in an embodiment of this application.
[0049] Figure 7 This is a schematic flowchart of a computer network collaborative scheduling feedback adjustment method provided in an embodiment of this application.
[0050] Figure 8 This is a flowchart illustrating another computer network collaborative scheduling method provided in an embodiment of this application.
[0051] Figure 9 This is a schematic diagram of the structure of a computer network collaborative scheduling device provided in an embodiment of this application.
[0052] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0053] The following section will describe in detail the computer network collaborative scheduling method, apparatus, equipment, medium, and program products provided in this application, with reference to the accompanying drawings.
[0054] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.
[0055] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.
[0056] The terms "at least one," "at least one of," etc., used in the specification and claims of this application refer to any one, any two, or a combination of two or more of the included items. For example, at least one of a, b, and c can mean: "a," "b," "c," "a and b," "a and c," "b and c," and "a, b, and c," where a, b, and c can be single or multiple. Similarly, "at least two" refers to two or more items, and its meaning is similar to that of "at least one."
[0057] In the description of this application, unless otherwise stated, "a plurality of" means two or more.
[0058] The embodiments of this application provide a computing network collaborative scheduling method, apparatus, device, medium, and program product that can be applied to the Agent Internet scenario.
[0059] In existing technologies, a typical intelligent agent collaboration and business orchestration method exists in the field of intelligent agent Internet. Its processing flow mainly includes the following key steps: First, the system receives user business requests, which may involve different types such as stateless or stateful service requests. Next, for different business scenarios and task types, the system adopts a case-by-case customized implementation approach, designing requirement parsing rules, service orchestration logic, and resource scheduling strategies to parse business requirements, decompose tasks, and select services. Then, the decomposed tasks are matched with available intelligent agent resources, and resource allocation and task scheduling are performed. Finally, multiple intelligent agents complete the business task through dynamic collaboration based on the allocated resources and scheduling results. In this process, requirement parsing, application orchestration, and resource scheduling are deeply coupled, and the global constraints involved in user requirements, such as end-to-end latency, service quality, and execution sequence, directly depend on service deployment location, network link status, and computing power capabilities.
[0060] However, in the above process, because the system needs to customize the requirements parsing, service orchestration and resource scheduling strategies for different business scenarios and different task types, the system's versatility and scalability are poor, resulting in low overall resource utilization and business execution efficiency.
[0061] To address the aforementioned technical issues, this application provides a network collaborative scheduling method, apparatus, device, medium, and program product. The method in this embodiment first parses service requests and decomposes global constraints into node and link quantification constraints. Then, it distinguishes between two types of requests for separate processing: the former is autonomously scheduled and its workload reduced by the network controller, while the latter is uniformly planned and issued with collaborative instructions by the orchestration and scheduling module. This solution decouples the processing flow, connects the demand to the scheduling instruction link, and improves system adaptability, resource utilization, and scheduling stability.
[0062] The following description, in conjunction with the accompanying drawings, details the computer network collaborative scheduling method, apparatus, equipment, medium, and program products provided in the embodiments of this application.
[0063] Figure 1 This illustration shows the network architecture of a computer network collaborative scheduling method provided in an embodiment of this application. For example... Figure 1 As shown, the network architecture includes: an orchestration and scheduling device 101, a computing power controller 102, a network controller 103, and a terminal device 104. The orchestration and scheduling device 101 is interconnected with the computing power controller 102, the network controller 103, and the terminal device 104.
[0064] In some embodiments, the orchestration and scheduling device 101 may be a server, a computer, or a processor or processing unit within a server or computer. The server may be a single server or a server cluster consisting of multiple servers. It should be noted that the embodiments of this application do not limit the specific device form of the orchestration and scheduling device 101. Figure 1 The orchestration and scheduling device 101 is shown as a single server as an example.
[0065] In some embodiments, the computing power controller 102 and the network controller 103 may each be a functional module in a server, computer, virtual machine, container or cloud platform, or a processor or processing unit in a server or computer. Figure 1 The example shown is that the computing power controller 102 and the network controller 103 are each an independent server.
[0066] In some embodiments, the terminal device may be a mobile phone, tablet computer, laptop computer, handheld computer, in-vehicle electronic device, mobile internet device (MID), augmented reality (AR) / virtual reality (VR) device, robot, wearable device, personal computer (PC), ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc., and the embodiments of this application do not specifically limit it. Figure 1 The example shown is a mobile phone, with terminal device 104 as the example.
[0067] In some embodiments, the orchestration and scheduling device 101 is used to receive service request information initiated by the terminal device 104, and determine the request type and structured requirement information according to the service request information, wherein the request type includes a stateless service request or a stateful service request; the orchestration and scheduling device 101 is also used to decompose the global constraint information in the structured requirement information to obtain constraint decomposition information; when the request type is a stateless service request, the orchestration and scheduling device 101 transmits the node information, connection information and constraint decomposition information in the structured requirement information to the network controller 103; when the request type is a stateful service request, the orchestration and scheduling device 101 performs node selection and network path selection according to the node information, connection information and constraint decomposition information, generates a computing network scheduling instruction containing the node selection result and the network path selection result, and sends the computing network scheduling instruction to the computing power controller 102 and the network controller 103 respectively. The computing power controller 102 is used to receive and execute the computing power scheduling part of the computing network scheduling instruction issued by the orchestration and scheduling device 101, and allocate computing power resources to the selected nodes. The network controller 103 is used to receive and execute the network scheduling portion of the network scheduling instruction issued by the orchestration and scheduling device 101, or to receive node information, connection information, and constraint decomposition information transmitted transparently by the orchestration and scheduling device 101, and to autonomously perform network scheduling according to the actual network status. The terminal device 104 is used to initiate service request information to the orchestration and scheduling device 101, and the service request information carries a description of the user's needs.
[0068] It should be noted that the network architecture described in the embodiments of this application is for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and does not constitute a limitation on the technical solutions provided in the embodiments of this application. As network architectures evolve, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.
[0069] See Figure 2 This is a flowchart illustrating a computer network collaborative scheduling method provided in an embodiment of this application. Figure 2 As shown, the computer network collaborative scheduling method provided in this application embodiment can be implemented by the above-mentioned orchestration and scheduling device, wherein the orchestration and scheduling device is configured with an orchestration and scheduling module, specifically including the following steps 201 to 204.
[0070] Step 201: The orchestration and scheduling device determines the request type and structured requirement information based on the service request information through the orchestration and scheduling module. The structured requirement information includes node information, connection information between nodes, and global constraint information. The request type includes stateless service requests or stateful service requests.
[0071] Step 202: The orchestration and scheduling device decomposes the global constraint information through the orchestration and scheduling module to obtain constraint decomposition information;
[0072] Step 203: When the request type is a stateless service request, the orchestration and scheduling device transmits node information, connection information and constraint decomposition information to the network controller through the orchestration and scheduling module, so that the network controller performs network scheduling according to the structured requirement information.
[0073] Step 204: When the request type is a stateful service request, the orchestration and scheduling device selects nodes and network paths based on node information, connection information and constraint decomposition information through the orchestration and scheduling module, generates a computing network scheduling instruction containing node selection results and network path selection results, and sends the computing network scheduling instruction to the computing power controller and the network controller, so that the computing power controller and the network controller allocate computing network resources according to the computing network scheduling instruction.
[0074] In this embodiment, the execution entity is the orchestration and scheduling module. This module is a core functional entity in the intelligent agent internet collaborative architecture. It receives service request information from above and interacts with the network controller and computing power controller to jointly complete end-to-end network-computer collaborative scheduling. The network controller is responsible for the perception, configuration, and management of network resources, while the computing power controller is responsible for the perception, configuration, and management of computing power resources.
[0075] First, the orchestration and scheduling device determines the request type and structured requirement information based on the service request information through the orchestration and scheduling module.
[0076] In this embodiment, the structured requirement information includes node information, connection information between nodes, and global constraint information. The request type includes stateless service requests or stateful service requests.
[0077] Specifically, upon receiving a business request, the orchestration and scheduling module first parses the request to determine the request type and structured requirements. In this embodiment, the business request information can originate from the user terminal, the upper-layer application system, or other intelligent agents. Based on the above parsing process, the fuzzy requirements input by the user, which may contain scenario-specific information, can be transformed into a standardized data structure that the system can recognize and process, laying the foundation for subsequent differentiated processing and constraint decomposition.
[0078] Based on the above, it can be seen that structured requirement information includes at least the following three types of information: node information, connection information between nodes, and global constraint information. Specifically:
[0079] Node information is used to characterize the various entities participating in this business request. Node types may include, but are not limited to: user terminal nodes (such as the source node initiating the request), service nodes (such as intelligent agent service nodes providing stateless services), computing power deployment nodes (such as the target node requiring the deployment of AI inference services), storage nodes, etc. Each node's attribute information may include node location (such as edge regions, core clouds, etc.), computing resource information (such as the number of CPU cores, memory size, etc.), and node processing capacity information, etc.
[0080] Connection information between nodes is used to characterize the interaction relationships and network links between nodes. Each connection can represent a communication path between two nodes, and its attribute information can include network protocol type (such as SRv6 protocol), bandwidth requirements, network latency requirements, etc. Connection information reflects the path and quality requirements for data transmission of service requests between nodes.
[0081] Global constraint information is used to characterize the overall requirements of users for the entire end-to-end execution process of a business request. Global constraints may include, but are not limited to: end-to-end latency constraints (such as the end-to-end latency of the entire business flow not exceeding a preset threshold), bandwidth constraints, and quality of service constraints. Unlike node-level or connection-level local constraints, global constraints are holistic constraints spanning multiple nodes and multiple connections, and whether they are satisfied depends on the combined results of the attributes of each node and each connection.
[0082] It should be noted that, optionally, structured requirement information can be generated using a predefined requirement description template. This template specifies a unified data format and field definitions for node information, connection information, and global constraint information, ensuring that user requirements can be converted into a data model with the same structure regardless of the user's specific needs.
[0083] Furthermore, the orchestration and scheduling module in this embodiment also determines the request type based on the service request information. Request types include stateless service requests and stateful service requests. Specifically:
[0084] Stateless service requests refer to requests where the service itself does not maintain session state with the client, each request is processed independently, and there is no need for state synchronization between service instances. The characteristics of stateless service requests are: lightweight request processing, high real-time requirements, dynamic scaling of service instances, and the ability of any service instance to handle any request.
[0085] Stateful service requests refer to business requests that require maintaining session or resource state, including resource deployment requests and agent requests. Resource deployment requests involve deploying applications, services, or computing resources on specific nodes, such as deploying an AI inference service cluster or a microservice application. Agent requests involve collaborative processing by agents and require state maintenance, such as multi-agent collaboration to complete tasks or agent service scheduling. Stateful service requests are characterized by heavy resource consumption, the need for fixed deployment locations, involvement of state maintenance or exclusive resource usage, and long lifecycles.
[0086] In practical applications, the request type can be determined either by parsing the type identifier field carried in the business request information, or by automatically classifying the request content after semantic analysis. For example, a task type field can be set in the standardized requirement description template, and its value can be "stateless" (indicating a stateless service request) or "stateful" (indicating a stateful service request, such as a resource deployment request or an agent request).
[0087] Preferably, when determining structured demand information, the orchestration and scheduling module can perform the following operations:
[0088] First, extract the participating entities from the user requirements, map them to nodes in a standardized topology model, and assign a unique identifier to each node (such as N001, N002, etc.).
[0089] Secondly, extract the interaction relationships from user needs, map them to connections in a standardized topology model, and assign a unique identifier (such as C001) to each connection.
[0090] Next, extract the global constraints from the user requirements, map them to global constraints in the standardized topology model, and record the constraint type (such as delay constraints, bandwidth constraints), constraint formula (such as the calculation formula for end-to-end delay), constraint threshold and unit in the global constraints.
[0091] For example, for a stateless service request that “access a certain agent service with an end-to-end latency of ≤10ms”, the orchestration and scheduling module can parse the requirement as follows: Node information includes the user terminal node (N001, type: source, location: edge A area) and the agent service node (N002, type: stateless-service, location: core cloud); connection information includes the access link from N001 to N002 (C001, protocol: SRv6, bandwidth: 10 Gbps); global constraint information includes the end-to-end latency constraint (i.e., the sum of the computation latency of all nodes and the sum of the network transmission latency of all links, with a threshold of ≤10ms, in milliseconds).
[0092] For example, for a resource deployment request of "deploying AI inference service with end-to-end latency ≤ 10ms + computing resource allocation", the orchestration and scheduling module can parse the requirement into: node information including user terminal node (N001, type source) and AI inference computing power deployment node (N002, type stateful-service, which needs to be allocated CPU and memory resources); connection information including the deployment link from N001 to N002; and global constraint information including end-to-end latency constraints and computing resource allocation constraints.
[0093] Through the above methods, different types of user needs are converted into structured requirement information in a unified format, thereby achieving standardized expression of user needs and decoupling from the scenario.
[0094] After obtaining structured requirement information, the orchestration and scheduling module decomposes the global constraint information to obtain constraint decomposition information. This step can resolve the coupling problem between global constraints and nodes / connections. As mentioned earlier, global constraints (such as end-to-end latency) are holistic constraints spanning multiple nodes and connections, and their satisfaction depends on the combined attributes of each node and connection. Before scheduling, the service deployment location and network path are not yet determined, so it is impossible to directly provide specific constraint values for each node and connection. This step transforms global constraints into structured constraint decomposition information that can be used for subsequent scheduling under the conditions of "uncertain candidate locations and dynamically changing network states."
[0095] After obtaining the constraint decomposition information, the orchestration and scheduling module executes differentiated processing flows based on the determined request type. Specifically, the orchestration and scheduling module adopts completely different processing logic for stateless service requests and stateful service requests.
[0096] Stateless service requests are typically characterized by lightweight nature and high real-time requirements. In these scenarios, the user's core need is to "access an available service instance with minimal latency." The specific deployment location of the service is transparent to the user, and the real-time status of the network link is the key factor determining service quality. Therefore, in this scenario, the orchestration and scheduling module only undertakes the functions of demand parsing and information transmission, and does not participate in specific resource selection and scheduling decisions.
[0097] Specifically, the orchestration and scheduling module extracts network scheduling-related information from the structured demand information, including network location-related attributes from node information, network transmission-related attributes from connection information, and constraint decomposition information. After trimming and assembling this information, the orchestration and scheduling module directly transmits it to the network controller, adding a pass-through identifier to this information. This pass-through identifier explicitly informs the network controller that this request is initiated by the network controller to autonomously select a service instance based on the actual network status, and that the selection result is verified against the passed-through constraint decomposition information. During this process, there is no interaction between the orchestration and scheduling module and the computing power controller.
[0098] After receiving the transparent information, the network controller autonomously selects service instances and network paths that meet the various screening conditions in the constraint decomposition information based on the real-time status of the actual network, such as link quality and node reachability, and completes network scheduling.
[0099] Stateful service requests are typically characterized by high resource consumption and the need for fixed deployment. In these scenarios, it's crucial to consider not only network connectivity quality but also multiple factors such as available resources, computing power, and deployment location of computing nodes. This results in a deep coupling between computing power and network requirements. Therefore, the orchestration and scheduling module must be involved in orchestration and scheduling decisions throughout the entire process.
[0100] Specifically, the orchestration and scheduling module performs joint scheduling calculations for computing node selection and network path selection based on node information, connection information, and constraint decomposition information. In the node selection phase, the module selects target nodes from the candidate node set corresponding to the node information that meet the selection criteria in the constraint decomposition information (e.g., the upper limit constraint on cumulative node processing latency), and allocates specific computing resources to these target nodes (e.g., allocating 8 CPU cores and 32GB of memory). In the network path selection phase, the module selects target network paths that meet the transmission latency constraints of each connection segment based on the determined target node locations, and configures specific network parameters (e.g., bandwidth, protocol) for these network paths. Simultaneously, the module also associates and binds the linkage rules in the constraint decomposition information with the selected target nodes and target network paths.
[0101] After completing the above selection and configuration, the orchestration and scheduling module generates a computing network scheduling instruction containing node selection results and network path selection results. This instruction can be further encapsulated into two parts: a computing power scheduling instruction and a network scheduling instruction. The orchestration and scheduling module sends the computing power scheduling instruction to the computing power controller and the network scheduling instruction to the network controller. Upon receiving the instructions, the computing power controller and the network controller only execute the specific resource allocation and configuration operations according to the instruction content and do not participate in the scheduling decision.
[0102] Based on the above methods, various business requests are uniformly accessed through the orchestration and scheduling module, and differentiated processing flows are automatically switched according to the request type, achieving the separation of demand parsing, service orchestration, resource scheduling, and computing network control. For stateless service requests, a lightweight pass-through processing mode is adopted, with the network controller autonomously scheduling based on real-time network conditions, ensuring efficient response for low-latency services. For stateful service requests, a full-process orchestration and scheduling mode is adopted, with the orchestration and scheduling module uniformly completing the selection of computing power nodes, network path selection, and constraint association binding, ensuring the determinism and reliability of resource-intensive service deployment. Simultaneously, by adopting a decomposition strategy of "first splitting the structure and quota, then allocating specific values" for global constraints, the technical challenge of decomposing and mapping global constraints under unknown service deployment locations can be effectively solved, ensuring that global constraints such as end-to-end latency are propagated, verifiable, and satisfied throughout the entire scheduling process.
[0103] In some embodiments, global constraint information can be structurally decomposed to address the deep coupling between end-to-end global constraints (such as latency) and uncertain service deployment locations and network states in intelligent agent Internet scenarios. Specifically:
[0104] First, the decomposition of global constraint information does not involve directly breaking down global constraints into specific values, but rather into two levels: "constraint structure" and "quota ratio".
[0105] In one specific embodiment, for end-to-end latency constraints, the orchestration and scheduling module does not directly allocate specific milliseconds when the service deployment location is unknown. Instead, it first decomposes the constraint into a fixed structural expression. For example, the serial service A→B→C is decomposed into "E2E = NodeA + NetAB + NodeB + NetBC + NodeC ≤ T". Based on this, the total latency quota is proportionally allocated according to a preset strategy (such as an industry-wide strategy or a custom strategy). For example, the network total latency quota is set to 60%, and the node total latency quota is set to 40%, thus obtaining the "constraint decomposition information".
[0106] Secondly, the constraint decomposition information specifically includes linkage rule information and latency quota information. The latency quota information specifically includes the total node processing latency quota and the total network transmission latency quota. Based on the above example, with a total latency T = 10ms, the decomposition yields a total node processing latency quota of 4ms and a total network transmission latency quota of 6ms. The total network transmission latency quota can also be evenly distributed or otherwise weighted according to the number of links in the service path, generating transmission latency constraints for each connection segment (e.g., NetAB ≤ 3ms, NetBC ≤ 3ms). These constraints will serve as filtering conditions for subsequent network path selection.
[0107] Finally, the linkage rule information is used to describe the coupling relationship and linkage adjustment logic between global constraints, node constraints, and connection constraints. Specifically, the linkage rule information includes processing delay linkage rules between nodes, bandwidth linkage rules between different connection segments, and global dynamic constraint linkage rules.
[0108] Inter-node processing latency linkage rules can be used to describe the constraints that different nodes should meet in terms of processing latency. For example, it can be stipulated that the computation latency of N002 should not be less than that of N001. Inter-connection bandwidth linkage rules are used to describe the constraints that different network connections should meet in terms of bandwidth and other attributes. Global dynamic constraint linkage rules are for global constraints that cannot be completely broken down to a single node or connection. In other words, when the attribute of a certain node or connection changes, other nodes or connections should perform linkage adjustment actions. For example, "when the transmission latency of a certain connection increases by 2ms, the corresponding node's processing latency is compressed by 2ms," which serves as the basis for subsequent verification and dynamic adjustment.
[0109] Based on the above method, by decomposing global constraint information into linkage rule information and latency quota information, the transformation of global constraints from top-level abstract requirements to implementable attribute thresholds and executable linkage rules is realized. Based on this decomposition method, even when the service deployment location and network status are unknown, it is still possible to ensure that global constraints such as end-to-end latency are propagated, verifiable, and satisfied during the scheduling process.
[0110] In the context of computing-network collaboration for the Internet of Intelligent Agents, user service requests exhibit significant differences. For stateless service requests, each request can be processed independently by any available service instance, with weaker requirements for computing resources but greater sensitivity to the real-time quality of network connections. Using the same end-to-end orchestration and scheduling method as for stateful service requests would result in unnecessary processing redundancy, increased end-to-end response latency, and failure to fully leverage the network controller's advantages in real-time network status awareness. Therefore, this embodiment introduces a request-type-based differentiated processing mechanism in the orchestration and scheduling module.
[0111] Figure 3 This is a schematic diagram illustrating the transparent transmission processing flow of an orchestration and scheduling module under a stateless service request, provided as an embodiment of this application. Figure 3 As shown, the method includes the following steps:
[0112] Step 301: Extract network-related information from the structured requirement information. The network-related information includes attributes related to network location in the node information and attributes related to network transmission in the connection information.
[0113] Step 302: Network-related information, linkage rule information, and latency quota information are transmitted as transparent information. After adding a transparent identifier to the transparent information, it is sent to the network controller. The transparent identifier is used to instruct the network controller to select a service instance independently according to the actual network status and to verify it according to the linkage rule information and latency quota information.
[0114] First, the orchestration and scheduling module extracts network-related information from the structured demand information. In this embodiment, the node information includes both attributes related to network location and attributes related to computing power; while the connection information includes attributes related to network transmission.
[0115] As can be seen from the above embodiments, for stateless service requests, computing power-related attributes (such as the number of CPU cores and memory size) are unnecessary because stateless services do not need to be deployed on specific nodes and occupy computing resources. Service instances are pre-deployed, and users only need to access them, not deploy them. Therefore, when the orchestration and scheduling module performs the extraction operation, it removes computing power-related fields from the node information and retains only information related to network scheduling.
[0116] Optionally, the specific rules for pruning can be predefined in a configuration file. For example, a "whitelist of fields to be retained in stateless service request scenarios" or a "blacklist of fields to be removed" can be pre-defined, and the orchestration and scheduling module will automatically execute the pruning operation based on this configuration. This allows for flexible adjustment of the pruning strategy according to different business scenarios without requiring modification of the code logic.
[0117] Then, the orchestration and scheduling module will extract the network-related information, along with the constraint decomposition information (linkage rule information and latency quota information), and assemble them into transparent transmission information.
[0118] During transparent transmission, the orchestration and scheduling module does not perform any orchestration, scheduling, or resource selection processing on this information; instead, it transmits it to the network controller as is (or after necessary format conversion). The constraint decomposition information needs to be transmitted along with this information because, although global constraints (such as end-to-end latency ≤ 10ms) have been structurally decomposed on the orchestration and scheduling module side, the network controller still needs to verify these constraints when selecting service instances. For example, after selecting a service instance, the network controller needs to verify whether the transmission latency of the selected path meets the decomposed latency quota requirements.
[0119] Optionally, the format of the transparent information can adopt a structured data format such as JSON or Protobuf to ensure that the data exchange between the orchestration and scheduling module and the network controller has good parsing and scalability. The transparent information should include at least the following fields: topology information (node location, connection relationship), latency quota (upper limit of transmission latency for each connection segment), and linkage rules (constraint relationships between nodes and connections).
[0120] Finally, the orchestration and scheduling module adds a pass-through identifier to the assembled pass-through information, and then sends the pass-through information with the pass-through identifier to the network controller.
[0121] In this embodiment, the purpose of the pass-through identifier is to inform the network controller that this request is stateless and the orchestration and scheduling module has not made any resource selection. The network controller should select a service instance independently based on the actual network status and verify it according to the linkage rule information and latency quota information.
[0122] After receiving the transparent transmission information with the transparent transmission identifier, the network controller can autonomously select a service instance and perform verification operations based on constraints. Specifically, the network controller first selects one or more optimal service instances from the available service instance set to respond to the user request based on the current real-time network status (such as the actual transmission latency of each link, bandwidth utilization, node load, etc.). After selecting the service instance and the corresponding network path, the network controller verifies whether the selected scheme meets the decomposed constraints based on the linkage rule information and latency quota information in the transparent transmission information—for example, verifying whether the end-to-end latency exceeds the quota limit.
[0123] Figure 4 This is a flowchart illustrating a method for generating global constraint quota decomposition and scheduling filtering conditions, provided in an embodiment of this application. Figure 4 As shown, the method includes:
[0124] Step 401: Based on the service path represented by the connection information, allocate the total network transmission delay quota to each connection segment, generate transmission delay constraints for each connection segment, and use the transmission delay constraints of each connection segment as the filtering condition for network path selection.
[0125] Step 402: Use the total node processing latency quota as the cumulative upper limit constraint for node processing latency when selecting nodes.
[0126] In this embodiment, after decomposing the global constraint information to obtain the total node processing delay quota and the total network transmission delay quota, these two quotas are used as the filtering conditions for subsequent node selection and network path selection.
[0127] Specifically: During the node selection phase, the total node processing delay quota is used as the upper limit constraint that the cumulative processing delay of all candidate nodes must not exceed—that is, the sum of the calculated processing delays of all nodes on the selected path must be less than or equal to the total quota; During the network path selection phase, the total network transmission delay quota is allocated according to the number of links in the service path (either by equal distribution or by weight), generating transmission delay constraint thresholds for each connection segment—that is, the actual transmission delay of each connection segment must be less than or equal to its allocated quota threshold.
[0128] Through the above quota-based decomposition, the global end-to-end latency constraint is transformed into quantifiable screening thresholds. During the scheduling process, the orchestration and scheduling module filters candidate computing power nodes and candidate network paths based on these quota thresholds, selecting only candidate schemes that can simultaneously meet the cumulative upper limit of node processing latency and the transmission latency thresholds of each segment connection to enter the subsequent scheduling stage.
[0129] In one optional embodiment, the computing network scheduling instructions include computing power scheduling instructions and network scheduling instructions.
[0130] Figure 5 This is a schematic flowchart illustrating a method for generating computer network scheduling instructions, provided in an embodiment of this application. Figure 5 As shown, the method includes the following steps.
[0131] Step 501: Determine the target node that satisfies the transmission delay constraint and cumulative upper limit constraint from the candidate node set corresponding to the node information, and allocate computing resources to the target node according to the node information.
[0132] Step 502: Select the target network path based on the location of the target node, and configure network parameters for the target network path based on the connection information.
[0133] Step 503: Associate the linkage rule information with the target node and the target network path.
[0134] Step 504: Generate computing power scheduling instructions and network scheduling instructions based on the target node, allocated computing power resources, target network path, configured network parameters, and associated linkage rule information.
[0135] In the context of the intelligent agent Internet, for stateful service requests, the orchestration and scheduling module needs to execute a complete orchestration and scheduling process, which includes the orchestration and scheduling of computing resources, the orchestration and scheduling of network resources, the binding of coupling rules, and the generation of execution instructions.
[0136] First, the orchestration and scheduling module makes a comprehensive decision based on the constraint decomposition information obtained in the preceding steps, combined with the real-time status of the computing network. Specifically, it selects candidate nodes from the set of candidate nodes corresponding to the node information that meet the transmission delay constraint and the cumulative upper limit constraint of node processing delay, and finally determines the target node by combining the real-time computing load, resource utilization, and other status information of each candidate node. After determining the target node, the orchestration and scheduling module allocates specific computing resources to the target node according to the computing resource attributes (such as the number of CPU cores, memory size, etc.) recorded in the node information.
[0137] After determining the location of the target node, the orchestration and scheduling module needs to select a specific network path. Specifically, based on the deployment location of the target node, it determines the target network path connecting each relevant node from the available network links, and configures specific network parameters for the target network path according to the network attributes recorded in the connection information (such as network protocol type, bandwidth requirements, etc.).
[0138] After completing node selection and network path selection, the orchestration and scheduling module also needs to bind and associate the linkage rule information obtained from the above embodiments with the selected target node and target network path. This association step ensures that when the actual operating attributes of the selected node or connection exceed the preset constraint threshold, the system can execute the corresponding linkage adjustment action according to the linkage rules, thereby ensuring that the global constraints are propagable, verifiable, and satisfyable during the end-to-end execution process.
[0139] Based on the determined target nodes, allocated computing resources, selected target network paths, configured network parameters, and associated linkage rules, the orchestration and scheduling module generates computing power scheduling instructions for the computing power controller and network scheduling instructions for the network controller. The computing power scheduling instructions include target node information and a specific computing power resource allocation scheme, while the network scheduling instructions include target network path information and a specific network parameter configuration scheme. The orchestration and scheduling module sends these computing power scheduling instructions and network scheduling instructions to the computing power controller and network controller, respectively, enabling them to execute specific computing and network resource allocation operations according to the received instructions.
[0140] In practical applications, global constraints on user requirements (e.g., end-to-end latency ≤ 10ms) typically cannot be directly bound to specific service deployment locations or network paths before scheduling. Therefore, the aforementioned embodiments only decompose them into structured quota thresholds, such as the total node processing latency quota and the total network transmission latency quota, and use them as threshold conditions for filtering candidate nodes and candidate network paths. However, once the target node and target network path are finally determined, the orchestration and scheduling module needs to further refine the above quota thresholds into executable latency parameters so that the computing power controller and network controller can accurately allocate and configure resources accordingly.
[0141] Figure 6 This is a schematic diagram illustrating the process of converting a global constraint quota into an executable latency parameter, as provided in an embodiment of this application. Figure 6 As shown, the method includes the following steps.
[0142] Step 601: Based on the actual computing power of each target node, allocate the total processing latency quota of the node to each target node, and generate the executable computing latency parameters for each target node.
[0143] Step 602: Based on the actual transmission delay of each segment connection in the target network path, allocate the total network transmission delay quota to each segment connection, and generate executable network delay parameters for each segment connection.
[0144] Step 603: The executable computation delay parameter and the executable network delay parameter are respectively carried in the computing power scheduling instruction and the network scheduling instruction.
[0145] In this embodiment, on the node side, the orchestration and scheduling module allocates the total processing latency quota to each target node based on the actual computing power of each target node (e.g., CPU clock speed, current load, instruction execution efficiency, etc.), generating an executable computing latency parameter for each target node. This parameter represents the upper limit of computing latency allowed for the node to process the allocated services, and serves as the basis for the computing power controller to perform process scheduling, thread allocation, and computing resource isolation during the execution phase.
[0146] On the network side, the orchestration and scheduling module allocates the total network transmission delay quota to each connection segment based on the actual transmission delay of each connection segment in the target network path (e.g., based on current link measurement results, historical statistical delay, or real-time status reported by the network controller), generating executable network delay parameters for each connection segment. This parameter represents the upper limit of transmission delay allowed for that link segment when transmitting service data, and serves as the basis for the network controller to perform path adjustment, traffic shaping, or queue scheduling during the execution phase.
[0147] Finally, the orchestration and scheduling module sends the aforementioned executable computation latency parameters to the computing power controller in the computing power scheduling command, and sends the executable network latency parameters to the network controller in the network scheduling command. Through this transformation, global constraints gradually converge from abstract thresholds before scheduling to specific parameters that are understandable, configurable, and guaranteed by each execution entity, thereby ensuring that global constraints such as end-to-end latency have a deterministic executable basis in actual operation.
[0148] In one optional embodiment, during the network-computer collaborative scheduling process, the orchestration and scheduling module also has feedback adjustment capabilities, which can dynamically adjust the scheduling strategy according to the actual execution of the network controller or computing power controller.
[0149] Figure 7 This is a schematic flowchart of a computer network collaborative scheduling feedback adjustment method provided in an embodiment of this application. Figure 7 As shown, the method includes the following steps:
[0150] Step 701: Receive feedback signals generated by the network controller or computing power controller based on the computing network scheduling instructions. The feedback signals are used to characterize the actual network status or actual computing power resource status, which cannot meet the linkage rule information or latency quota information.
[0151] Step 702: In response to the feedback signal, adjust the split ratio of the global constraint information or perform expansion / shrinkage operations, and re-execute node selection and network path selection according to the adjusted split ratio, generate new computing network scheduling instructions and issue them.
[0152] In this embodiment, after the orchestration and scheduling module sends the network scheduling instruction to the network controller and the computing power controller, the network controller and the computing power controller execute specific network resource scheduling operations and computing power resource scheduling operations according to the received instructions.
[0153] During execution, the network controller verifies the network resource allocation scheme in the scheduling command based on the actual network status, such as determining whether the actual network links can meet the transmission delay or bandwidth constraints carried in the command. The computing power controller verifies the computing power resource allocation scheme in the scheduling command based on the actual computing power resource status, such as determining whether the actual available computing power of the target node can meet the resource requirements allocated in the command. When the actual network status or the actual computing power resource status cannot meet the linkage rule information or delay quota information carried in the scheduling command, the network controller or the computing power controller generates a feedback signal and sends the feedback signal to the orchestration and scheduling module.
[0154] Upon receiving the aforementioned feedback signal, the orchestration and scheduling module responds by executing corresponding adjustment operations. Specifically, based on the actual state information carried in the feedback signal, the module can adjust the splitting ratio used in the previous steps when decomposing the global constraint information. For example, it can adjust the allocation ratio between the total node processing latency quota and the total network transmission latency quota, or adjust the latency quota allocation weights between different connection segments, so that the adjusted quota thresholds better reflect the current actual state of the computing network. Simultaneously, the module can also perform scaling up or down operations based on insufficient computing or network resources. For example, it can trigger horizontal or vertical scaling up of computing resources, or trigger expansion of network bandwidth resources, to bridge the gap between the current resource status and the constraint requirements.
[0155] After completing the above adjustment operations, the orchestration and scheduling module re-executes node selection and network path selection based on the adjusted split ratio or the resource status after scaling up or down, generates new computing network scheduling instructions, and redistributes the new computing network scheduling instructions to the network controller and computing power controller. Through the above feedback adjustment mechanism, this embodiment can promptly detect and resolve constraint non-compliance issues caused by dynamic changes in computing network status during scheduling execution, ensuring that end-to-end global constraints are always propagated, verifiable, and satisfied during scheduling, thereby significantly improving the determinism and reliability of multi-agent collaboration.
[0156] Figure 8 This is a flowchart illustrating another computer network collaborative scheduling method provided in an embodiment of this application. The following is a summary of the process. Figure 8 The present application provides a detailed description of the computer network collaborative scheduling method provided in the embodiments.
[0157] like Figure 8 As shown in the embodiments of this application, the computer network collaborative scheduling method includes the following process.
[0158] Perform standardized parsing of business request information:
[0159] The orchestration and scheduling device receives service request information initiated by terminal devices, performs standardized parsing of the service request information, extracts the participating entities, interaction relationships, and global constraints from the user requirements, and generates structured requirement information. This structured requirement information includes node information, connection information between nodes, and global constraint information. Node information represents the various entities participating in this service request and their attributes (such as node location, computing resource requirements, etc.), connection information represents the interaction relationships between nodes and network link attributes (such as network protocols, bandwidth requirements, etc.), and global constraint information represents the user's overall requirements for the entire end-to-end execution process of the service request (such as end-to-end latency constraints).
[0160] For example, after receiving a service request for "accessing a certain intelligent agent service with an end-to-end latency of ≤10ms", the orchestration and scheduling device parses it into the following structured information: node information includes user terminal node N001 and service node N002; connection information includes access link C001 from N001 to N002; global constraint information includes end-to-end latency constraints, the coupling formula of which is E2E_delay = ∑node.compute_delay + ∑connect.network_delay, with a threshold of ≤10ms.
[0161] Global constraint quota decomposition generates node / connection / linkage rules:
[0162] The orchestration and scheduling device decomposes global constraint information into quotas, transforming global constraints from "top-level abstract requirements" into "implementable attribute thresholds and executable linkage rules".
[0163] Continuing with the example above, the orchestration and scheduling device decomposes the end-to-end latency constraint into a fixed structure "E2E = N001 + C001 + N002 ≤ 10ms", and then allocates quotas according to a preset strategy: the total quota for node processing latency is 4ms, and the total quota for network transmission latency is 6ms. Simultaneously, it generates linkage rule information, such as "when the transmission latency of C001 increases, the processing latency of N002 is compressed accordingly," as the basis for subsequent verification and dynamic adjustment.
[0164] Based on the above decomposition method, even when the service deployment location and network status are unknown, it is still possible to ensure that global constraints such as end-to-end latency are propagated, verifiable, and satisfied during the scheduling process.
[0165] Request type determination and differential processing:
[0166] The orchestration and scheduling unit determines the request type based on the service request information. The request type includes stateless service requests or stateful service requests, and executes differentiated processing procedures according to the request type.
[0167] (a) Stateless service requests - pass-through processing.
[0168] If the request type is a stateless service request (such as accessing an agent service), the orchestration and scheduling device performs transparent transmission processing: it removes computing power-related fields from the structured demand information, retains only network-related information, adds a transparent transmission identifier, and then sends the transparent transmission information to the network controller. After receiving the transparent transmission information, the network controller autonomously selects the service instance and network path based on the actual network status, and performs verification based on the constraint decomposition information in the transparent transmission information to complete network scheduling and execution.
[0169] In this processing method, there is no interaction between the orchestration and scheduling device and the computing power controller, which ensures efficient response to lightweight, stateless services with high real-time requirements.
[0170] Stateful service requests – end-to-end orchestration and scheduling:
[0171] If the request type is a stateful service request (such as deploying an AI inference service, which requires 16 CPU cores, 64GB of memory, and an end-to-end latency of ≤10ms), the orchestration and scheduling device performs full-process orchestration and scheduling: First, based on the cumulative upper limit constraint of node processing latency and the transmission latency constraint of each connection segment, the target node is selected from the candidate node set and specific computing resources are allocated to it; Second, the target network path is selected based on the location of the target node, and network parameters are configured; Then, the linkage rule information is associated and bound with the selected target node and target network path; Finally, a computing network scheduling instruction containing the node selection result and the network path selection result is generated and sent to the computing power controller and the network controller respectively, and the computing power controller and the network controller perform specific resource allocation and configuration operations.
[0172] Furthermore, after the target nodes and target network paths are determined, the orchestration and scheduling device further refines the quota threshold into executable latency parameters: based on the actual computing power of each target node, the total node processing latency quota is allocated to each target node, generating executable computing latency parameters for each target node and carrying them in the computing power scheduling instructions; based on the actual transmission latency of each connection segment in the target network path, the total network transmission latency quota is allocated to each connection segment, generating executable network latency parameters for each connection segment and carrying them in the network scheduling instructions. Thus, the global constraints gradually converge from abstract thresholds before scheduling to specific parameters that are understandable, configurable, and guaranteed by each executing entity.
[0173] Feedback adjustment:
[0174] In actual execution, when the network controller or computing power controller finds that the actual network status or actual computing power resource status cannot meet the linkage rule information or latency quota information, it generates a feedback signal and sends it to the orchestration and scheduling device. In response to the feedback signal, the orchestration and scheduling device adjusts the splitting ratio of the global constraint information (such as adjusting the allocation ratio of the total node processing latency quota to the total network transmission latency quota), or performs scaling up or down operations. Then, based on the adjusted splitting ratio, it re-executes node selection and network path selection, generates new network scheduling instructions, and reissues them.
[0175] Through the above feedback adjustment mechanism, this embodiment can promptly detect and resolve constraint non-compliance issues caused by dynamic changes in the computing network state during the scheduling and execution process, ensuring that end-to-end global constraints are always propagated, verifiable, and satisfied.
[0176] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
[0177] In some embodiments, this application also provides a computer network collaborative scheduling device. This device may include one or more functional modules for implementing the computer network collaborative scheduling method described in the above embodiments.
[0178] For example, Figure 9 This is a schematic diagram of the structure of a computer network collaborative scheduling device provided in an embodiment of this application. Figure 9 As shown, the computer network collaborative scheduling device 900 includes: an orchestration and scheduling module 901.
[0179] The aforementioned orchestration and scheduling module 901 is used to determine the request type and structured requirement information based on the service request information. The structured requirement information includes node information, connection information between nodes, and global constraint information. The request type includes stateless service requests or stateful service requests. The orchestration and scheduling module decomposes the global constraint information to obtain constraint decomposition information. In the case of a stateless service request, the orchestration and scheduling module transmits the node information, connection information, and constraint decomposition information to the network controller, enabling the network controller to perform network scheduling based on the structured requirement information. In the case of a stateful service request, the orchestration and scheduling module selects nodes and network paths based on the node information, connection information, and constraint decomposition information, generates a computing network scheduling instruction containing the node selection results and network path selection results, and sends the computing network scheduling instruction to the computing power controller and the network controller, enabling the computing power controller and the network controller to allocate computing network resources according to the computing network scheduling instruction.
[0180] In one optional embodiment, the constraint decomposition information includes linkage rule information and latency quota information.
[0181] The orchestration and scheduling module 901 is specifically used to decompose and process the global constraint information to obtain linkage rule information and latency quota information. The linkage rule information includes the processing latency linkage rules between nodes, the bandwidth linkage rules between each segment connection, and the global dynamic constraint linkage rules. The latency quota information includes the total node processing latency quota and the total network transmission latency quota.
[0182] In an optional embodiment, the orchestration and scheduling module 901 is specifically used to extract network-related information from the structured demand information. The network-related information includes attributes related to network location in the node information and attributes related to network transmission in the connection information. The network-related information, linkage rule information, and latency quota information are used as transparent transmission information. After adding a transparent transmission identifier to the transparent transmission information, it is sent to the network controller. The transparent transmission identifier is used to instruct the network controller to autonomously select a service instance according to the actual network status and to verify it according to the linkage rule information and latency quota information.
[0183] In an optional embodiment, before selecting nodes and network paths based on node information, connection information, and constraint decomposition information, the orchestration and scheduling module 901 is specifically used to allocate the total network transmission delay quota to each connection segment according to the service path represented by the connection information, generate transmission delay constraints for each connection segment, and use the transmission delay constraints for each connection segment as the filtering condition for network path selection; and use the total node processing delay quota as the cumulative upper limit constraint of node processing delay when selecting nodes.
[0184] In one optional embodiment, the computing network scheduling instructions include computing power scheduling instructions and network scheduling instructions.
[0185] The orchestration and scheduling module 901 is specifically used to determine the target node that meets the transmission delay constraint and cumulative upper limit constraint from the candidate node set corresponding to the node information, and allocate computing resources to the target node according to the node information; select the target network path according to the location of the target node, and configure network parameters for the target network path according to the connection information; associate the linkage rule information with the target node and the target network path; and generate computing power scheduling instructions and network scheduling instructions according to the target node, the allocated computing power resources, the target network path, the configured network parameters, and the associated linkage rule information.
[0186] In an optional embodiment, after selecting nodes and network paths based on node information, connection information, and constraint decomposition information, and generating a computing network scheduling instruction containing node selection results and network path selection results, the orchestration and scheduling module 901 is specifically used to allocate the total node processing delay quota to each target node according to the actual computing power of each target node, and generate executable computing delay parameters for each target node; allocate the total network transmission delay quota to each segment connection according to the actual transmission delay of each segment connection in the target network path, and generate executable network delay parameters for each segment connection; and carry the executable computing delay parameters and executable network delay parameters in the computing power scheduling instruction and the network scheduling instruction, respectively.
[0187] In an optional embodiment, the orchestration and scheduling module 901 is further configured to receive a feedback signal generated by the network controller or computing power controller based on the computing network scheduling instruction. The feedback signal is used to characterize the actual network status or actual computing power resource status as failing to meet the linkage rule information or latency quota information. In response to the feedback signal, the module adjusts the splitting ratio of the global constraint information or performs scaling up or down operations, and re-executes node selection and network path selection according to the adjusted splitting ratio, generates a new computing network scheduling instruction, and issues it.
[0188] It should be noted that the computer network collaborative scheduling device can implement all the processes implemented in the above method embodiments and achieve the same beneficial effects. To avoid repetition, it will not be described again here.
[0189] In the case where the functions of the integrated modules described above are implemented in hardware, this application provides a possible structural schematic diagram of the electronic device involved in the above embodiments. For example... Figure 10 As shown, the electronic device 1000 includes: a processor 1002, a communication interface 1003, and a bus 1004. Optionally, the electronic device 1000 may also include a memory 1001.
[0190] Processor 1002 may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 1002 may be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 1002 may also be a combination that implements computing functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.
[0191] Communication interface 1003 is used to connect with other devices via a communication network. This communication network can be Ethernet, wireless access network, wireless local area network (WLAN), etc.
[0192] The memory 1001 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.
[0193] As one possible implementation, the memory 1001 can exist independently of the processor 1002. The memory 1001 can be connected to the processor 1002 via the bus 1004 and is used to store instructions or program code. When the processor 1002 calls and executes the instructions or program code stored in the memory 1001, it can implement the computer network collaborative scheduling method provided in this application embodiment.
[0194] In another possible implementation, the memory 1001 can also be integrated with the processor 1002.
[0195] Bus 1004 can be an Extended Industry Standard Architecture (EISA) bus, etc. Bus 1004 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 10 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.
[0196] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer network collaborative scheduling device can be divided into different functional modules to complete all or part of the functions described above.
[0197] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described computer network collaborative scheduling method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be described again here.
[0198] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.
[0199] This application also provides a readable storage medium storing a program or instructions. When executed by a computer, the program or instructions implement the computer network collaborative scheduling method provided in the above embodiments. It is understood that all or part of the processes in the above method embodiments can be executed by computer instructions instructing related hardware. The readable storage medium can be any of the foregoing embodiments or memory. The readable storage medium can also be an external storage device of the computer network collaborative scheduling device, such as a plug-in hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer network collaborative scheduling device. Further, the readable storage medium can include both internal storage units and external storage devices of the computer network collaborative scheduling device. The readable storage medium is used to store the computer program and other programs and data required by the computer network collaborative scheduling device. The readable storage medium can also be used to temporarily store data that has been output or will be output.
[0200] This application also provides a computer program product, which is stored in a storage medium and implements the computer network collaborative scheduling method provided in the above embodiments when the computer program product is executed by a computer.
[0201] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.
[0202] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.
[0203] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.
Claims
1. A computer network collaborative scheduling method, characterized in that, include: The orchestration and scheduling module determines the request type and structured requirement information based on the business request information. The structured requirement information includes node information, connection information between nodes, and global constraint information. The request type includes stateless service requests or stateful service requests. The global constraint information is decomposed by the orchestration and scheduling module to obtain constraint decomposition information; When the request type is the stateless service request, the node information, the connection information and the constraint decomposition information are transmitted to the network controller through the orchestration and scheduling module, so that the network controller performs network scheduling according to the structured requirement information. When the request type is a stateful service request, the orchestration and scheduling module selects nodes and network paths based on the node information, connection information, and constraint decomposition information, generates a computing network scheduling instruction containing the node selection results and network path selection results, and sends the computing network scheduling instruction to the computing power controller and the network controller, so that the computing power controller and the network controller allocate computing network resources according to the computing network scheduling instruction.
2. The method according to claim 1, characterized in that, The constraint decomposition information includes linkage rule information and time delay quota information; The step of decomposing the global constraint information to obtain constraint decomposition information includes: The orchestration and scheduling module decomposes the global constraint information to obtain the linkage rule information and the delay quota information. The linkage rule information includes processing delay linkage rules between nodes, bandwidth linkage rules between different segments, and global dynamic constraint linkage rules; the delay quota information includes the total node processing delay quota and the total network transmission delay quota.
3. The method according to claim 2, characterized in that, The step of transmitting the node information, the connection information, and the constraint decomposition information to the network controller includes: Network-related information is extracted from the structured requirements information, including network location-related attributes in the node information and network transmission-related attributes in the connection information; The network-related information, the linkage rule information, and the latency quota information are transmitted as transparent information. A transparent identifier is added to the transparent information and then sent to the network controller. The transparent identifier is used to instruct the network controller to autonomously select a service instance based on the actual network status and to verify it based on the linkage rule information and the latency quota information.
4. The method according to claim 2, characterized in that, Before selecting nodes and network paths based on the node information, the connection information, and the constraint decomposition information, the process further includes: Based on the service path represented by the connection information, the total network transmission delay quota is allocated to each connection segment, generating transmission delay constraints for each connection segment, and using these constraints as filtering conditions for network path selection; and The total node processing latency quota is used as the cumulative upper limit constraint for node processing latency when selecting a node.
5. The method according to claim 4, characterized in that, The computing network scheduling instructions include computing power scheduling instructions and network scheduling instructions; The step of selecting nodes and network paths based on the node information, the connection information, and the constraint decomposition information, and generating a network scheduling instruction containing node selection results and network path selection results, includes: From the candidate node set corresponding to the node information, determine the target node that satisfies the transmission delay constraint and the cumulative upper limit constraint, and allocate computing resources to the target node according to the node information; Select a target network path based on the location of the target node, and configure network parameters for the target network path based on the connection information; Associate the linkage rule information with the target node and the target network path; Based on the target node, the allocated computing resources, the target network path, the configured network parameters, and the associated linkage rule information, computing power scheduling instructions and network scheduling instructions are generated.
6. The method according to claim 5, characterized in that, After performing node selection and network path selection based on the node information, connection information, and constraint decomposition information, and generating a network scheduling instruction containing node selection results and network path selection results, the method further includes: Based on the actual computing power of each target node, the total processing latency quota of the node is allocated to each target node, and the executable computing latency parameters of each target node are generated. Based on the actual transmission delay of each segment of the target network path, the total network transmission delay quota is allocated to each segment of the connection, and executable network delay parameters for each segment of the connection are generated. The executable computation latency parameter and the executable network latency parameter are respectively carried in the computing power scheduling instruction and the network scheduling instruction.
7. The method according to claim 2, characterized in that, Also includes: Receive a feedback signal generated by the network controller or the computing power controller based on the computing network scheduling instruction. The feedback signal is used to characterize that the actual network status or the actual computing power resource status cannot meet the linkage rule information or the latency quota information. In response to the feedback signal, the splitting ratio of the global constraint information is adjusted or a scaling up / down operation is performed. Based on the adjusted splitting ratio, node selection and network path selection are re-executed, and a new computing network scheduling instruction is generated and issued.
8. A computer network collaborative scheduling device, characterized in that, include: An orchestration and scheduling module, wherein the orchestration and scheduling module is used to perform the method as described in any one of claims 1 to 7.
9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing programs or instructions that can run on the processor, the programs or instructions being executed by the processor to implement the computer network collaborative scheduling method as described in any one of claims 1-7.
10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions, which, when executed by a computer, implement the computer network collaborative scheduling method as described in any one of claims 1-7.
11. A computer program product, characterized in that, The computer program product is stored in a storage medium, and when the computer program product is executed by a computer, it implements the computer network collaborative scheduling method as described in any one of claims 1-7.
