Hierarchical control network architecture and multi-domain resource collaborative management method based on industrial SDN

Abstract

The application discloses a layered control network architecture based on an industrial software-defined network (SDN), proposes a double-time-scale network slicing method and a multi-domain resource collaborative management scheme, and has a significant industrial application prospect. The layered control architecture allows flexible configuration of the SDN controller according to specific industrial environments, and realizes multi-level resource management. The SDN controllers at different levels can be dynamically adjusted according to actual needs, and adapt to complex and changeable industrial scenes. The double-time-scale network slicing method decouples resource reservation and orchestration, improves the efficiency of resource scheduling. By monitoring the network state in real time, optimizing resource allocation, and ensuring the satisfaction of heterogeneous QoS requirements, the utilization rate of resources is significantly improved.

Description

Technical Field

[0001] This invention mainly relates to the field of communication technology, and in particular to a hierarchical control network architecture based on industrial SDN, a dual-time-scale network slicing method, and a multi-domain resource collaborative management and control method. Background Technology

[0002] Network systems in modern industrial environments are becoming increasingly complex, placing higher demands on the management and allocation of network resources. To ensure the heterogeneous Quality of Service (QoS) requirements of different types of industrial applications, network slicing technology has gradually become a research hotspot in both academia and industry. Network slicing divides physical network resources into multiple virtual networks on demand, enabling flexible scheduling and management of resources to adapt to the needs of different applications. However, existing technologies still face some key challenges in dealing with dynamic and ever-changing industrial environments.

[0003] In industrial internet scenarios, networks are massive in scale, with numerous nodes, and each node has significantly different business requirements. Traditional single-controller architectures face many challenges, such as high control latency, high signaling overhead, and difficulty in adapting to large-scale dynamic network changes. Furthermore, existing network slicing research mainly focuses on the independent optimization of resource slicing and scheduling, failing to explore their joint optimization. Due to the interdependence between resource slicing and scheduling decisions, neglecting their joint optimization leads to low resource utilization efficiency and makes it difficult to fully meet the heterogeneous QoS requirements in industrial scenarios.

[0004] Meanwhile, most existing network slicing solutions employ static or historical data-based optimization methods, which are ill-suited to the spatiotemporal variations of network conditions in industrial settings. For instance, communication links in factories may be constrained by physical environments, equipment movement, and other factors, leading to highly uncertainties in resource demands and service quality. Against this backdrop, achieving efficient resource allocation and scheduling while ensuring service quality has become a pressing issue in the Industrial Internet.

[0005] To address the aforementioned issues, Software-Defined Networking (SDN), as an emerging network architecture, offers more flexible network management and control by separating the control plane and data plane. In industrial environments, SDN architecture can centrally control network resources, making network management and control more efficient. However, existing SDN technologies typically employ a single controller, making it difficult to adapt to the resource coordination needs of large-scale industrial environments. Furthermore, the scalability and flexibility of traditional SDN architectures are limited when facing multi-domain, multi-layered resource management. Therefore, a layered control network architecture based on industrial SDN has become an important direction for solving resource management problems in large-scale network environments.

[0006] To address the aforementioned issues, this invention proposes a hierarchical control network architecture based on industrial SDN, and further proposes a dual-timescale network slicing method and a multi-domain resource collaborative management and control scheme. Through this hierarchical architecture, SDN controllers at different levels can be responsible for resource management and control within different scopes, achieving efficient resource allocation and scheduling. The dual-timescale slicing method improves the flexibility and response speed of resource scheduling by decoupling resource reservation and resource orchestration. Furthermore, the multi-domain resource collaborative management and control scheme further optimizes the coordination and management of cross-domain resources, enhancing the system's scalability and stability. Summary of the Invention

[0007] In view of the above-mentioned deficiencies of the prior art, the technical problem to be solved by the present invention includes:

[0008] How to design a hierarchical control network architecture, network slicing method, and resource co-management method to solve the above-mentioned technical problems.

[0009] To achieve the above objectives, the present invention provides a hierarchical control network architecture based on industrial SDN, including a terminal-side SDN controller, an edge-side SDN controller, and an industrial cloud-side SDN controller;

[0010] The terminal-side SDN controller is deployed at the access point, switch, and gateway, with a relatively small coverage area, and is responsible for resource scheduling for terminal users.

[0011] The edge-side SDN controller is responsible for coordinating multiple terminal-side SDN controllers to achieve higher-level network operations;

[0012] The industrial cloud-side SDN controller, as the highest-level controller, obtains network status information provided by the lower-level controllers to formulate global rules and coordinate across domains for the entire network.

[0013] A dual-timescale network slicing method based on the aforementioned network architecture divides resource allocation into different time domains for processing, including resource reservation methods and resource orchestration methods.

[0014] Furthermore, the resource reservation method specifically involves the following steps: at the beginning of each slice window w, the industrial cloud-side SDN controller makes a resource reservation decision based on network status feedback information from the edge-side SDN controller and global service requirements. Where ε represents the edge-side SDN controller, and s represents the index of the network slice.

[0015] Furthermore, the resource reservation method also includes: the reservation decision remains unchanged throughout the entire network slice window; after each network slice window ends, the industrial cloud-side SDN controller evaluates system performance through feedback and adjusts the resource allocation decision for the next network slice window based on changes in business needs.

[0016] Furthermore, the resource orchestration method specifically involves the edge-side SDN controller scheduling resources in real time within each network slice window based on the spatiotemporal changes in network topology, link status, and service requirements.

[0017] Furthermore, the resource orchestration method specifically involves: real-time scheduling of resources, which specifically involves allocating network resources to the terminal-side SDN controller and flexibly adjusting resources between different time slots.

[0018] Furthermore, the resource orchestration method primarily focuses on the correlation between the edge-side SDN controller and the terminal-side SDN controller and the efficiency of resource allocation.

[0019] A multi-domain resource collaborative management method based on the aforementioned dual-timescale network slicing method includes the following steps:

[0020] Step 1: Based on the current network status and resource requirements, decompose the dual-timescale network slicing problem into a resource reservation subproblem and a resource orchestration subproblem, and decompose the relevant constraints accordingly.

[0021] Step 2: At the resource orchestration layer, due to the isolation of slices, a resource orchestration algorithm based on multi-knapsack matching is adopted.

[0022] Step 3: In the resource reservation layer, the resource orchestration strategy is substituted as an incentive into the resource reservation subproblem, which is transformed into a closed expression that is only related to the reservation ratio, and then solved using the iterative block coordinate gradient descent algorithm.

[0023] Furthermore, step 2 specifically includes:

[0024] Step 2.1: Construct a bipartite graph G = (Γ) t ,E t ,Ξ t ), where Γ t It is a collection of sampling terminals, E t It is a collection of edge-side SDN controllers, Ξ t It is a connection Γ t and E t Given the set of edges of the mid-vertex; based on the expected service delay and relevant network resource constraints, give the multi-domain network resources required to satisfy the requirements under a given association decision <γ,ε>.

[0025]

[0026] in, The required communication, computing, and storage resources are further defined as follows: γ represents the terminal-side SDN controller, ε represents the edge-side SDN controller, b represents communication resources, c represents computing resources, m represents storage resources, and t represents time slots.

[0027] Step 2.2: For each terminal-side SDN controller, determine the required resources according to different associations. The edge-side SDN controllers are sorted in ascending order to construct a preference list, that is, the terminal-side SDN controllers prefer edge-side SDN controllers with higher resource utilization.

[0028] Step 2.3: Each terminal-side SDN controller sends a request to its currently preferred edge-side SDN controller and then removes that edge-side SDN controller from the preference list;

[0029] Step 2.4: Each edge-side SDN controller examines all received requests, including new requests and requests accepted in previous iterations; given a set of proposals, each request has a required amount of resources. and a priority weight p γ Each edge-side SDN controller needs to determine which requests to accept in order to avoid violating the total weight constraint. Maximize total value under the following circumstances;

[0030] Step 2.5: For all unqualified terminal-side SDN controllers, proceed to step 2.3; the matching process ends when all terminal-side SDN controllers are matched with edge-side SDN controllers or when all available resources of edge-side SDN controllers are allocated.

[0031] Furthermore, step 3 specifically includes:

[0032] Step 3.1: During the initialization phase, randomly generate initial resource reservation values ​​that meet the requirements for each slice. And set the learning factor α and the number of iterations n = 0;

[0033] Step 3.2: Calculate the gradient of the objective function based on the current resource reservation decision and network state. And update the reserved variables. Ensure that the updated reserved values ​​are within a valid range;

[0034] Step 3.3: Calculate the difference in the objective function values ​​of the current iteration. If the number of iterations is less than the set convergence threshold or the current iteration count reaches the maximum iteration count, then stop the iteration; otherwise, return to step 3.2 and proceed to the next iteration.

[0035] Compared with existing technical solutions, the technical advantages of the present invention are as follows:

[0036] This invention, by introducing a hierarchical SDN controller architecture, effectively improves the scalability and response speed of SDN networks in industrial environments. The multi-layered controller can flexibly schedule resources according to different communication scenarios and QoS requirements, reducing signaling overhead and optimizing network status acquisition efficiency. Furthermore, this architecture ensures efficient collaborative management and control of cross-domain resources, thereby improving overall network performance and quality of service. This hierarchical control approach solves the inefficiency problem of a single SDN controller managing a large range of resources in existing technologies, significantly improving network performance in complex industrial scenarios.

[0037] This invention introduces a dual-time-scale network slicing method to achieve joint optimization of resource slicing and scheduling, resulting in the following technical effects:

[0038] Improve resource utilization: Due to global resource reservation on a large time scale and fine-grained resource orchestration on a small time scale, the system can utilize available resources more efficiently and reduce resource waste.

[0039] Meeting heterogeneous QoS requirements: Through a flexible resource allocation mechanism, this invention can dynamically adjust for different service scenarios, ensuring that the QoS requirements of each slice are guaranteed and improving the service quality of the network.

[0040] Enhanced system scalability: The dual-time-scale optimization method enables the architecture to cope with changes in resource requirements under different business models and maintain stable performance even when network load and business requirements change dynamically.

[0041] Reduced signaling overhead: By reducing frequent decision adjustments at the resource reservation layer, this invention significantly reduces the system's signaling overhead and improves the overall efficiency of network operation.

[0042] This innovative network slicing method addresses the inefficiency caused by the separation of resource slicing and scheduling in existing technologies, and enhances the system's adaptability to the demands of complex industrial networks.

[0043] By introducing a multi-domain resource collaborative management and control scheme, this invention brings the following technical effects to the implementation of the dual-timescale network slicing method:

[0044] Improved optimization efficiency: By breaking down the problem into simpler subproblems, the computational complexity of resource orchestration and reservation decisions is reduced from exponential to linear, greatly improving the algorithm's execution efficiency.

[0045] Improved convergence: The convergence speed of resource reservation decisions is optimized by using the iterative block coordinate gradient descent algorithm, which improves the stability and reliability of the system.

[0046] Enhance resource utilization: Through precise resource allocation strategies, ensure that network resources are used more efficiently to meet heterogeneous QoS requirements.

[0047] High adaptability: This solution can dynamically adjust resource allocation strategies to adapt to constantly changing network conditions and business needs, thereby improving the flexibility and responsiveness of industrial network systems.

[0048] In summary, this invention provides an innovative and effective solution to the complex problem of dual-timescale network slicing. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of a hierarchical control network architecture based on industrial software-defined networking (SDN) provided in an embodiment of the present invention;

[0050] Figure 2 This is a schematic diagram of multi-domain resource collaborative management under dual-timescale network slicing provided in an embodiment of the present invention;

[0051] Figure 3 This is a schematic diagram illustrating the working principle of the multi-domain resource collaborative management and control scheme provided in this embodiment of the invention;

[0052] Figure 4 This is a flowchart of a resource orchestration algorithm based on multi-knapsack matching provided in an embodiment of the present invention;

[0053] Figure 5 This is a flowchart of the resource reservation algorithm based on iterative block coordinate gradient descent provided in an embodiment of the present invention. Detailed Implementation

[0054] The following description, with reference to the accompanying drawings, illustrates several preferred embodiments of the present invention to make its technical content clearer and easier to understand. The present invention can be embodied in many different forms, and the scope of protection of the present invention is not limited to the embodiments mentioned herein.

[0055] Figure 1 This paper demonstrates a hierarchical control network architecture based on industrial software-defined networking (SDN) for resource management and scheduling in complex industrial scenarios. Taking the measurement of moving surface angles during aircraft final assembly as an example, this architecture aims to improve the overall system performance and meet the Quality of Service (QoS) requirements of different industrial applications through hierarchical control and resource collaboration. This SDN-based hierarchical control network architecture mainly includes three layers: a field perception layer, an edge computing layer, and an industrial cloud decision layer. Each layer has a clear functional division to achieve efficient management and scheduling of cross-domain resources.

[0056] Field Sensing Layer: The field sensing layer consists of multiple terminal clusters and terminal-side SDN controllers (TSCs). The main task of this layer is to perform real-time sensing and data acquisition of equipment status in the industrial field. Taking the measurement of the moving surface angle in aircraft assembly as an example, multiple sensors are deployed on the moving surface to collect key data such as angle changes and position deviations in real time. This data undergoes local preprocessing and preliminary analysis by the terminal-side SDN controller to ensure timely and reliable data transmission. Simultaneously, the TSCs are responsible for interacting with the upper-layer network to ensure immediate response to field tasks.

[0057] Edge Computing Layer: The edge computing layer consists of multiple edge-side SDN controllers (ESCs) and edge servers. Its main function is to further process and analyze data from the perception layer, providing real-time computing, storage, and network resource allocation capabilities. In aircraft final assembly scenarios, the edge computing layer can complete a large amount of data processing work in a local data center close to the production site, reducing the system's dependence on cloud resources and minimizing transmission latency. Furthermore, ESCs are responsible for optimizing network resource scheduling based on network topology and business needs under dynamic spatiotemporal changes, enabling real-time response and resource allocation for on-site tasks.

[0058] Industrial Cloud Decision Layer: The industrial cloud decision layer consists of an industrial cloud-side SDN controller (CSC) and cloud servers, responsible for the overall resource scheduling and optimization decisions of the entire system. At the beginning of each network slice window, the CSC determines the proportion of communication, computing, and storage resources that need to be reserved for each edge server based on the QoS requirements of different applications. In the aircraft final assembly scenario, the CSC can dynamically adjust resource allocation strategies by integrating feedback information from multiple edge servers to ensure the execution quality of critical tasks and optimal overall system performance. Simultaneously, the industrial cloud decision layer is also responsible for long-term analysis of historical data to formulate global strategies and optimization plans.

[0059] This hierarchical control network architecture based on industrial SDN allows the system to flexibly allocate communication, computing, and storage resources according to the needs of different application scenarios, achieving efficient resource utilization, reducing system response time, and improving resource utilization and task completion quality. Furthermore, the hierarchical control strategy of this architecture ensures the independence and collaboration of each layer, supports resource scheduling and optimization at different time scales, and helps to cope with dynamic changes in network conditions in industrial environments.

[0060] Figure 2 This diagram illustrates the collaborative management and control of multi-domain resources under a dual-timescale network slicing architecture. The left side of the diagram shows the backbone of the SDN-based hierarchical control network architecture, while the right side shows the dual-timescale network slicing framework.

[0061] Left side: Backbone diagram of the hierarchical control network architecture

[0062] This diagram illustrates the closed-loop interaction mechanism of data flow and decision flow in an SDN-based architecture. The upward data flow is responsible for feeding back real-time device status and network information from the perception layer and edge computing layer; the downward decision flow involves the SDN controllers on the industrial cloud and edge sides issuing resource reservation instructions and orchestration decisions to ensure the system operates as expected. Through this closed loop of data and decision flow, the system achieves global perception and intelligent scheduling of network status, ensuring efficient resource utilization.

[0063] Right side: Network slicing framework with dual time scales

[0064] The dual timescale consists of a large timescale and a small timescale, each corresponding to different levels of control and optimization:

[0065] ①Large-scale resource reservation

[0066] On a large timescale, the pre-allocation of resources is primarily managed jointly by the Industrial Cloud SDN Controller (CSC) and the Edge SDN Controllers (ESCs). The CSC predicts future resource demands based on long-term historical data and business requirements, reserving corresponding communication, computing, and storage resources for each network slice. At the beginning of each slice window, the CSC makes a global resource reservation decision, determining the resource allocation ratio for different slices to ensure that the system achieves global optimization while meeting QoS requirements.

[0067] ② Small timescale resource orchestration

[0068] At a small timescale, dynamic resource orchestration is handled by edge-side SDN controllers (ESCs) and terminal-side SDN controllers (TSCs). Based on actual service requirements and network conditions, the ESC performs resource scheduling in each time slot, allocating reserved network resources to each terminal. During this process, the ESC dynamically adjusts the relationships between terminals and performs fine-grained allocation of bandwidth, computing resources, and storage resources to ensure system real-time performance and efficient resource utilization.

[0069] At both large and small timescales, resource reservation and orchestration decisions are interdependent, forming a tightly coupled control mechanism. This dual-timescale network slicing design ensures both global resource reservation optimization and rapid response to dynamically changing business requirements, exhibiting strong adaptability and flexibility. This architecture effectively addresses complex spatiotemporal dynamics in industrial scenarios, enabling collaborative optimization of cross-domain resources and improving network performance and resource utilization in the industrial internet environment.

[0070] Figure 3This paper demonstrates the working principle of a multi-domain resource collaborative management and control scheme. This scheme decomposes the complex dual-timescale network slicing problem into resource reservation and resource orchestration sub-problems, achieving collaborative resource management and control across different domains and optimizing resource allocation efficiency. The core of the scheme comprises two key algorithms: a resource orchestration algorithm based on multi-knapsack matching and a resource reservation algorithm based on iterative block coordinate gradient descent, which work in tandem. The resource orchestration algorithm dynamically adjusts the association between edge-side SDN controllers (ESCs) and terminal-side SDN controllers (TSCs) and the allocation of network resource blocks through four steps: list construction, request sending, reverse selection, and cyclical requests, to meet real-time service demands. The resource reservation algorithm converges to the optimal resource reservation ratio through partial derivative calculation and iterative updates, ensuring that resource allocation in each slice within each slice window meets long-term needs. The results of resource orchestration are fed back to the resource reservation algorithm, providing reference data for the next global resource reservation, forming a closed-loop optimization mechanism. Through this decomposition and collaboration mechanism, this solution can effectively handle the complexity of resource scheduling in complex industrial internet scenarios, significantly improve the efficiency of resource scheduling, and support the efficient allocation and optimization of resources, meeting the needs of multi-timescale and multi-domain resource collaboration in industrial scenarios.

[0071] Figure 4 The flowchart of the resource orchestration algorithm based on multi-knapsack matching includes the following steps:

[0072] Step 1: Initialization phase, obtain network status, terminal resource requirements and edge server resource status;

[0073] Step 2: Construct the bipartite graph G = (Γ) t ,E t ,Ξ t ), where Γ t It is a collection of sampling terminals, E t It is a collection of ESCs, Ξ t It is a connection Γ t and E t The set of edges of the middle vertex;

[0074] Step 3: Resource requirement assessment. Based on the expected service latency and relevant network resource constraints, calculate the network resources required to meet the requirements under a given associated decision <γ,ε>.

[0075]

[0076] Step 4: Sorting terminal preferences, generating a preference list for each TSC, and sorting the resources required under different associations. Sort ESCs in ascending order;

[0077] Step 5: Sending a request. Each TSC sends an association request to its favorite ESC and removes that ESC from its preference list.

[0078] Step 6: Request Inspection. Each ESC inspects all received requests, including new requests and those accepted in previous iterations, and performs a resource availability assessment, including the amount of resources required for each TSC. and priority weight p γ ;

[0079] Step 7: Matching Decision. Each ESC determines the list of TCSs to accept the request based on resource constraints and request weights, so as not to violate the overall weight constraint. Maximize total value under the following circumstances;

[0080] Step 8: Loop through the process. If a TCS request is rejected but ECSs still have available resources, return to step 5 to continue processing; otherwise, end the matching process.

[0081] Step 9: Output the results, output the resource orchestration results, and update the network status information.

[0082] Figure 5 This is a flowchart of a resource reservation algorithm based on iterative block coordinate gradient descent, including the following steps:

[0083] Step 1: Initialization phase, randomly generate initial resource reservation values ​​that meet the requirements for each slice. And set the learning factor α and the number of iterations n = 0;

[0084] Step 2: Calculate the gradient. Based on the current resource reservation decision and network state, calculate the gradient of the objective function.

[0085] Step 3: Update the reserved value using the formula. Update the reserved variables to ensure that the updated reserved values ​​are within the valid range;

[0086] Step 4: Calculate the difference in objective function values ​​for the current iteration.

[0087] Step 5: Convergence detection, determine whether the difference in objective function values ​​is less than the set convergence threshold or whether the current iteration number has reached the maximum iteration number;

[0088] Step 6: Iteration termination check. If the stopping condition is met, the algorithm ends and the final resource reservation scheme is output; otherwise, return to step 2 and proceed to the next iteration.

[0089] Step 7: Output the updated resource reservation values.

[0090] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0091] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A multi-domain resource collaborative management method under a double-time-scale network slicing mechanism based on an industrial SDN hierarchical control network architecture, characterized in that: The network architecture comprises a terminal side SDN controller, an edge side SDN controller and an industrial cloud side SDN controller; The terminal side SDN controller is deployed at an access end, a switch and a gateway, has a small coverage range and is responsible for resource scheduling of terminal users; The edge side SDN controller is responsible for coordinating multiple terminal side SDN controllers and realizing higher-level network operation; The industrial cloud side SDN controller is the highest-level controller, formulates global rules and realizes cross-domain coordination for the entire network by acquiring network state information provided by the bottom controllers; The multi-domain resource collaborative management method under the dual-time-scale network slicing mechanism comprises dividing the allocation of resources into different time domains for processing, including a resource reservation method and a resource arrangement method; The multi-domain resource collaborative management method under the dual-time-scale network slicing mechanism comprises the following steps: Step 1, according to the current network state and resource demand, the dual-time-scale network slicing problem is decomposed into a resource reservation sub-problem and a resource arrangement sub-problem, and the related constraint conditions are also decomposed; Step 2, at the resource arrangement layer, due to the isolation of slicing, a resource arrangement algorithm based on multiple knapsack matching is adopted; Step 3, at the resource reservation layer, the resource arrangement strategy is substituted into the resource reservation sub-problem as an incentive, which is converted into a closed expression related only to the reservation ratio, and then an iterative block coordinate gradient descent algorithm is adopted for solving; Step 2 specifically comprises the following steps: Step 2.1, constructing a bipartite graph wherein is a set of sampling terminals, is a set of edge-side SDN controllers, is a set of connections and between vertices in ; given a desired service delay and related network resource constraints, a given association decision the multi-domain network resources required to meet the requirements in case ； wherein, The required communication, computing and storage resources are further, a terminal-side SDN controller, a communication resource, a computing resource, a storage resource, t represents a time slot; Step 2.2, for each terminal-side SDN controller, sort the edge-side SDN controllers in ascending order of the required resources under different associations, and construct a preference list, i.e., the terminal-side SDN controller prefers the edge-side SDN controller with higher resource utilization rate; Step 2.3, the terminal-side SDN controller sends a request to the edge-side SDN controller in the preference list, and the edge-side SDN controller returns a response to the terminal-side SDN controller. Step 2.3, each terminal side SDN controller sends a request to the edge side SDN controller that it currently likes most, and then deletes this edge side SDN controller from the preference list; Step 2.4: Each edge-side SDN controller examines all received requests, including new requests and requests accepted in previous iterations; given a set of proposals, each request has a required amount of resources. and a priority weight Each edge-side SDN controller needs to determine which requests to accept in order to avoid violating the total weight constraint. Maximize total value under the following circumstances; Step 2.5, all unqualified terminal side SDN controllers go to Step 2.3; when all terminal side SDN controllers are matched to edge side SDN controllers or the available resource allocation of all edge side SDN controllers is completed, the matching process ends.

2. The method of claim 1, wherein the method is performed in a dual-time-scale network slicing mechanism. The resource reservation method is specifically as follows: at the start of each slice window , the industrial cloud side SDN controller makes a resource reservation decision according to network state feedback information from the edge side SDN controller and global service demand , wherein represents the edge side SDN controller, represents the index of the network slice.

3. The method of claim 2, wherein the method further comprises: The resource reservation method further comprises the following steps: during the entire network slicing window, the reservation decision remains unchanged; after each network slicing window ends, the industrial cloud side SDN controller evaluates the system performance through a feedback evaluation system and adjusts the resource allocation decision of the next network slicing window based on changes in business demand.

4. The method of claim 1, wherein the method is implemented in a dual-time-scale network slicing mechanism. The resource arrangement method specifically comprises the following steps: within each network slicing window, the edge side SDN controller performs real-time scheduling of resources according to network topology, link state and spatiotemporal changes in business demand.

5. The method of claim 4, wherein the method further comprises: The resource arrangement method specifically comprises the following steps: the real-time scheduling of resources specifically comprises allocating network resources to terminal side SDN controllers and flexibly adjusting resources between different time slots.

6. The method of claim 4, wherein the method further comprises: The resource arrangement method mainly focuses on the association between edge side SDN controllers and terminal side SDN controllers and the efficiency of resource allocation.

7. The method of claim 1, wherein the method further comprises: Step 3 specifically comprises the following steps: Step 3.1, in the initialization phase, randomly generate initial resource reservation values that meet the requirements for each slice and set the learning factor and the number of iterations ; Step 3.

2. Calculate the gradient of the objective function according to the current resource reservation decision and network state and update the reservation variable , ensuring that the updated reservation value is within the valid range. Step 3.3, calculate the target function value difference value of the current iteration If it is less than the set convergence threshold value or the current iteration number reaches the maximum iteration number, stop the iteration, otherwise, return to step 3.2 for the next iteration.

Images