Method for fine-grained slice isolation in air-sea cross-domain network

Abstract

The application discloses a fine-grained slice isolation method in an air-sea cross-domain network, and belongs to the technical field of network communication. The method comprises the following steps: firstly, collecting resource data to construct a cross-domain heterogeneous resource graph; secondly, screening and matching nodes by the K shortest path to select an optimal slice deployment scheme; thirdly, realizing slice hard isolation by means of resource data isolation technology, network resource isolation technology and wireless link resource isolation technology; and finally, monitoring the running state and triggering dynamic adjustment of the scheme when an abnormality occurs. The system comprises three modules to ensure seamless switching of the slice. By means of the deterministic optimization algorithm and the multi-level isolation mechanism, the application realizes accurate and efficient allocation and strict isolation of the heterogeneous resources, and significantly improves the QoS guarantee capability and system stability of the business in the high-dynamic cross-domain network.

Description

Technical Field

[0001] This invention belongs to the field of network communication technology, and specifically relates to a fine-grained slicing isolation method in air-sea cross-domain networks. Background Technology

[0002] This invention relates to resource management technology in air-sea cross-domain networks. With the development of integrated air, sea, and space networks, there is a need to provide differentiated and highly reliable network services for diverse missions such as reconnaissance, communication, and remote sensing in a complex heterogeneous environment composed of satellites, UAVs, ships, and ground stations. Network slicing technology, as a key enabling technology, creates multiple logically isolated virtual networks on shared physical infrastructure to meet the Quality of Service (QoS) requirements of different services. However, existing technologies have significant shortcomings when applied to the specific scenario of air-sea cross-domain networks.

[0003] Currently, the main technologies for achieving network slicing resource isolation include:

[0004] (1) Network slicing technology based on 5G core network architecture. This technology is mainly based on Network Functions Virtualization (NFV) and Software-Defined Networking (SDN). On a relatively fixed terrestrial network infrastructure, it partitions logically independent network resources for different services (such as eMBB, uRLLC, and mMTC). However, this technology has the following problems: First, its resource model and scheduling strategy are mainly designed for homogeneous terrestrial networks, making it difficult to uniformly and accurately model and perceive heterogeneous air and sea resources such as long-latency links of satellites, energy limitations of UAVs, and the movement trajectories of ships. Second, its orchestration mechanism assumes that the network topology is relatively stable, which cannot adapt to the drastic topology changes caused by high-speed node movement and frequent link disconnections in air and sea networks. This leads to frequent failures of resource allocation strategies, and the slicing isolation effect and QoS cannot be guaranteed.

[0005] (2) Traditional cross-domain resource scheduling and coordination technologies. These technologies typically rely on inter-domain gateway protocols (such as BGP) and pre-configured static service level agreements (SLAs) for routing and traffic engineering between different autonomous systems. However, these technologies have the following problems: First, the coordination mechanism has a slow response time, usually on the order of minutes or even hours, which cannot meet the real-time creation and second-level adjustment requirements of slices in highly dynamic scenarios. Second, each domain manages its own resources independently, lacking an end-to-end, unified resource view and control plane, making it extremely difficult to perform fine-grained resource reservation and isolation for cross-domain slices. When resources are scarce within a domain, global optimization scheduling is impossible, easily leading to business congestion and disrupting resource isolation between slices.

[0006] Therefore, how to implement a fine-grained slice resource isolation method in highly dynamic and heterogeneous air-sea cross-domain networks that can quickly respond to topology changes, accurately allocate heterogeneous resources, and provide end-to-end hard isolation guarantees is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention proposes a fine-grained slicing isolation method for cross-domain air-sea networks. This method involves collecting multi-dimensional resource data and constructing a cross-domain heterogeneous resource graph model. Then, a slice mapping mechanism uses a two-stage algorithm (KSP candidate path selection, node resource matching, and a global optimal decision algorithm to calculate a comprehensive score) to determine the optimal deployment scheme. Finally, a resource isolation execution framework translates the optimal deployment scheme into configuration instructions, achieving end-to-end hard isolation through three-layer isolation technology, and dynamically replanning and migrating upon event triggering. This provides deterministic QoS guarantees for network slices in complex and dynamic environments, effectively ensuring service continuity.

[0008] The technical solution adopted in this invention:

[0009] A fine-grained slice isolation method for cross-domain air-sea networks is proposed. This method constructs a unified cross-domain heterogeneous resource graph model, employs a two-stage combinatorial optimization algorithm to implement an intelligent slice mapping mechanism, and leverages a cross-layer collaborative resource isolation execution framework to achieve end-to-end resource isolation. This provides deterministic QoS guarantees for network slices in complex and dynamic environments. The method includes the following steps:

[0010] Step S1. Collect multi-dimensional resource data of cross-domain network nodes and links in the air and sea, and construct a cross-domain heterogeneous resource graph model based on the dynamic weighted graph model;

[0011] Step S2. Based on the combinatorial optimization strategy, design a slice mapping mechanism. After receiving the network QoS requirements of the slice, candidate paths are selected through K-shortest path filtering. The path cost and node resource matching degree are calculated by the node resource matching and global optimal decision algorithm to select the global optimal deployment scheme.

[0012] Step S3. Construct a multi-level, end-to-end resource isolation execution framework, transform the optimal deployment scheme into configuration instructions, and achieve end-to-end hard isolation between slices with the help of three-layer isolation technology;

[0013] Step S4. Monitor events; the optimal deployment plan will be dynamically adjusted after an abnormal event is triggered.

[0014] Preferably, the construction of the cross-domain heterogeneous resource graph model in step S1 is as follows:

[0015] Multi-dimensional resource data on nodes and links in air-sea cross-domain networks are collected. Based on a dynamic weighted graph model, a cross-domain heterogeneous resource graph model is constructed to uniformly represent cross-domain heterogeneous physical resources. The cross-domain heterogeneous resource graph model is represented as G=(V,E), where... For a set of nodes, As an edge set, this cross-domain heterogeneous resource graph model defines the attributes of the node set and edge set in multiple dimensions, as follows:

[0016] Node set : Represents a physical entity, each node Associate a multidimensional attribute vector ,in Represents the total computing power of a node. Indicates available computing power. Represents total memory. Indicates available memory. Indicates current energy consumption;

[0017] Edge set : Represents a physical or logical link; each edge Associate a multidimensional attribute vector ,in Indicates the total bandwidth of the link. Indicates available bandwidth. Indicates the propagation and processing delay. Indicates latency jitter. This indicates the reliability of the link.

[0018] Preferably, in step S2, the process of selecting the globally optimal deployment scheme is as follows:

[0019] The network QoS requirements containing multi-dimensional resource data slices are received, and a two-stage slice mapping mechanism is used to map formal slice service requests to physical resources, as follows:

[0020] Phase 1: Selection of K-shortest paths based on multi-objective constraints

[0021] The first stage process includes preliminary filtering and calculation of compound path costs, as detailed below:

[0022] 1) Preliminary filtering: Remove links that do not meet the QoS constraints based on the network QoS requirements of the slice;

[0023] 2) Calculate the cost of the composite path: Using the K-shortest path selection algorithm, the cost Cost(P) of each candidate path P is calculated using the composite path cost formula, and the K candidate paths with the lowest costs are selected; the composite path cost formula is:

[0024]

[0025] in, It is a path One of the links in; , These are links latency, available bandwidth, and reliability; This is the bandwidth of the slice request; It is a normalized weighting factor adjusted according to the business type;

[0026] Phase Two: Node Resource Matching Based on Optimal Adaptation and Global Optimal Decision

[0027] Based on the candidate paths selected by the K-shortest path algorithm, a node resource matching and global optimal decision-making algorithm is adopted; this algorithm is used for each candidate path. A comprehensive score is calculated by combining path cost and node resource matching degree. To simultaneously evaluate the matching degree between network cost and node resources, and select the globally optimal deployment scheme that balances network performance and computing resource utilization, the calculation formula is as follows:

[0028]

[0029] in, This is the composite path cost calculated in the first stage; and This refers to the computational and memory resource requirements of slicing; and Nodes on the path Available resources; This represents the resource matching degree of the node with the most scarce resources on the path. The closer the value is to 1, the better the resource matching and the less redundancy. and It is the weighting coefficient.

[0030] Preferably, in step S3, the multi-level, end-to-end resource isolation execution framework includes computing resource isolation technology, network resource isolation technology, and wireless link resource isolation technology. It transforms the globally optimal deployment scheme selected in step S2 into configuration instructions executable on the underlying heterogeneous devices and issues them for execution, achieving end-to-end isolation from computing and wired networks to wireless links, and realizing hard isolation between slices, as detailed below:

[0031] 1) Resource data isolation technology: By calling the container engine API through the orchestrator and utilizing the namespace and control group mechanism of the Linux kernel, an independent process and network protocol stack view is created for each slice, and its CPU time slice and memory usage are limited;

[0032] 2) Network resource isolation technology: Based on a software-defined network architecture, flow table rules are issued to switching nodes that support the OpenFlow protocol through a central controller; the flow table rules are based on slice ID and five-tuple information to direct traffic from different slices to different queues and virtual channels, and configure bandwidth guarantee and priority policies;

[0033] 3) Wireless link resource isolation technology: For satellite and UAV wireless communication scenarios, by linking with the underlying MAC protocol controller, orthogonal frequency division multiple access and time division multiple access technologies are used to allocate mutually orthogonal time and frequency resource blocks to different slices, thereby eliminating signal interference at the physical layer.

[0034] Preferably, the monitoring events specifically include:

[0035] After the resource isolation execution framework is completed, the running status of the slice and the network resource status are monitored in real time. When abnormal events such as link interruption, performance degradation, node failure, and slice requirement change occur, the migration is dynamically replanned and automatically re-executed in step S2. Based on the updated resource map, a new optimal deployment scheme is found. Subsequently, step S3 is re-executed to update the configuration of relevant devices, so as to achieve seamless switching and continuity assurance of slices.

[0036] Compared with existing technologies, this invention proposes a fine-grained slicing isolation method for cross-domain air-sea networks. The advantages of this method are:

[0037] (1) This invention solves the problems of mismatched heterogeneous resource models and inaccurate perception. Addressing the difficulty of unified modeling of heterogeneous air and sea resources using 5G slicing technology in the background, this invention constructs a unified cross-domain heterogeneous resource map (CD-HRG) to accurately quantify and dynamically update heterogeneous attributes such as long-latency satellite links and energy limitations of UAVs. This enables the system to comprehensively and in real-time perceive the true state of the entire cross-domain network, providing an accurate data foundation for subsequent resource allocation and fundamentally solving the problem of unreasonable resource allocation caused by model mismatch.

[0038] (2) It overcomes the problems of slow response and low efficiency of resource scheduling in highly dynamic environments; In response to the problems of slow response and inability to adapt to drastic changes in network topology of traditional cross-domain collaborative technology in the background technology, the slice mapping mechanism based on combinatorial optimization strategy is adopted, which can complete the calculation and decision of complex slice requests within seconds; This mechanism can quickly respond to dynamic events such as node movement and link interruption, and re-plan the optimal resource deployment scheme, ensuring the continuity of services and the stability of QoS, which is significantly better than the response speed of traditional technology in minutes or even hours.

[0039] (3) It achieves fine-grained end-to-end resource isolation and improves QoS guarantee capability. In view of the problem that the resource isolation granularity is coarse and the cross-domain coordination is weak in the background technology, which leads to the inability to guarantee QoS, a multi-level, end-to-end resource isolation execution framework is used to achieve all-round hard isolation from computing, network to wireless link. Through the coordination of technologies such as SDN, containerization and physical layer resource partitioning, it ensures that the resources of a slice (such as CPU, bandwidth and time slot) will not be occupied by other slices. Even when resources are scarce in a certain domain, scheduling can be carried out through the global view, thereby providing predictable and highly reliable end-to-end QoS guarantee and solving the problem of isolation failure caused by the lack of global coordination in traditional technology.

[0040] (4) Enhanced the system’s global optimization capabilities and resource utilization; Through a unified resource view and centralized optimization decision-making, the barriers of independent management and information sharing among autonomous domains in traditional technologies have been broken down; This enables the system to allocate resources from a global perspective, avoid local congestion and resource waste, significantly improve the resource utilization efficiency of the entire air and sea cross-domain network, and make it possible to carry more and higher quality services on limited resources. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the overall framework of the fine-grained slicing isolation method in the air-sea cross-domain network of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of this application will be further described clearly and completely below with reference to the accompanying drawings. It should be noted that the described embodiments are only some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0043] To make the inventive objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be further described in detail below with reference to the accompanying drawings: In order to better understand the above-mentioned objectives, features, and advantages of this invention, the advantages of this invention will be further illustrated below by comparing the embodiments with the accompanying drawings and specific implementation methods.

[0044] This invention proposes a fine-grained slicing isolation method in air-sea cross-domain networks, such as... Figure 1 As shown, this method constructs a unified cross-domain heterogeneous resource graph model for heterogeneous resources, employs a two-stage combinatorial optimization algorithm to implement an intelligent slice mapping mechanism, and leverages a cross-layer collaborative resource isolation execution framework to achieve end-to-end resource isolation, thereby providing deterministic QoS guarantees for network slices in complex and dynamic environments. The steps of this method are described in detail below:

[0045] Step S1. Collect multi-dimensional resource data of cross-domain network nodes and links in the air and sea, and construct a cross-domain heterogeneous resource graph model based on the dynamic weighted graph model;

[0046] Step S2. Based on the combinatorial optimization strategy, design a slice mapping mechanism. After receiving the network QoS requirements of the slice, candidate paths are selected through K-shortest path filtering. The path cost and node resource matching degree are calculated by the node resource matching and global optimal decision algorithm to select the global optimal deployment scheme.

[0047] Step S3. Construct a multi-level, end-to-end resource isolation execution framework, transform the optimal deployment scheme into configuration instructions, and achieve end-to-end hard isolation between slices with the help of three-layer isolation technology;

[0048] Step S4. Monitor events; the optimal deployment plan will be dynamically adjusted after an abnormal event is triggered.

[0049] Specifically, in step S1, the cross-domain heterogeneous resource graph (CD-HRG) model is constructed as follows:

[0050] Collect multi-dimensional resource data on nodes and links in cross-domain air and sea networks;

[0051] The multi-dimensional resource data includes: ① Node information: Through agent programs deployed on ships, drones, satellite gateways, and ground stations, their resource status is reported periodically (e.g., once per second); for example, the drone agent will report its current available computing power, remaining memory, battery power, etc.; ② Link information: Through software-defined network controllers, network monitoring tools, and link prediction models, the link status is obtained in real time, such as the available bandwidth, propagation delay, and link reliability of the wireless link between ships and drones, and the space link between drones and satellites.

[0052] The collected information is uniformly modeled into a dynamic weighted graph model. Based on this dynamic weighted graph model, a cross-domain heterogeneous resource map model is constructed to uniformly represent cross-domain heterogeneous physical resources such as air, sea, and space. The cross-domain heterogeneous resource map model is represented as G=(V, E), where... For a set of nodes, As an edge set, this cross-domain heterogeneous resource graph model provides a multi-dimensional quantitative definition of the attributes of the node set and edge set, as detailed below:

[0053] Node set : Represents a physical entity, such as a drone ,satellite Ships Wait, each node Associate a multidimensional attribute vector ,in Represents the total computing power of a node. Indicates available computing power. Represents total memory. Indicates available memory. Indicates current energy consumption;

[0054] Edge set : Represents a physical or logical link; each edge Associate a multidimensional attribute vector ,in Indicates the total bandwidth of the link. Indicates available bandwidth. Indicates the propagation and processing delay. Indicates latency jitter. Indicates link reliability (such as the probability of successful transmission).

[0055] Specifically, in step S2, the process of selecting the globally optimal deployment scheme is as follows:

[0056] When receiving network QoS requirements containing multi-dimensional resource data slices, the network QoS requirements of the slices include end-to-end latency, bandwidth, reliability, and node computing and memory resource requirements. Based on the resource graph in the cross-domain heterogeneous resource graph model constructed in step S1, a two-stage slice mapping mechanism is used to efficiently and accurately map the formal slice service request (SLA) to physical resources. The specific process is as follows:

[0057] Phase 1: K-Shortest Path (KSP) Selection Algorithm Based on Multi-Objective Constraints

[0058] The first stage process includes preliminary filtering and calculation of compound path costs, as detailed below:

[0059] 1. Preliminary filtering: Based on the network QoS requirements of the slice, remove links in the resource graph that do not meet the basic requirements of the slice (such as available bandwidth being lower than the requested value, reliability being lower than the threshold);

[0060] 2. Calculate the cost of the composite path: An improved K-shortest path selection algorithm is adopted, the core of which lies in a composite path cost formula specifically designed for heterogeneous networks; the cost Cost(P) of each candidate path P is calculated using the composite path cost formula, and K candidate paths with the lowest costs are selected; the composite path cost formula is:

[0061]

[0062] in, It is a path One of the links in; These are links latency, available bandwidth, and reliability; This is the bandwidth of the slice request; It is a normalized weighting factor adjusted according to the business type;

[0063] This formula unifies the costs of latency, bandwidth availability (in reciprocal form, the smaller the available bandwidth, the higher the cost), and reliability (in logarithmic form, converted into additive cost), thereby selecting the K lowest-cost candidate paths that are optimal at the network layer.

[0064] Phase Two: Node Resource Matching and Global Optimal Decision Algorithm Based on Best-Fit

[0065] Based on the candidate paths selected by the K-shortest path algorithm, a node resource matching and global optimal decision-making algorithm is adopted; this algorithm is used for each candidate path. A comprehensive score is calculated by combining path cost and node resource matching degree. To simultaneously evaluate the matching degree between network cost and node resources, and select the globally optimal deployment scheme that balances network performance and computing resource utilization, the calculation formula is as follows:

[0066]

[0067] in, This is the composite path cost calculated in the first stage; and This refers to the computational and memory resource requirements of slicing; and Nodes on the path Available resources; The resource matching degree represents the node with the most scarce resources on the path (bottleneck node). The closer the value is to 1, the more matched the resources are and the less redundant they are. It is used to evaluate whether the computing and memory resources of the node (drone in this example) that carries the virtual network function on the path are sufficient and not excessively redundant. and These are weighting coefficients; the algorithm maximizes the overall score. Select the globally optimal deployment scheme that balances network performance and computing resource utilization;

[0068] Optimal solution decision: The system selects the "path-node" combination with the highest comprehensive score as the final deployment solution; for example, the final decision may be to select the data flow path "ship -> UAV A -> satellite 1 -> ground station B", while specifying that the video compression function is deployed on UAV A.

[0069] Specifically, in step S3, the multi-level, end-to-end resource isolation execution framework includes computing resource isolation technology, network resource isolation technology, and wireless link resource isolation technology. It transforms the globally optimal deployment scheme selected in step S2 into configuration instructions executable on the underlying heterogeneous devices and issues them for execution, achieving end-to-end isolation from computing and wired networks to wireless links, and realizing hard isolation between slices, as detailed below:

[0070] ① Computational resource isolation technology: By calling the container engine API (such as Docker) through the orchestrator, a new container is created to run the video compression function. Through cgroups technology, the CPU time slice and memory usage of the container are strictly limited to ensure that it does not over-consume the drone's computing resources. The Linux kernel's namespaces and cgroups mechanism is used to create an independent process and network protocol stack view for each slice, which is isolated from other applications and its CPU time slice and memory usage are precisely limited.

[0071] ② Network resource isolation technology: Based on the software-defined networking (SDN) architecture, a software-defined networking (SDN) controller issues precise flow table rules to switching nodes that support the OpenFlow protocol. These rules, based on information such as slice ID and 5-tuple, direct traffic from different slices to different queues and virtual channels, opening a virtual channel with a 20 Mbps bandwidth guarantee and high priority for the data flow of the slice, and configuring strict bandwidth guarantee and priority policies.

[0072] ③ Wireless link resource isolation technology: For the wireless link between the ship and UAV A, by linking with the underlying MAC protocol controller, orthogonal frequency division multiple access (OFDMA) and time division multiple access (TDMA) technologies are used to allocate mutually orthogonal time and frequency resource blocks to different slices, thereby eliminating signal interference at the physical layer.

[0073] Specifically, the monitoring events include:

[0074] After the resource isolation execution framework is completed, the running status of the slices and the network resource status are monitored in real time. When abnormal events such as link interruption, performance degradation, node failure, and slice requirement change occur, the migration is dynamically replanned and automatically re-executed in step S2. Based on the updated resource map, a new optimal deployment scheme is found. Subsequently, step S3 is re-executed to update the configuration of relevant devices, so as to achieve seamless switching and continuity assurance of slices.

[0075] The link is interrupted: if UAV A flies out of the communication range with satellite 1, the link will be updated accordingly in the map;

[0076] The performance degradation refers to the fact that if the link between the ship and the drone is disrupted, the available bandwidth drops below the required value.

[0077] Through the collaborative work of the above four steps, this embodiment of the invention constructs a complete closed-loop system from resource modeling and intelligent decision-making to isolated execution, which can provide strictly isolated network slicing services with deterministic quality of service assurance for critical businesses in highly dynamic and heterogeneous air and sea cross-domain networks.

[0078] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0079] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for fine-grained slice isolation in an air-sea cross-domain network, comprising: Specifically comprising: Step S1. Collecting multi-dimensional resource data of air-sea cross-domain network nodes and links, and constructing a cross-domain heterogeneous resource graph model based on a dynamic weighting graph model; Step S2. Based on a combination optimization strategy, designing a slice mapping mechanism, after receiving the network QoS requirements of the slice, screening candidate paths through K shortest path, calculating path cost and node resource matching degree comprehensive score through node resource matching and global optimal decision algorithm, and selecting a global optimal deployment scheme; The process of selecting a global optimal deployment scheme is as follows: Receiving network QoS requirements of a slice containing multi-dimensional resource data, and using a two-stage slice mapping mechanism to map formal slice service requests to physical resources, specifically as follows: First stage: K shortest path screening based on multi-objective constraints The process of the first stage includes preliminary filtering and calculating composite path cost, specifically as follows: 1) Preliminary filtering: for the network QoS requirements of the slice, remove links that do not meet the QoS constraints; 2) Calculate the composite path cost: use the K shortest path screening algorithm to calculate the cost Cost(P) of each candidate path P through the composite path cost formula, and select the K candidate paths with the lowest cost; the composite path cost formula is: ； wherein, is a path of one link in the path; , are the latency, available bandwidth and reliability of the link , respectively; is the bandwidth of the slice request; is a normalized weight factor adjusted according to the traffic type; Second stage: node resource matching based on best adaptation and global optimal decision On the basis of the candidate paths screened out by the K shortest path algorithm, a node resource matching and global optimal decision algorithm is adopted The algorithm calculates the comprehensive score by combining the path cost and the node resource matching degree to simultaneously evaluate the network cost and the node resource matching degree, and selects a globally optimal deployment scheme that takes into account the network performance and the computing resource utilization rate, and the calculation formula is: ； wherein, is the composite path cost calculated in the first stage; and is the resource requirement of the slice pair for computation and memory; and is the available resource of the node on the path; represents the resource matching degree of the node with the most stressed resource on the path, and the value closer to 1 indicates that the resource is more matched and the redundancy is smaller; and is the weight coefficient; Step S3. Constructing a multi-level, end-to-end resource isolation execution framework, converting the optimal deployment scheme into configuration instructions, and realizing end-to-end hard isolation between slices by means of three-layer isolation technology; Step S4. Monitoring events, and dynamically adjusting the optimal deployment scheme after an abnormal event is triggered.

2. The method of claim 1, wherein, In the step S1 of constructing a cross-domain heterogeneous resource graph model, specifically as follows: The multi-dimensional resource data of node and link information in the air-sea cross-domain network is collected, a cross-domain heterogeneous resource graph model for uniformly representing cross-domain heterogeneous physical resources is constructed based on a dynamic weighting graph model, and the cross-domain heterogeneous resource graph model is represented as G=(V,E), wherein V is a node set, E is an edge set, and the cross-domain heterogeneous resource graph model quantitatively defines the attributes of the node set and the edge set in multiple dimensions, as follows: set of nodes : represents a physical entity, each node is associated with a multidimensional attribute vector wherein represents the total computing power of the node, represents the available computing power, represents the total memory, represents the available memory, represents the current energy consumption; edge set : represents a physical or logical link; each edge is associated with a multidimensional attribute vector where denotes the total bandwidth of the link, denotes the available bandwidth, denotes the propagation and processing delay, denotes the delay jitter, denotes the link reliability.

3. The method of claim 1, wherein, In the step S3, the multi-level, end-to-end resource isolation execution framework includes computing resource isolation technology, network resource isolation technology and wireless link resource isolation technology, converts the global optimal deployment scheme selected in step S2 into configuration instructions executable for underlying heterogeneous devices, and executes them to realize end-to-end isolation from computing, wired network to wireless link, realize hard isolation between slices, specifically as follows: 1) Resource data isolation technology: call the container engine API through the orchestrator, and use the namespace and control group mechanism of the Linux kernel to create an independent process and network protocol stack view for each slice, and limit its CPU time slice and memory usage; 2) Network resource isolation technology: based on the software-defined network architecture, the central controller issues flow table rules to the switch nodes supporting the OpenFlow protocol; the flow table rules are based on slice ID and five-tuple information, and guide the traffic of different slices to different queues and virtual channels, and configure bandwidth guarantee and priority strategy; 3) Wireless link resource isolation technology: for satellite and unmanned aerial vehicle wireless communication scenarios, through linkage with the underlying MAC protocol controller, orthogonal frequency division multiple access and time division multiple access technology are used to allocate mutually orthogonal time-frequency resource blocks for different slices, to eliminate signal interference at the physical layer.

4. The method of claim 1, wherein, The monitoring events specifically include: After the resource isolation execution framework is completed, the running state of the slice and the network resource state are monitored in real time. When abnormal events such as link interruption, performance decline, node failure and slice demand change occur, dynamic re-planning migration is triggered, the slice mapping mechanism of step S2 is automatically re-executed, a new optimal deployment scheme is found based on the updated resource atlas, then step S3 is re-executed, the configuration of the related device is updated, and seamless switching and continuity of the slice are realized.

Images