Deployment Method of Link-State Aware Dynamic Service Function Chain in F6G Low-Earth Orbit Satellite Optical Network

Abstract

This invention proposes a link-state-aware dynamic service function chain deployment method in an F6G low-Earth orbit (LEO) satellite optical network, relating to the technical field of LEO satellite optical network service function chain deployment. The invention includes: obtaining TLE data of space debris based on the network topology and network resource information of the F6G LEO satellite optical network; preprocessing the TLE data of space debris and performing clustering using the KMeans clustering algorithm; predicting the risk level of risky debris, calculating the interruption duration of risky links and the comprehensive risk value of the links; constructing a weighted satellite optical network topology based on the logical distance of the links; obtaining k optimal pre-deployment paths using the K-shortest path algorithm; constructing an optimization problem for joint optimization of flow routing and VNF placement; and successfully deploying the service if the constraints are met and no high-risk links cause service interruption. This invention effectively improves the overall service acceptance rate and resource utilization of the system while reducing the average latency of service deployment.

Description

Technical Field

[0001] This invention relates to the technical field of F6G low-Earth orbit satellite optical network service function chain deployment, and more particularly to a method for dynamic service function chain deployment in a low-Earth orbit satellite optical network. Background Technology

[0002] With the rapid development of Internet of Things (IoT) technology and the ever-growing demand for global connectivity, a large number of IoT devices are being deployed in remote areas to provide a wide range of emerging services. However, the rapid increase in the number of users, the continuous improvement in access speeds, and the increasingly diversified service requests mean that traditional terrestrial networks, due to long-distance transmission and high latency issues, are struggling to fully meet the urgent needs of these emerging applications for low latency, high reliability, and seamless access. The sixth-generation fixed communication network (F6G), as an important direction for future network development, proposes a new integrated space-ground network architecture. By deeply integrating satellite networks with terrestrial fiber optic networks, F6G constructs a globally covering, three-dimensional communication infrastructure, overcoming the limitations of traditional terrestrial networks in terms of coverage area and construction costs, and providing users with truly ubiquitous high-speed broadband access services. As an important component of the F6G architecture, low-Earth orbit (LEO) satellite optical networks exhibit high robustness, strong resilience, and global coverage across time and space communication capabilities. As an effective supplement to terrestrial networks, they are widely used in military and civilian fields such as emergency communications, environmental monitoring, geological surveys, and military communications.

[0003] However, satellite optical networks exhibit characteristics such as large scale, high dynamism, multi-dimensional and heterogeneous resources, and limited onboard capacity, while also offering diverse service types. This undoubtedly presents unprecedented challenges to resource management. For low-Earth orbit satellite networks, traditional hardware-based network function deployment methods are particularly unsuitable, as they not only increase payload and maintenance costs but also struggle to adapt to the complex and ever-changing space environment. To address this, Network Function Virtualization (NFV) technology has emerged and been introduced into the satellite network field, aiming to meet the diverse needs of IoT users in a more flexible and cost-effective manner. NFV technology decouples network functions from physical hardware, abstracting the computing, storage, and bandwidth resources of the physical network into virtual resource pools for flexible, on-demand allocation of network resources. A series of Virtual Network Functions (VNFs) are arranged in a specific order to form Service Function Chains (SFCs), guiding user traffic sequentially through different instances. This provides personalized solutions for different services, ensuring the quality of service for differentiated services through flexible network topology, and improving network flexibility and adaptability.

[0004] Furthermore, satellite networks are highly dynamic and latency-tolerant networks, characterized by time-varying topologies, complex space environments, and intermittent connectivity. In particular, the high density of space debris in low Earth orbit poses a serious threat to inter-satellite laser link communication. These issues can lead not only to increased latency and resource waste but also to service interruptions, severely threatening service reliability and stability. Therefore, deploying SFCs in dynamic satellite networks requires joint consideration of routing reliability and VNF placement effectiveness. While some researchers have proposed joint optimization problems for flow routing planning and VNF placement, most focus on the connection and interruption states of links, often neglecting the impact of multi-dimensional link attributes on service quality. This makes it difficult to fully guarantee the reliability of service deployment in the actual network under the F6G integrated space-ground scenario.

[0005] Patent application number 202310417320.6 discloses a method for optimizing the deployment of service function chains in a space-ground integrated network. The method includes: constructing a problem with service quality requirements, latency, and multi-dimensional resources as constraints, and maximizing total service profit as the optimization objective; solving for the optimization objective by placing VNFs and mapping virtual link routes for service requests; and finding the optimal deployment scheme. However, this invention does not consider the reliability of service deployment in dynamic satellite networks, especially under the F6G space-ground integrated network architecture. Ensuring the stability and reliability of services in highly dynamic and complex satellite networks is a problem that urgently needs to be solved. Summary of the Invention

[0006] To address the technical challenges of low efficiency and reliability in deploying service function chains in complex space scenarios of existing satellite optical networks, particularly in situations where link disruptions caused by high-density space debris occur, this invention proposes a dynamic service function chain deployment method with link status awareness in the F6G low-Earth orbit satellite optical network. When the inter-satellite laser link communication status fluctuates, flexible virtual network function pre-migration can be performed, thereby improving the service reliability of the entire F6G space-based backbone network, increasing the service deployment acceptance rate and resource utilization, while reducing the average service latency.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows: a method for deploying dynamic service function chains with link status awareness in an F6G low-Earth orbit satellite optical network, comprising the following steps:

[0008] Step 1: Based on the network topology and network resource information of the F6G low-Earth orbit satellite optical network, obtain the TLE data of space debris and initialize the service function chain request information;

[0009] Step 2: Preprocess the TLE data of spatial fragments and use the KMeans clustering algorithm to cluster the preprocessed data;

[0010] Step 3: Construct a space debris risk perception model based on the Transformer model and predict the risk level of the debris, determine the spatial relationship between the debris trajectory and the inter-satellite laser link, and calculate the duration of the risk link interruption and the comprehensive risk value of the link based on the inter-satellite laser link interruption risk assessment model.

[0011] Step 4: Construct the link logical distance based on multi-attribute decision theory, and construct the weighted satellite optical network topology based on the link logical distance;

[0012] Step 5: Use the K-shortest path algorithm to process the weighted satellite optical network topology to obtain k optimal pre-deployment paths;

[0013] Step 6: Construct an optimization problem with the constraints of latency, computation, storage, and communication resources of service function chain requests, and the goal of minimizing the average service latency. Perform joint optimization of flow routing and VNF placement.

[0014] Step 7: If the service deployment process meets the constraints and there are no high-risk links causing service interruption, the service deployment is successful, the service deployment plan is obtained, and network resources are updated; otherwise, the service deployment fails.

[0015] Preferably, the network topology information includes a set of satellite nodes, a set of satellite-to-ground nodes, a set of inter-satellite laser links and satellite-to-ground links and their connection relationships, physical distances between links, and propagation delays; the network resource information includes the CPU computing resources and storage resources currently available for each physical node, as well as the available bandwidth resources for each physical link.

[0016] The network topology of a dynamic low-Earth orbit satellite optical network is represented using a time-spreading graph model. The physical network is abstracted as a set of weighted undirected graphs, and the mathematical model of the physical network is as follows: ,in, For a weighted undirected graph, Indicates a time range. This represents the t-th discrete time slice. Represents a set of physical nodes. Represents the set of physical links, for a node The set of attributes includes CPU computing resources and storage resources Inter-satellite laser link between node m and node n The set of attributes includes available bandwidth resources. , This represents a set of satellite nodes that have computing and storage capabilities and can support VNF ​​instances;

[0017] The service function chain request information includes: the number of service function chain requests, the VNF type and processing order corresponding to each service request, the computing and storage resources required by each VNF, the bandwidth requirements of the virtual links between VNFs, and the maximum end-to-end tolerable latency; N service function chain requests are generated using a Poisson distribution model, and the service lifetime is generated using an exponential distribution model; the service function chain request is described as a six-tuple. ,in, Indicates the composition of a service request. An ordered VNF set Represents a set of virtual links. and These represent the computational and storage resources required by the VNF, respectively. This indicates the bandwidth resources required for the virtual link. Indicates a service request Maximum tolerable end-to-end delay; Define binary decision variables Indicates whether a VNF node has been placed on an LEO node, defining a binary decision variable. Indicates virtual link Is it mapped to an inter-satellite laser link? superior.

[0018] The satellite / ground nodes, inter-satellite and satellite-to-ground links, and their available computing, storage, and bandwidth resources within each time slice are organized into a unified data structure, and links that are invalid or have no available bandwidth are eliminated. At the same time, the generated arrival time, lifetime, and VNF sequence and resource requirements of each SFC are combined into a standardized service request queue.

[0019] Preferably, the method for preprocessing the TLE data of space debris is as follows: cleaning the records with missing or abnormal TLE data of space debris and retaining samples with complete orbital elements; filtering out information highly related to the interruption of inter-satellite laser links, including the number of times the space debris orbits the Earth per day and the first time derivative of the average motion, and randomly assigning each space debris a debris size within 1m to construct an unlabeled debris feature dataset.

[0020] The method of using the KMeans clustering algorithm to cluster the debris dataset is as follows: the preprocessed daily number of orbits around the Earth, the first time derivative of the average motion, and the three-dimensional features of the debris size are used as input data for the KMeans clustering algorithm. From binary classification to multi-class classification, the number of clusters K=3 is determined by combining two indicators: the silhouette coefficient and the sum of squared center distances. The space debris is divided into three categories: zero risk, low risk, and high risk, and the corresponding data are labeled with 0, 1, and 2 respectively.

[0021] Preferably, the method for predicting the risk level of debris is as follows: The clustered space debris dataset is used as the input parameter of the space debris risk perception model to perform risk level classification prediction of space debris. Specifically, debris feature data already clustered using the KMeans clustering algorithm and assigned risk labels of 0, 1, and 2 is input into the space debris risk perception model. During the model training phase, the cross-entropy loss function is selected as the loss function, and the Adam optimizer is used to update the model parameters. The number of training epochs is set to 1000, and the learning rate is adjusted within the range of 0.001 to 0.1. The number of times the Earth orbits per day, the first-order time derivative of the average motion, and the debris size are input into the trained space debris risk perception model to obtain classification prediction results for three risk levels: 0, 1, and 2. Data with zero risk is filtered out.

[0022] For space debris with non-zero risk, the risk assessment model for inter-satellite laser link interruption caused by space debris is used to calculate the duration of link interruption, and the start and end times of the corresponding link interruption, as well as the number of the interrupted link, are obtained accordingly.

[0023] Preferably, the method for determining the spatial relationship between the debris trajectory and the inter-satellite laser link is as follows: a three-dimensional spatial geometric model is established based on the precise orbital parameters of the debris and the satellite; by calculating the parameters of the orbital plane intersection line and the minimum distance, it is determined whether the spatial relationship is perpendicular intersection, general intersection, or special tangency; according to the spatial relationship between the space debris trajectory and the inter-satellite laser link, the link interruption scenario is divided into three cases: perpendicular intersection, general intersection, and special tangency.

[0024] Calculate the link interruption duration in a vertically intersecting scenario ;

[0025] Calculate the link interruption duration in typical intersecting scenarios ;

[0026] Calculate the duration of link interruption in special tangential scenarios ;

[0027] Among them, the orbital central angle parameter ; For fragment size, Represents the orbital radius of the space debris. Indicates the width of the laser beam. Indicates the speed at which space debris travels. This represents the angle between the trajectory of space debris and the inter-satellite laser link. The arc length representing the trajectory of space debris; and ,

[0028] When the inter-satellite laser link is in a high-risk, low-risk, or zero-risk state, a link risk impact factor is set. The combined risk value of the inter-satellite laser link is represented by infinity, a constant value, and 1, respectively. in, This indicates the distance between space debris and the satellite at the launch site.

[0029] Preferably, the link logical distance The calculation formula is:

[0030]

[0031] Among them, w1, w2, and w3 are weighting factors; Indicates LISL link Available bandwidth resources in the current time slot; Indicates LISL link Maximum bandwidth resources; Represents the set of all LISL links; Indicates LISL link The instantaneous physical distance; Represents a set The maximum instantaneous physical distance of all LISL links in the network; Indicates LISL link The overall risk value; Indicates link The maximum overall risk value during the assessment period;

[0032] The method for constructing a weighted satellite optical network topology based on link logical distance is as follows: In the satellite network topology, the link logical distance W of each LISL link is used as the weight of the corresponding edge, forming a weighted undirected graph. The data on each edge represents the link logical distance value of the inter-satellite laser link within the current time slot. The entire satellite network cycle is discretized, and the network topology within a time slot is treated as a quasi-static network. A two-dimensional matrix of link weights is set for the t-th discrete time slot. for:

[0033] ;

[0034] in, , , , , , , , , , , These represent the weights of the corresponding edges between physical nodes 1, 2, ..., m within the t-th discrete time slice.

[0035] Preferably, within each time slot, the logical distance W of each LISL is recalculated based on the latest values ​​of the current available link bandwidth, physical distance, and overall link risk value, and the two-dimensional matrix is ​​updated. The corresponding elements ensure that the weighted satellite optical network topology always reflects the latest link status;

[0036] The method for obtaining k optimal pre-deployment paths is as follows: using the source satellite node and the destination satellite node as endpoints, the KSP algorithm is run on the weighted satellite optical network topology to iteratively find and return the path with the shortest cumulative link logical distance among the first k paths, thus obtaining k optimal pre-deployment paths.

[0037] Preferably, the optimization problem

[0038]

[0039]

[0040]

[0041]

[0042]

[0043]

[0044]

[0045]

[0046] in, and These represent service requests respectively. The Middle VNF nodes and the VNF nodes Place on a physical node Decision variables; Represents the set of all service requests The total number of service requests; This indicates that the k-th service request is constituted. An ordered VNF set; and These are the source VNF nodes. and target VNF ​​node The decision variable for the placement location; Indicates a service request Middle Virtual Link Is it mapped to a physical link? Decision variables; Indicates a service request The original end-to-end latency without VNF ​​migration; Indicates a service request Additional latency due to VNF migration; This indicates the number of service requests successfully deployed during the simulation period.

[0047] Preferably, the optimization problem is solved heuristically: given the physical network and link logical distance weights, the K-shortest path algorithm is used to obtain a set of candidate pre-deployment paths for each service request. On the candidate pre-deployment path set, the resource availability and latency of each candidate pre-deployment path are checked sequentially according to the constraints. Once a candidate pre-deployment path that satisfies all constraints is found, service pre-deployment is performed, determining the VNF placement location and virtual link mapping result for the service request, thus simultaneously providing a joint optimization solution for flow routing and VNF placement. Otherwise, the process continues to check whether other candidate pre-deployment paths are satisfied. If all pre-deployment paths in the candidate pre-deployment path set are not satisfied, deployment fails.

[0048] Service pre-deployment is for service requests on candidate pre-deployment paths. Allocate the corresponding physical node and physical link resources, and determine the corresponding decision variables. and The value of is selected and recorded as the deployment scheme for the current service request;

[0049] Determine if there are high-risk links in the pre-deployment path within the lifecycle of the service request. If so, immediately start the pre-migration mechanism to allow the data flow directly from the source node to the target migration node.

[0050] Preferably, the space debris risk perception model processes the three-dimensional features of space debris and the corresponding 0, 1, and 2 risk labels, and combines the link interruption duration and comprehensive risk value calculation results to form a risk status sequence for each LISL in each time slot; according to this risk status sequence, the lifetime interval of the service request is compared with the high-risk time interval of all inter-satellite laser links on the pre-deployed path. If a link is marked as high-risk during the service lifetime, it is determined that there is a high-risk link in the pre-deployed path, triggering the pre-migration mechanism;

[0051] The pre-migration mechanism allows data to flow from the source node to the VNF migration node, and then to the destination node;

[0052] The method for updating network resources is as follows: After the service function chain request is successfully deployed, based on the final VNF placement result and virtual link mapping result, the occupied CPU computing resources are adjusted. and storage resources The virtual link bandwidth requirement is deducted from the available resources of the corresponding satellite node and from the available bandwidth of the physical link it maps to. When the service life ends and the service exits the network, the node resources and link bandwidth occupied by the service function chain request are returned to the available resources of the corresponding physical node and physical link.

[0053] Compared with existing technologies, the beneficial effects of this invention are as follows: Addressing the development needs of F6G integrated space-ground networks, this invention focuses on the impact of frequent link status information fluctuations and high-density space debris on service quality. It designs a link status-aware dynamic service function chain deployment method: First, the space debris dataset is preprocessed to filter out information highly correlated with inter-satellite laser link interruptions. Then, the unlabeled space debris data is clustered using the KMeans clustering algorithm, adding three categories of labels: zero risk, low risk, and high risk. The Transformer risk-aware model is used to classify and predict the clustering results, determining the number, start, and end times of interrupted links. The logical distance between inter-satellite links is constructed by comprehensively considering link physical distance, interruption risk, and bandwidth resources. Furthermore, a weighted undirected time spread graph (TEG) with logical distance as link weight is designed to describe the topology of the dynamic satellite optical network. The K-shortest path algorithm is used to determine k optimal pre-deployment paths. It is then determined whether the pre-deployment paths meet resource and latency constraints. If a high-risk link exists in a pre-deployment path during the service lifetime, a VNF pre-migration mechanism is immediately initiated to effectively mitigate the adverse effects of the migration process on latency. This invention effectively improves the overall service acceptance rate and resource utilization of the F6G system, while reducing the average latency of service deployment. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a schematic diagram of the process in this invention.

[0056] Figure 2 This is a schematic diagram of a topology snapshot based on link logical distance in this invention, wherein (a) is a topology snapshot based on link open or closed, and (b) is a topology snapshot based on link logical distance.

[0057] Figure 3 This is a schematic diagram of the pre-migration mechanism in this invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] like Figure 1 As shown, a dynamic service function chain deployment method with link status awareness in an F6G low-Earth orbit satellite optical network addresses the impact of space debris on inter-satellite laser link communication. It utilizes the KMeans clustering algorithm and the Transformer model to classify and predict the risks of space debris, obtaining data information and risk levels of interrupted links. To mitigate migration delays caused by local rerouting to avoid high-risk links, a pre-migration mechanism is introduced, allowing VNF nodes to migrate directly from the source node to the target migration location in advance based on predicted link status information. This improves the service deployment acceptance rate and reduces the average service latency. The implementation steps of this invention are as follows:

[0060] Step 1: For the F6G low-Earth orbit satellite optical network, based on the current network topology and network resource information of the low-Earth orbit satellite optical network, obtain the two-line orbit data (TLE) of space debris and initialize the service function chain request information.

[0061] The network topology information includes satellite node sets, satellite-to-ground node sets, inter-satellite laser links and satellite-to-ground link sets and their connection relationships, link physical distances, and propagation delays, which can be calculated based on satellite constellation parameters and orbital dynamics models. Network resource information includes the currently available CPU computing and storage resources of each physical node, as well as the available bandwidth resources of each physical link, which are monitored and reported in real time by the network resource management module. Physical network information refers to the aforementioned actual satellite / ground nodes, physical links, and the multidimensional resources they carry; virtual network information refers to the VNFs, virtual links, and their resource requirements defined by the SFC request. The initialization of virtual network information involves: generating the service request arrival sequence using a Poisson distribution based on the given average service arrival rate and average lifetime; generating the lifetime of each request using an exponential distribution model; and randomly generating the computing, storage, and bandwidth requirements of each VNF and the maximum end-to-end tolerable delay within a preset range, thereby obtaining the virtual network description corresponding to each request.

[0062] Step 1.1: Initialize the network topology information of the low-Earth orbit (LEO) satellite optical network and obtain the Time-Legacy Graph (TLE) data of space debris. This invention uses a Time-Spread Graph (TEG) model to represent the dynamic network topology of the LEO satellite optical network. The physical network is abstracted as a set of weighted undirected graphs, effectively shielding the dynamic nature of the underlying network and facilitating the use of mature graph algorithms for routing planning and service deployment. The mathematical model of the physical network is as follows: .in, For a weighted undirected graph, Indicates a time range, where, Let p represent the t-th discrete time slice, and p represent the physical network corresponding to the weighted undirected graph. Represents a set of physical nodes. Represents the set of physical links, for a node The set of attributes includes CPU computing resources and storage resources The link between node m and node n The set of attributes includes available bandwidth resources. . This represents a set of satellite nodes with computing and storage capabilities, capable of supporting VNF instances. In subsequent service function chain optimization, a weighted undirected graph... As the physical network carrier, its nodes and link resources constitute the constraints for VNF deployment and are used to calculate link logical distances, construct time-spread graph weights, and update network status. After preprocessing and clustering, they are used as input to the Transformer model to predict link outages and assess risks, providing support for high-reliability routing planning.

[0063] Step 1.2: Initialize Service Function Chain Request Information. Service function chain request information includes: the number of service function chain requests, the VNF type corresponding to each request and its processing order, the computing and storage resources required by each VNF, the bandwidth requirements of virtual links between VNFs, and the maximum end-to-end tolerable latency. N service function chain requests are generated using a Poisson distribution model, and the service lifetime is generated using an exponential distribution model. This simulates the random fluctuations of service load in real satellite networks over time, facilitating the evaluation of the proposed method's performance under different service intensities. The above random generation process requires pre-defined values ​​for the average service arrival rate, average lifetime, and the range of CPU computing resources, storage resources, bandwidth resources, and latency requirements for different types of services. This invention describes a service function chain request as a six-tuple. ,in, Indicates the composition of a service request. An ordered VNF set Represents a set of virtual links. and These represent the computing and storage resources required by a Virtual Network Function (VNF), respectively. This indicates the bandwidth resources required for the virtual link. Service request The maximum tolerable end-to-end delay. Define binary decision variables. Indicates whether a VNF node has been placed on an LEO node, defining a binary decision variable. Indicates virtual link Is it mapped to an inter-satellite laser link? superior.

[0064] exist Figure 1 In the illustrated process, preprocessing physical network information and service request information involves organizing satellite / ground nodes, inter-satellite and satellite-to-ground links, and their available computing, storage, and bandwidth resources within each time slice into a unified data structure, eliminating failed links or links with no available bandwidth. Simultaneously, the arrival time and lifetime generated through Poisson / exponential distribution are combined with the VNF sequence and resource requirements of each SFC to form a standardized service request queue. Updating network resources involves deducting the computing, storage, and bandwidth occupied by a service function chain from the available resources of the corresponding physical nodes and links when the service is successfully deployed; when the service ends, the occupied resources are released and written back to the resource table, ensuring that subsequent services always make deployment decisions based on the latest network resource status.

[0065] Step 2: Preprocess the TLE data of space debris, retain data information that is highly related to the interruption of inter-satellite laser links, and use the KMeans clustering algorithm to cluster the debris dataset into three categories: zero risk, low risk, and high risk.

[0066] The physical network environment is highly complex and dynamic. High-speed moving LEO satellites and high-density space debris cause intermittent connectivity and unpredictable interruptions in inter-satellite laser links (LISLs). While the operational information of space debris is predictable, the unpredictable failures of LISLs caused by debris are not. Both the operational status of space debris and satellites are predictable; by predicting the risk level of space debris in advance and determining its spatial relationship with LISLs, interruption information can be obtained.

[0067] The Time-Loop (TLE) data of space debris was preprocessed. First, 3710 sets of TLE data were selected as the original dataset. Records with missing or anomalies were cleaned, retaining samples with complete orbital elements. Then, information highly correlated with inter-satellite laser link disruption was selected, including the number of times the space debris orbits the Earth daily and the first-order time derivative of its average motion. Each debris was randomly assigned a size value within 1 meter. These three features served as input to the subsequent link disruption risk assessment model, characterizing the orbital motion and size information of the space debris, thus constructing an unlabeled debris feature dataset. The KMeans clustering algorithm was used to cluster the unlabeled debris dataset. The preprocessed number of daily orbits, the first-order time derivative of the average motion, and the three-dimensional features of the debris size were used as input data for the KMeans clustering algorithm. From binary classification to multi-class classification, the number of clusters (K=3) was determined by combining the silhouette coefficient and the sum of squared center distances. The debris was divided into three categories: zero risk, low risk, and high risk, and labeled 0, 1, and 2 respectively. Using the KMeans clustering algorithm for clustering can automatically complete the initial risk classification of a large amount of unlabeled fragment data without manual annotation, mapping continuous track features to discrete risk level labels. On the one hand, this provides training samples for the subsequent Transformer risk perception model, reducing the cost of manual annotation. On the other hand, it facilitates the rapid reference and processing of fragments of different risk levels in subsequent link risk assessment.

[0068] Step 3: Construct a space debris risk perception model based on the Transformer model and predict the risk level of the debris, determine the spatial relationship between the debris trajectory and the inter-satellite laser link, and calculate the duration of the risk link interruption and the comprehensive risk value based on the inter-satellite laser link interruption risk assessment model.

[0069] A Transformer model is used to classify and predict the risk level of space debris. The clustered space debris dataset is used as input parameters for the space debris risk perception model. Specifically, debris feature data (number of orbits around the Earth per day, first-order time derivative of average motion, and debris size) clustered using the KMeans clustering algorithm and labeled with risk levels of 0, 1, and 2 are input into the Transformer model. During model training, the cross-entropy loss function is selected as the loss function, and the Adam optimizer is used to update the model parameters. The training epochs are set to 1000, and the learning rate is adjusted within the range of 0.001 to 0.1 to obtain better prediction results. The input features of the space debris risk perception model are the number of orbits around the Earth per day, the first-order time derivative of average motion, and debris size. The output is three risk levels: 0, 1, and 2, used for supervised classification and prediction of space debris risk levels.

[0070] There is a fundamental difference between the risk levels predicted by the KMeans clustering algorithm and the space debris risk perception model: the risk level obtained by clustering analysis is an unsupervised classification result based on the similarity of orbital features, while the risk level predicted by the Transformer model is a supervised classification result obtained by learning the features of labeled data, which has higher accuracy and generalization ability. The risk level prediction results are used to screen debris that needs further analysis. Only debris marked as low-risk or high-risk will enter the subsequent geometric relationship calculation and risk assessment process. After obtaining the classification prediction results, the zero-risk dataset is filtered out, which can significantly improve the computational efficiency and avoid unnecessary geometric relationship calculations and risk assessments for a large number of debris that does not pose a threat. This allows the system to concentrate its computing resources on dealing with space debris that truly poses a potential threat. The spatial positional relationship between the space debris trajectory and the inter-satellite laser link is determined. Based on the precise orbital parameters of the debris and the satellite, a three-dimensional spatial geometric model is established. By calculating parameters such as the orbital plane intersection line and the minimum distance, it is determined whether the spatial positional relationship belongs to one of three cases: perpendicular intersection, general intersection, or special tangency. This provides a geometric relationship basis for subsequent outage time calculation. The outage time of LISL is also calculated. Combined risk value with LISL The specific calculation formula is as follows:

[0071]

[0072]

[0073]

[0074]

[0075] Among them, formula (1) is the formula for calculating the link interruption duration in the vertical intersection scenario, formula (2) is the formula for calculating the link interruption duration in the general intersection scenario, and formulas (3) and (4) are the formulas for calculating the link interruption duration in the special tangential scenario. Based on the spatial positional relationship between the space debris trajectory and the inter-satellite laser link, the link interruption scenario is divided into three cases: vertical intersection, general intersection, and special tangential, and the corresponding formulas are used for calculation in each case. Specifically, , Represents the orbital radius of the space debris. Indicates the central angle parameter of the orbit. Indicates the arc length of the trajectory of space debris. It refers to the size of the fragments. This indicates the width of the laser beam (typically 10 mm). This indicates the speed at which space debris travels.

[0076] The risk of inter-satellite laser link outage is primarily determined by the degree of debris obstruction and the duration of the resulting link outage. A link risk impact factor is set when the inter-satellite laser link is in a high-risk, low-risk, or zero-risk state. These represent infinity, a constant value, and 1, respectively. The Link Comprehensive Risk Value (LCR) calculation formula is as follows:

[0077]

[0078] in, This represents the distance between space debris and the launching satellite, which is obtained by calculating the difference in three-dimensional spatial coordinates between the space debris and the launching satellite in the geocentric inertial coordinate system.

[0079] Step 4: Construct the link logical distance based on multi-attribute decision theory, and construct the weighted satellite optical network topology based on the link logical distance.

[0080] Taking into account the impact of link state information such as link outage risk, physical distance, and available bandwidth resources on link stability, a normalization method is used to transform multiple indicators to a unified metric scale to calculate the logical distance of the LISL link. Since all indicators have an equally important impact on link communication quality, the weighting factors w1, w2, and w3 are all set to 1 / 3. The formula for calculating the logical distance of a LISL link is as follows:

[0081]

[0082] in, Indicates LISL link Maximum bandwidth resources; Indicates LISL link Available bandwidth resources in the current time slot; Indicates the LISL link number; Represents the set of all LISL links; Indicates LISL link The instantaneous physical distance is calculated from the satellite's orbital position; Represents a set The maximum physical distance of all links in the network; Indicates LISL link The overall risk value is calculated through space debris risk assessment; Indicates link The maximum overall risk value during the assessment period.

[0083] The logical distance W of the link is a comprehensive quantization of the link's communication resources. physical distance and risk value The impact of link attributes on link communication can integrate multi-dimensional link state information into a single metric. This allows a single scalar to simultaneously reflect link bandwidth sufficiency, transmission distance, and interruption risk in subsequent weighted topology modeling and K-shortest path algorithm searches, thus enabling comprehensive consideration of link quality differences during route planning.

[0084] Based on this, a weighted satellite optical network topology based on link logical distance was constructed. Specifically, in the satellite network topology, the logical distance W of each LISL link is used as the weight of the corresponding edge, forming a weighted undirected graph, such as... Figure 2 As shown in (b), the data on each edge is the link logical distance value of the inter-satellite laser link in the current time slot. Figure 2 (a) is a traditional topology snapshot obtained solely based on the open or closed state of the links, without distinguishing between different link qualities at each edge; while Figure 2 (b) adds logical distance weights on top of this, which can intuitively reflect the comprehensive differences in bandwidth, distance, and risk among different links. The entire satellite network cycle is discretized, and the network topology within a time slot is treated as a quasi-static network. A two-dimensional matrix of link weights within slot t is set for the t-th discrete time slice. It means, and:

[0085]

[0086] in, , , , , , , , , , , These represent the weights of the corresponding edges between physical nodes 1, 2, ..., m in the t-th discrete time slice. When there is an inter-satellite laser link between physical node i and physical node j, the logical distance of the link in the current time slot t is taken; otherwise, it is denoted as infinity.

[0087] Unlike traditional topology snapshots based on link connection and interruption states, the topology designed in this invention comprehensively considers the impact of multiple link state information on link communication quality by introducing a logical distance W, thus constructing a link logical distance quantification of link communication quality. The logical distance W integrates multiple state indicators such as available bandwidth, physical distance, and comprehensive risk value, and these indicators are normalized and weighted to merge into a unified metric.

[0088] Figure 1The start and end times of inter-satellite laser link interruptions are determined using a space debris risk perception model. After obtaining the classification results of the space debris risk perception model, for non-zero risk debris, the link interruption duration is calculated using a space debris-induced ISLL interruption risk assessment model, and the start and end times of the corresponding link interruption, as well as the number of the interrupted link, are obtained accordingly. Updating the link logical distance means that, within each time slot, based on the latest values ​​of the current available bandwidth, physical distance, and comprehensive risk value, the logical distance W of each inter-satellite laser link is recalculated using equation (6), and the matrix is ​​updated. Corresponding element This ensures that the weighted topology always reflects the latest link status.

[0089] Step 5: Use the K-shortest path algorithm to process the weighted satellite optical network topology to obtain k optimal pre-deployment paths.

[0090] Based on link risk awareness and the impact of multi-dimensional link attribute changes on routing reliability, a link-state-aware reliable flow routing planning method based on KSP is designed. The K-shortest path algorithm (KSP) is used to calculate k optimal pre-deployment paths. The data processed by the KSP algorithm includes the weighted satellite optical network topology constructed in step 4, with link logical distance W as the weight. This topology integrates the real-time bandwidth, logical distance, and risk status of the links. By using the source and destination satellite nodes as endpoints, the KSP algorithm is run on this weighted topology to iteratively find and return the top k paths with the shortest cumulative link logical distance, thus obtaining the k optimal pre-deployment paths.

[0091] Step 6: Construct an optimization problem with the constraints of service request latency, computation, storage and communication resources, and the goal of minimizing the average service latency. Perform joint optimization of flow routing and VNF placement.

[0092] By performing flow routing planning and VNF placement optimization on each service request separately, the best deployment strategy is found to meet resource constraints, thereby ensuring acceptance rate and service quality.

[0093] Deploying SFC (Service Controller Center) in low-Earth orbit (LEO) satellite optical networks requires ensuring its effectiveness and reliability while meeting constraints. When the physical network topology changes, the migration of VNFs (Virtual Network Functions) must be completed in real-time and efficiently to guarantee service quality. Service request latency includes computation latency and communication latency, with communication latency encompassing satellite-to-ground and inter-satellite transmission and propagation latency. For pre-deployment paths with high-risk links, VNF migration latency needs to be considered.

[0094] The deployment method for a link-state-aware dynamic service function chain with link endpoints includes the following steps:

[0095] Step 6.1: Define the optimization objective and constraints as follows:

[0096]

[0097]

[0098]

[0099]

[0100]

[0101]

[0102]

[0103]

[0104]

[0105] in, and A binary decision variable representing a service request. The Middle VNF nodes and the VNF nodes Place on a physical node Above; k is the index of the service request; Represents the set of all service requests The total number of service requests; This indicates that the k-th service request is constituted. An ordered VNF set; and These are the source VNF nodes. and target VNF ​​node The placement location decision variable; A binary decision variable representing a service request. Middle Virtual Link Is it mapped to a physical link? superior; Indicates a service request The original end-to-end latency without VNF ​​migration is obtained by summing the communication latency and the processing latency; Indicates a service request The additional latency caused by VNF migration is calculated by the migration latency model; This indicates the number of service requests successfully deployed during the simulation period.

[0106] Equation (8) constrains all VNF nodes to be instantiated and placed on a single physical node. Equation (9) is the placement constraint of the source and target VNF ​​nodes. Equation (10) constrains the order of service data flow. Equations (11) to (13) are the constraints of computing resources, storage resources and link bandwidth resources during SFC deployment. Equation (14) is the latency constraint of service deployment. Equation (15) is the original latency during service deployment and the latency generated by VNF migration. The optimization objective is to minimize the average latency of service deployment.

[0107] The above optimization problem is solved heuristically using the inter-satellite laser link state-aware dynamic service function chain deployment algorithm proposed above: First, under the given physical network and link logical distance weights, the K shortest path algorithm is used to obtain the candidate pre-deployment path set for each service request. On this basis, the resource availability and latency of each candidate path are checked sequentially according to the constraints (8)-(14) to see if they meet the requirements. Once a path that meets all the constraints is found, the VNF placement location and virtual link mapping result of the service request are determined, thus giving the joint optimization solution of flow routing and VNF placement at the same time, without having to directly solve the integer programming model.

[0108] Step 6.2: Determine whether the candidate pre-deployment path meets the resource constraints and service request latency constraints. If the constraints are met, then pre-deploy the service, that is, provide the service request on the candidate pre-deployment path. Allocate the corresponding physical node and physical link resources, and determine the corresponding decision variables. and The system selects the value of [value] and records it as the current service deployment scheme; otherwise, it continues to check whether other pre-deployment paths are satisfied. If all pre-deployment paths in the candidate pre-deployment path set obtained by the K-shortest path algorithm in step 5 are not satisfied, the deployment fails. Service requests that fail to deploy will be recorded and abandoned, and their resource requirements will not be allocated. The system will continue to process the next service request.

[0109] Determine whether the node's computing and storage resource constraints and the service's latency constraints are met. For services that do not meet the constraints, check whether the next pre-deployment path meets the constraints. If a path that meets the constraints exists, the service is pre-deployed; otherwise, the service deployment fails.

[0110] Step 6.3: After service pre-deployment, it is necessary to determine whether there are high-risk links in the pre-deployment path within the life cycle of the service request based on the link risk perception model. If so, the pre-migration mechanism should be started immediately to allow the data flow to go directly from the source node to the target migration node, thereby reducing the impact of the migration process on the overall latency.

[0111] The link risk perception model is composed of a Transformer-based space debris risk perception model and a space debris-induced inter-satellite laser link (LISL) interruption risk assessment model. It is obtained by training on the clustered space debris feature data and their risk level labels from steps 2 and 3, and is used to output the risk level of each LISL within the simulation time, as well as information such as the start and end times of interruption. The space debris risk perception model processes the three-dimensional features of space debris and their corresponding 0, 1, and 2 risk labels. Combined with the link interruption duration and the comprehensive risk value calculation results, a risk state sequence for each LISL within each time slot is formed. Based on this risk state sequence, the service request's lifetime interval is compared with the high-risk time intervals of all inter-satellite laser links on its pre-deployed path. If a link is marked as high-risk within its service lifetime, it is determined that a high-risk link exists on the pre-deployed path, requiring the triggering of a pre-migration mechanism.

[0112] The specific execution process of the pre-migration mechanism is as follows: Figure 3 As shown, Slot1, Slot2, and Slot3 refer to the discretized time slots, respectively. , , When the satellite network topology changes dynamically due to the influence of space debris, inter-satellite laser links become more vulnerable. exist It can become a high-risk link at any time, and The situation may then revert to a low-risk or risk-free state. This uncertainty necessitates that SFC data streams must be routed through satellite nodes. Storing and waiting for the next connection inevitably introduces additional latency and may cause service interruptions. To address this issue, the local rerouting strategy changes the deployment path to... However, this will increase latency by additional hops. The pre-migration mechanism involves migrating the VNF from node 7 to node 6 and changing the deployment path to... However, this introduces additional migration latency. This invention employs a VNF pre-migration mechanism, that is, in... The VNF migration process is initiated at specific times, allowing data to flow from the source node to the VNF migration node, and then to the destination node. Specifically, based on the interruption information predicted by the link risk awareness model, at specific times... The VNF instance originally deployed on node 7 is pre-migrated and instantiated to the target node 6. At the same time, the original deployment path is switched to a new path 1→5→6→10→11 through node 6, so that the subsequent data flow is forwarded in the order of "source node → VNF migration node → destination node". This completes the migration before the link enters a high-risk state, reducing the impact of the migration process on the overall latency.

[0113] Step 7: If the service deployment process meets resource and latency constraints and no high-risk links cause service interruptions, the service deployment is successful. The deployment plan for the service request is obtained, and the multidimensional heterogeneous network resources are updated. Otherwise, the service deployment fails if the service deployment constraints are not met.

[0114] Upon successful service deployment, the service deployment plan is recorded and network resources are updated; otherwise, service deployment fails. Updating network resources refers to adjusting the allocated CPU computing resources based on the final VNF placement and virtual link mapping results after successful service deployment. and storage resources The virtual link bandwidth requirement is deducted from the available resources of the corresponding satellite node, and the virtual link bandwidth requirement is deducted from the available bandwidth of its mapped physical link. The resources are deducted from the network; when the service life ends and it exits the network, the node resources and link bandwidth occupied by the service are returned to the available resources of the corresponding physical node and physical link, thereby maintaining the real-time update of the physical network resource status for subsequent service deployment decisions.

[0115] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for deploying dynamic service function chains with link status awareness in an F6G low-Earth orbit satellite optical network, characterized in that: Includes the following steps: Step 1: Based on the network topology and network resource information of the F6G low-Earth orbit satellite optical network, obtain the TLE data of space debris and initialize the service function chain request information; Step 2: Preprocess the TLE data of spatial fragments and use the KMeans clustering algorithm to cluster the preprocessed data; Step 3: Construct a space debris risk perception model based on the Transformer model and predict the risk level of the debris, determine the spatial relationship between the debris trajectory and the inter-satellite laser link, and calculate the duration of the risk link interruption and the comprehensive risk value of the link based on the inter-satellite laser link interruption risk assessment model. Step 4: Construct the link logical distance based on multi-attribute decision theory, and construct the weighted satellite optical network topology based on the link logical distance; Step 5: Use the K-shortest path algorithm to process the weighted satellite optical network topology to obtain k optimal pre-deployment paths; Step 6: Construct an optimization problem with the constraints of latency, computation, storage, and communication resources of service function chain requests, and the goal of minimizing the average service latency. Perform joint optimization of flow routing and VNF placement. Step 7: If the service deployment process meets the constraints and there are no high-risk links causing service interruption, the service deployment is successful, the service deployment plan is obtained, and network resources are updated; otherwise, the service deployment fails.

2. The method for deploying dynamic service function chains with link status awareness in the F6G low-Earth orbit satellite optical network according to claim 1, characterized in that: The network topology information includes the set of satellite nodes, the set of satellite-to-ground nodes, the set of inter-satellite laser links and satellite-to-ground links and their connection relationships, the physical distance of the links and the propagation delay; the network resource information includes the CPU computing resources and storage resources currently available for each physical node, as well as the available bandwidth resources for each physical link. The network topology of a dynamic low-Earth orbit satellite optical network is represented using a time-spreading graph model. The physical network is abstracted as a set of weighted undirected graphs, and the mathematical model of the physical network is as follows: ,in, For a weighted undirected graph, Indicates a time range. This represents the t-th discrete time slice. Represents a set of physical nodes. Represents the set of physical links, for a node The set of attributes includes CPU computing resources and storage resources Inter-satellite laser link between node m and node n The set of attributes includes available bandwidth resources. , This represents a set of satellite nodes that have computing and storage capabilities and can support VNF ​​instances; The service function chain request information includes: the number of service function chain requests, the VNF type and processing order corresponding to each service request, the computing and storage resources required by each VNF, the bandwidth requirements of the virtual links between VNFs, and the maximum end-to-end tolerable latency; N service function chain requests are generated using a Poisson distribution model, and the service lifetime is generated using an exponential distribution model; the service function chain request is described as a six-tuple. ,in, Indicates the composition of a service request. An ordered VNF set Represents a set of virtual links. and These represent the computational and storage resources required by the VNF, respectively. This indicates the bandwidth resources required for the virtual link. Indicates a service request Maximum tolerable end-to-end delay; Define binary decision variables Indicates whether a VNF node has been placed on an LEO node, defining a binary decision variable. Indicates virtual link Mapped to inter-satellite laser link superior; The satellite / ground nodes, inter-satellite and satellite-to-ground links, and their available computing, storage, and bandwidth resources within each time slice are organized into a unified data structure, and links that are invalid or have no available bandwidth are eliminated. At the same time, the generated arrival time, lifetime, and VNF sequence and resource requirements of each SFC are combined into a standardized service request queue.

3. The method for deploying dynamic service function chains with link status awareness in the F6G low-Earth orbit satellite optical network according to claim 2, characterized in that: The method for preprocessing the TLE data of space debris is as follows: cleaning the records with missing or abnormal TLE data of space debris and retaining samples with complete orbital elements; filtering out information highly related to the interruption of inter-satellite laser links, including the number of times the space debris orbits the Earth per day and the first time derivative of the average motion, and randomly assigning each space debris a debris size within 1m to construct an unlabeled debris feature dataset. The method of using the KMeans clustering algorithm to cluster the debris dataset is as follows: the preprocessed daily number of orbits around the Earth, the first time derivative of the average motion, and the three-dimensional features of the debris size are used as input data for the KMeans clustering algorithm. From binary classification to multi-class classification, the number of clusters K=3 is determined by combining two indicators: the silhouette coefficient and the sum of squared center distances. The space debris is divided into three categories: zero risk, low risk, and high risk, and the corresponding data are labeled with 0, 1, and 2 respectively.

4. The method for deploying dynamic service function chains with link status awareness in the F6G low-Earth orbit satellite optical network according to claim 2 or 3, characterized in that: The method for predicting the risk level of risk fragments is as follows: the clustered spatial fragment dataset is used as the input parameter of the spatial fragment risk perception model to perform risk level classification prediction of spatial fragments. Fragment feature data, which have been clustered using the KMeans clustering algorithm and assigned risk labels of 0, 1, and 2, are input into the space debris risk perception model. During the model training phase, the cross-entropy loss function is selected as the loss function, and the Adam optimizer is used to update the model parameters. The number of training epochs is set to 1000, and the learning rate is adjusted within the range of 0.001 to 0.

1. The number of times the Earth orbits each day, the first-order time derivative of the average motion, and the size of the debris are input into the trained space debris risk perception model to obtain classification prediction results for three risk levels: 0, 1, and 2. Filter out data with zero risk; For space debris with non-zero risk, the risk assessment model for inter-satellite laser link interruption caused by space debris is used to calculate the duration of link interruption, and the start and end times of the corresponding link interruption, as well as the number of the interrupted link, are obtained accordingly.

5. The method for deploying dynamic service function chains with link status awareness in the F6G low-Earth orbit satellite optical network according to claim 4, characterized in that: The method for determining the spatial relationship between the debris trajectory and the inter-satellite laser link is as follows: a three-dimensional spatial geometric model is established based on the precise orbital parameters of the debris and the satellite. By calculating the parameters of the orbital plane intersection line and the minimum distance, it is determined whether the spatial relationship is perpendicular intersection, general intersection, or special tangency. Based on the spatial relationship between the trajectory of space debris and the inter-satellite laser link, the link interruption scenario is divided into three cases: perpendicular intersection, general intersection, and special tangency. Calculate the link interruption duration in a vertically intersecting scenario ; Calculate the link interruption duration in typical intersecting scenarios ; Calculate the duration of link interruption in special tangential scenarios ; Among them, the orbital central angle parameter ; For fragment size, Represents the orbital radius of the space debris. Indicates the width of the laser beam. Indicates the speed at which space debris travels. This represents the angle between the trajectory of space debris and the inter-satellite laser link. The arc length representing the trajectory of space debris; and , When the inter-satellite laser link is in a high-risk, low-risk, or zero-risk state, a link risk impact factor is set. The combined risk value of the inter-satellite laser link is represented by infinity, a constant value, and 1, respectively. in, This indicates the distance between space debris and the satellite at the launch site.

6. The method for deploying dynamic service function chains with link status awareness in the F6G low-Earth orbit satellite optical network according to claim 5, characterized in that: The link logical distance The calculation formula is: Among them, w1, w2, and w3 are weighting factors; Indicates LISL link Available bandwidth resources in the current time slot; Indicates LISL link Maximum bandwidth resources; Represents the set of all LISL links; Indicates LISL link The instantaneous physical distance; Represents a set The maximum instantaneous physical distance of all LISL links in the network; Indicates LISL link The overall risk value; Indicates link The maximum overall risk value during the assessment period; The method for constructing a weighted satellite optical network topology based on link logical distance is as follows: In the satellite network topology, the link logical distance W of each LISL link is used as the weight of the corresponding edge, forming a weighted undirected graph. The data on each edge represents the link logical distance value of the inter-satellite laser link within the current time slot. The entire satellite network cycle is discretized, and the network topology within a time slot is treated as a quasi-static network. A two-dimensional matrix of link weights is set for the t-th discrete time slot. for: ； in, , , , , , , , , , , These represent the weights of the corresponding edges between physical nodes 1, 2, ..., m within the t-th discrete time slice.

7. The method for deploying dynamic service function chains with link status awareness in the F6G low-Earth orbit satellite optical network according to claim 6, characterized in that: Within each time slot, based on the latest values ​​of available link bandwidth, physical distance, and overall link risk, the logical link distance W for each LISL is recalculated, and the two-dimensional matrix is ​​updated. The corresponding elements ensure that the weighted satellite optical network topology always reflects the latest link status; The method for obtaining k optimal pre-deployment paths is as follows: using the source satellite node and the destination satellite node as endpoints, the KSP algorithm is run on the weighted satellite optical network topology to iteratively find and return the path with the shortest cumulative link logical distance among the first k paths, thus obtaining k optimal pre-deployment paths.

8. The method for deploying a link-state-aware dynamic service function chain in an F6G low-Earth orbit satellite optical network according to any one of claims 5-7, characterized in that: The optimization problem ； ； ； ； ； ； ； ； in, and These represent service requests respectively. The Middle VNF nodes and the VNF nodes Place on a physical node Decision variables; Represents the set of all service requests The total number of service requests; This indicates that the k-th service request is constituted. An ordered VNF set; and These are the source VNF nodes. and target VNF ​​node The decision variable for the placement location; Indicates a service request Middle Virtual Link Is it mapped to a physical link? Decision variables; Indicates a service request The original end-to-end latency without VNF ​​migration; Indicates a service request Additional latency due to VNF migration; This indicates the number of service requests successfully deployed during the simulation period.

9. The method for deploying dynamic service function chains with link status awareness in the F6G low-Earth orbit satellite optical network according to claim 8, characterized in that: The optimization problem is solved heuristically: given the physical network and link logical distance weights, the K-shortest path algorithm is used to obtain a set of candidate pre-deployment paths for each service request. On this set, the resource availability and latency of each candidate pre-deployment path are checked sequentially according to the constraints. Once a candidate pre-deployment path that satisfies all constraints is found, service pre-deployment is performed, determining the VNF placement location and virtual link mapping result for the service request, thus providing a joint optimization solution for flow routing and VNF placement. Otherwise, other candidate pre-deployment paths are checked. If all pre-deployment paths in the candidate pre-deployment path set fail, deployment fails. Service pre-deployment is for service requests on candidate pre-deployment paths. Allocate the corresponding physical node and physical link resources, and determine the corresponding decision variables. and The value of is selected and recorded as the deployment scheme for the current service request; Determine if there are high-risk links in the pre-deployment path within the lifecycle of the service request. If so, immediately start the pre-migration mechanism to allow the data flow directly from the source node to the target migration node.

10. The method for deploying dynamic service function chains with link status awareness in the F6G low-Earth orbit satellite optical network according to claim 9, characterized in that: The space debris risk perception model processes the three-dimensional features of space debris and the corresponding 0, 1, and 2 risk labels, and combines the link interruption duration and comprehensive risk value calculation results to form a risk status sequence for each LISL in each time slot. Based on this risk status sequence, the lifetime interval of the service request is compared with the high-risk time interval of all inter-satellite laser links on the pre-deployed path. If a link is marked as high-risk during the service lifetime, it is determined that there is a high-risk link in the pre-deployed path, triggering the pre-migration mechanism. The pre-migration mechanism allows data to flow from the source node to the VNF migration node, and then to the destination node; The method for updating network resources is as follows: After the service function chain request is successfully deployed, based on the final VNF placement result and virtual link mapping result, the occupied CPU computing resources are adjusted. and storage resources The virtual link bandwidth requirement is deducted from the available resources of the corresponding satellite node and from the available bandwidth of the physical link it maps to. When the service life ends and the service exits the network, the node resources and link bandwidth occupied by the service function chain request are returned to the available resources of the corresponding physical node and physical link.

Images