SFC deployment and migration method in giant low earth orbit satellite constellation of mobile perception

Abstract

The invention discloses an SFC deployment and migration method in a giant low-orbit satellite constellation of mobile perception, and the method comprises the steps: 1, carrying out the information exchange of all controllers according to an SFC set to be deployed, and constructing a simplified topology; step 2, executing a subnet selection algorithm by each controller in parallel; step 3, performing network information exchange by each controller to obtain a VNF (Virtual Network Function); 4, the current controller obtains all the tasks to be executed; and step 5, each controller performs network information exchange again, the SFC set failed in deployment is put into the SFC set waiting for deployment of the next execution process according to information transmitted by other subnets, the step 1 is returned, the satellite network is divided into a plurality of subnets according to the orbit and the coverage range of the satellite, and the sub-networks are distributed in the next execution process. The satellites in each subnet are relatively fixed and are close to each other, so that relatively low communication delay can be ensured, and a simplified topological structure is designed in each subnet, so that a controller in the subnet can efficiently manage and deploy the SFC.

Description

Technical Field

[0001] The present invention belongs to the technical field of service function chain deployment, and in particular relates to a SFC deployment and migration method in a mobile-aware giant low-orbit satellite constellation. Background Art

[0002] As emerging applications become more and more complex, it is difficult to effectively support large-scale services by relying solely on ground networks, especially for ultra-long-distance real-time services. Traditional ground networks have long transmission distances and high delay sensitivity, making it difficult to ensure end-to-end service continuity. Therefore, satellite networks will play an important role in the sixth-generation (6G) communication system, especially in providing services to users in remote areas of the ocean and sparsely populated areas. In satellite systems, low-orbit satellites have received widespread attention from the scientific and industrial communities because of their lower development costs, launch costs and lower latency compared to geostationary orbit satellites and medium-earth orbit satellites. For example, SpaceX's Starlink plans to launch more than 40,000 low-Earth orbit (LEO) satellites for global network services.

[0003] In order to better manage such large-scale networks, Network Function Virtualization (NFV) and Software Defined Network (SDN) technologies have been widely used. NFV aims to replace dedicated network devices by building virtual network functions (VNFs) in software. VNFs can be automatically deployed and configured to achieve automated lifecycle management. Since a task is often completed by multiple network functions, these VNFs are linked in sequence to form a service function chain. SDN aims to separate the network control plane from the data forwarding plane and allow network administrators to manage network behavior programmatically through an open northbound interface. By combining SDN and VNF technologies, dynamic and cost-effective business guidance can be provided between VNFs, achieving efficient and flexible service function chain (SFC) deployment without traditional manual configuration.

[0004] After introducing SDN and NFV technologies into satellite systems, the scientific and industrial communities have proposed many management architectures. In large-scale networks such as LEO constellations, the distance between satellites is relatively large. If the control layer only contains a single controller, the resulting delay is very high, making it difficult to monitor the network status in real time. Therefore, a distributed control layer is necessary. At the same time, due to the high orbital altitude of geostationary orbit (GEO) satellites, the communication round-trip time (RTT) exceeds 600 milliseconds, which is too long for some services. The average RTT of LEO satellites is 50 milliseconds, which can provide network bandwidth with low latency. Therefore, deploying controllers on low-orbit satellites can achieve real-time network status monitoring and lower signaling delays.

[0005] In terms of SDN architecture, many studies have proposed a series of control methods for satellite networks. For single-layer satellite networks, there are two different ways to establish the SDN control plane of the LEO constellation, namely out-of-band control and in-band control. In the out-of-band control method, a separate channel is established for the control service, namely the ground station or GEO and MEO satellites, and the LEO satellite only responds to data transmission. Typical out-of-band control groups LEO satellites and manages them in clusters. The SDN controller is deployed in the data center of the ground station, using artificial intelligence technology to collect network information and distribute the flow table calculated by the control center. In in-band control, the control plane and the data plane share LEO satellites with different priorities instead of using ground stations. Typical in-band control often considers two costs. The first is the migration cost in the dynamic controller placement scenario, which is caused by the change of controller location and switching to controller allocation. The second is the reconfiguration cost caused by the reconfiguration messages exchanged between the SDN controller and the satellite SDN switch.

[0006] In terms of SFC deployment solutions, they can be classified from the aspects of optimization objectives, network architecture and optimization algorithms. From the perspective of optimization objectives, they can be divided into delay, bandwidth utilization, resource consumption, reliability, etc. The goal of the delay minimization solution is to minimize the total delay of data packets when passing through SFC, such as path selection and function deployment based on delay constraints. The goal of the bandwidth maximization solution is to maximize the utilization of link bandwidth so that network resources remain balanced and avoid network bottlenecks. The resource minimization solution focuses on minimizing the consumption of resources such as computing, storage and energy to ensure the interests of service providers. The reliability maximization solution focuses on the robustness of deployment and the reliability of services, such as improving the fault tolerance of the network through multi-path routing and redundant deployment. From the perspective of network architecture, it can be divided into centralized deployment solutions, distributed deployment solutions and hybrid deployment solutions. In the centralized deployment algorithm, all decisions are made by a central control node, which is suitable for networks with centralized resource management and control. In the distributed deployment algorithm, decisions are made by multiple nodes in a distributed manner, which is suitable for self-organizing networks and dynamically changing network environments. The hybrid deployment algorithm combines the advantages of centralized and distributed methods and can be flexibly applied under different network conditions. From the perspective of optimization algorithms, it can be divided into mathematical programming algorithms, heuristic and meta-heuristic algorithms, and machine learning algorithms. Mathematical programming algorithms use mathematical methods such as linear programming and integer programming to solve SFC deployment problems. Heuristic algorithms can quickly find better solutions using large-scale networks, such as greedy algorithms and local search. Meta-heuristic algorithms use systematic exploration and utilization strategies to perform global searches in the solution space to find the best possible solution, such as genetic algorithms, particle swarm optimization (PSO), simulated annealing (SA), etc. Machine learning algorithms are also used to predict and optimize SFC deployment, such as reinforcement learning (RL) and deep learning (DL) to achieve automated deployment and adapt to dynamic and complex network environments.

[0007] Compared with the method of directly deploying in LEO satellites, the out-of-band control method, which places the controller in GEO or ground stations, has higher latency, poor link reliability, low bandwidth utilization, insufficient dynamics and flexibility, limited fault recovery and redundancy mechanisms, and scalability issues. Due to the high altitude of GEO satellites, the round-trip time of signals from LEO to GEO satellites is long, which becomes a bottleneck restricting this control method. Higher latency affects real-time control and rapid response. Although the distance between the ground station and the LEO satellite is short, due to the rapid movement of LEO satellites, the communication link between the ground station and the satellite needs to be switched frequently, which will also introduce additional latency and switching overhead. In terms of link reliability, although GEO satellites provide a relatively stable link, due to the longer distance, signal attenuation and interference risks increase, which may lead to reduced communication quality. Due to the continuous movement of LEO satellites, the communication link between the ground station and the LEO satellite needs to be established and switched frequently, which increases the risk of communication interruption and data packet loss. In terms of bandwidth utilization, when the controller is deployed on the GEO satellite, the control signaling needs to be transmitted through a longer distance link, occupying precious satellite communication bandwidth resources. , which reduces the bandwidth utilization efficiency of the overall network. In terms of flexibility, due to the fixed position of GEO satellites, it is difficult to flexibly adapt to the dynamic changes of the LEO satellite network, especially the topology changes of the fast-moving LEO satellite network. The ground station coverage is limited, and it is difficult to grasp the status and changes of the LEO satellite network in real time and globally, affecting the flexibility and real-time performance of network control. In terms of fault recovery and redundancy, once a GEO satellite fails, the recovery and replacement costs are high and the time is long, which affects the reliability of the overall network. Ground station failure or communication interruption with LEO satellites will directly affect the transmission of control signaling, and the establishment of redundant links requires more ground infrastructure support, increasing costs and complexity. In terms of scalability, GEO satellites: Due to the limited number of GEO satellites, it is difficult to support the control needs of large-scale LEO satellite networks, which limits the scalability of the network. The ground station control method requires the addition of a large number of ground stations to cover a wide range of LEO satellite networks, which increases the complexity of deployment and maintenance. These shortcomings make the controller deployment method of GEO satellites or ground stations less effective than the method of directly deploying controllers in LEO satellites in the rapidly changing and high-demand LEO satellite network environment.

[0008] Centralized control methods often have problems such as single point failure, poor scalability, high latency and bandwidth overhead, and low flexibility and security. If a centralized controller fails, the control function of the entire network may be interrupted, resulting in the inability to deploy and manage the service function chain, which has a serious impact on the availability and reliability of the network, especially in critical missions and services, which may cause serious service interruptions. In terms of scalability, centralized control methods require all control decisions to go through a central controller, which may cause the controller to be overloaded in large-scale satellite networks. When the network scale expands, the centralized controller may not be able to handle all control requests, resulting in performance bottlenecks and prolonged response time. In terms of latency and bandwidth overhead, the centralized controller needs to collect status information from each node in the network and send control commands to them, which increases communication delays, especially in satellite networks. In the process of satellite communication, this problem is more serious due to the large communication delay between satellites and between satellites and the ground. Frequent status information collection and control command distribution will occupy a lot of bandwidth resources and reduce the efficiency of data transmission. In terms of flexibility and adaptability, satellite networks, especially LEO satellite networks, are characterized by dynamic changes. The rapid movement of satellites and frequent topological changes require rapid adaptation and adjustment. The centralized control method requires all decisions to be made through a central controller, so the perception of local changes in the network is not timely enough, and the SFC deployment strategy cannot be quickly adjusted to adapt to the current network status. In terms of security, once the centralized controller becomes a potential target of attack and is successfully attacked, the control function of the entire network may be paralyzed. The distributed control method improves the reliability, flexibility and scalability of the system by distributing control tasks, and is more suitable for dynamically changing and large-scale satellite network environments.

[0009] The service function chain deployment method based on the traditional ground network or the satellite network service function chain deployment scheme in static and dynamic scenarios cannot be well applied to large-scale satellite networks. Although these studies have good performance, they are not suitable for large-scale satellite network scenarios. Based on the above technical problems existing in the prior art, the present invention proposes a SFC deployment and migration method in a mobile-aware giant low-orbit satellite constellation. Summary of the invention

[0010] The purpose of the present invention is to address the deficiencies of the prior art and provide a method for SFC deployment and migration in a mobile-aware giant low-orbit satellite constellation. The satellite network is divided into subnets and a simplified topology is constructed to guide the controller to deploy SFCs. In order to provide continuous and low-latency services in dynamic scenarios, reduce system resource consumption, and minimize the number of service interruptions caused by migration, the concept of subnet proximity centrality in the time dimension is introduced. By utilizing the predictable characteristics of satellites, the SFC deployment and migration problems in VNF sharing scenarios are solved, and the frequency of migration is reduced.

[0011] The present invention adopts the following technical solution:

[0012] A method for deploying and migrating SFCs in a giant low-orbit satellite constellation with motion awareness, comprising:

[0013] Step 1: Based on the SFC set to be deployed, each controller exchanges information to build a simplified topology;

[0014] Step 2, each controller executes the subnet selection algorithm in parallel;

[0015] Step 3: Each controller exchanges network information to obtain the VNF that needs to be deployed in the current controller management subnet;

[0016] Step 4: The current controller obtains all tasks to be executed, executes the deployment algorithm in the subnet, and if the execution is successful, it is added to the successful execution list. If the execution fails, the deployment is not performed, and a network transmission success or failure signaling is sent;

[0017] In step 5, each controller exchanges network information again, updates the SFC sets of successful and failed deployments according to the information transmitted by other subnets, puts the SFC sets of failed deployments into the SFC sets waiting to be deployed in the next execution process, and returns to step 1.

[0018] Further, in step 2, according to The simplified node proximity centrality in the time dimension is used to indicate the importance of the subnet in the deployment process. The larger the reciprocal value of the sum of the shortest paths from a node to other nodes, the closer the node is to other nodes. The larger the proximity centrality between the simplified node and the simplified nodes representing the source and destination subnets, the shorter the path connecting the source and destination subnets through the destination subnet. Suppose there are P*Q subnets, and the simplified topology constructed has a total of P*Q nodes representing each subnet. Let [t1,t n ] The simplified node representing the source subnet is The simplified nodes representing the destination subnet are

[0019] In t i Calculate P*Q simplified nodes to the source at any time and destination The closeness centrality is presented in the form of a weight matrix:

[0020]

[0021] Indicates t i Simplify the node (p,q) to and The closeness centrality of , b(a,b) represents the number of hops between nodes a and b calculated based on the abstract link;

[0022] [t m ,t n ]The weight matrix within the time range is accumulated to obtain:

[0023]

[0024] Reflection m ,t n ] time to simplify the average distance value of the node (p, q) to each access subnet. The larger the average distance value, the smaller the average distance. Thus, the closeness centrality of the subnet is obtained.

[0025] Furthermore, in step 2, the subnet where the SFC is to be deployed is selected based on the closeness centrality of the subnet;

[0026] make Represents the cumulative weight matrix The subnet number with the largest value is taken, that is, the subnet number with the shortest average distance. If there are multiple subnet numbers with equal weights, the subnet with the most remaining resources is taken to achieve load balancing;

[0027] Assume t m Continuously calculate the service arrival time until:

[0028]

[0029] Via subnet connect The number of hops is greater than At this time, the subnet that the average shortest path passes through is not within the rectangular range formed by the shortest path at the next moment. Continuing to accumulate the overlapping subnets will cause additional path loss, and the accumulation will stop. n+1 The time service has left and the accumulation calculation will also stop;

[0030] The shortest path Subnets on The weights in the matrix are sorted in descending order to obtain the set Assume that the final selected deployment subnet is S. In order to make the remaining resources in the selected set S sufficient for SFC to request u to newly establish all VNFs and generate a smaller redundant path, the weights are ranked from large to small. The subnets in join S until:

[0031]

[0032] Assume that there are N*M satellites in each subnet. is the average resource of each satellite node, l is the average number of satellites passing through the subnet to connect two adjacent subnets, let {U1, U2, U3, U4} be the set of satellites on the four boundaries, P j ={(v j ,v j' )|v j ∈U j ,v j' ∈U j' ,j'≠j} is U j The path from the boundary satellite to other boundary satellites, len() represents the number of satellites on the path, |P j | represents the number of paths in the set, and l is calculated as follows:

[0033]

[0034] It is used to characterize the remaining resources on the satellites when traversing a subnet on average without redundant paths.

[0035] Furthermore, in step 3, the local subnet deploys the VNF, and the deployed VNF is added to the local tasks to be executed.

[0036] Furthermore, in step 3, other subnets deploy VNFs and transmit deployment information through network signaling.

[0037] Furthermore, in step 3, when each subnet deploys VNF, it first searches for reusable VNFs in the n-hop neighbor satellites adjacent to the access satellite. If this fails, it searches for nodes within the subnet where new VNFs can be established. During the search, the nodes are sorted according to the weight cist(v i ) are sorted, where

[0038] If v i For newly created VNF Satellite:

[0039]

[0040] If v i To reuse VNF in Satellite:

[0041]

[0042] Among them, I is related to v i The set of nodes with inter-satellite links, r() represents the remaining resources, α>β, and p(a,b) represents the link length between a and b calculated by the minimum hop model.

[0043] Furthermore, in step 1, a simplified topology is constructed to guide the distributed controller to make decisions. The simplified topology is used to simplify the state within the subnet, emphasize the connection relationship between subnets, abstract key nodes and links, and ignore secondary nodes and links.

[0044] Furthermore, in step 3, when the subnet controller receives the signaling of the deployment sub-SFC, the placement of the VNF and the routing selection between VNFs are completed, and the controller takes the central satellite on the border adjacent to the previous deployed subnet as the temporary source and the central satellite on the border adjacent to the next subnet as the temporary destination node.

[0045] Further, in step 3, if it is the first or last subnet, the satellite with the minimum hops from the source and destination nodes is selected as the temporary source and temporary destination nodes respectively.

[0046] The beneficial effects of the present invention are:

[0047] 1. The SFC deployment and migration method in the giant low-orbit satellite constellation with mobile perception described in the present invention divides the satellite network into multiple subnets according to the orbit and coverage of the satellite. The satellites in each subnet are relatively fixed and close to each other, which can ensure low communication delay. A simplified topology structure is designed inside each subnet, so that the controller in the subnet can efficiently manage and deploy SFC. At the same time, by establishing a reasonable control strategy between subnets, it is ensured that the SFC deployment across subnets can also proceed smoothly.

[0048] 2. The SFC deployment and migration method in the giant low-orbit satellite constellation with mobile awareness described in the present invention introduces a distributed algorithm, so that the controller can adjust the SFC deployment strategy in real time in a dynamically changing network environment to ensure service quality and resource utilization. In this way, not only the delay problem caused by centralized control is solved, but also the flexibility and scalability of the system are enhanced, which can better cope with future complex network requirements and challenges;

[0049] 3. The SFC deployment and migration method in the mobile-aware giant low-orbit satellite constellation of the present invention analyzes the time characteristics of each subnet in the satellite network through the concept of subnet proximity centrality, and determines the subnets with higher connection stability and lower communication delay in a specific time period. These subnets are preferentially selected for deploying VNF nodes, thereby minimizing the migration requirements of VNF nodes while ensuring the quality of service. In this way, it is ensured that the SFC can maintain stable and efficient operation in a dynamically changing network environment;

[0050] 4. The SFC deployment and migration method in the mobile-aware giant low-orbit satellite constellation described in the present invention utilizes the predictable orbit characteristics of the satellite to plan the migration path and time points of the VNF in advance, avoiding unnecessary frequent migration; in the actual deployment process, through the accurate prediction of the satellite orbit in the future, the optimal VNF deployment strategy can be formulated, so that when the satellite network changes, the overall performance of the SFC will not be significantly affected, which not only reduces the consumption of system resources, but also significantly reduces the service interruption caused by migration, ensuring that users can enjoy stable, low-latency network services at any time. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a schematic diagram of satellite orbit parameters in an embodiment of the present invention;

[0052] Figure 2 A schematic diagram of subnet division under distributed control in an embodiment of the present invention;

[0053] Figure 3 A schematic diagram of a simplified topology in an embodiment of the present invention;

[0054] Figure 4 A schematic diagram of a subnet switching process in an embodiment of the present invention;

[0055] Figure 5 It is a flow chart of a method for deploying and migrating SFCs in a giant low-orbit satellite constellation with mobile awareness according to an embodiment of the present invention;

[0056] Figure 6 A schematic diagram of bandwidth resources consumed by each SFC on average as the number of SFCs changes in an embodiment of the present invention;

[0057] Figure 7 A schematic diagram of node resources consumed by each SFC on average as the number of SFCs changes in an embodiment of the present invention;

[0058] Figure 8 A schematic diagram of the average bandwidth occupied by each SFC request during a change over time in an embodiment of the present invention;

[0059] Fig. 9 A schematic diagram of the node computing resources occupied by each SFC request on average over time in an embodiment of the present invention;

[0060] Fig.10 A schematic diagram of the SFC ratio that migrates over time in an embodiment of the present invention;

[0061] Fig.11 Schematic diagram of SFC acceptance rate changing over time in an embodiment of the present invention. DETAILED DESCRIPTION

[0062] In order to more clearly understand the above-mentioned purposes, features and advantages of the present invention, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0063] Example

[0064] The SFC deployment and migration method in the mobile-aware giant low-orbit satellite constellation includes three parts: building a simplified topology snapshot, selecting a deployment subnet, and an SFC deployment algorithm within the subnet;

[0065] To facilitate understanding, we first introduce the common key parameters in satellite networks and the formula for the number of hops between any two satellites:

[0066] A Walker Delta constellation can usually be expressed as α:PQ / P / F, where P means there are P orbits evenly distributed along the equator, Q means there are Q satellites evenly distributed in each orbit, and all satellites are distributed on circular orbits with the same orbital inclination α and altitude h, and F represents the phase difference between adjacent orbits;

[0067] Figure 1 The difference in right ascension (RAAN) between adjacent orbits Depending on the number of orbits, the phase difference between adjacent satellites in the same orbit Depending on the number of satellites in one orbit, the phase shift represents the difference in latitude between two neighbors in adjacent tracks;

[0068] According to the flight direction, all satellites can be divided into two categories: ascending satellites and descending satellites. The former fly in the direction of increasing latitude, while the latter fly in the direction of decreasing latitude. When the orbital inclination α is between 0 and 90 degrees, the directions of the ascending and descending orbit satellites are northeast and southeast respectively. If the satellite is in an ascending orbit, it is represented by A, and the descending orbit is represented by D. There are four modes: A2A, A2D, D2A, and D2D. The ascending or descending state of the satellite is not constant, but changes with the movement of the satellite. The sub-satellite point of a satellite can be expressed by longitude. And dimension λ represents:

[0069]

[0070] λ=ζ(u)+L0-ω e t,

[0071] Where u∈[-π,π] represents the phase difference between the satellite and the ascending node. Indicates that the satellite is flying northward in an ascending orbit. When ζ(u) is the longitude difference from the satellite to its ascending node, the longitude difference varies with the satellite phase. L0 is the initial longitude of the orbital ascending node, which is an absolute parameter for determining the position of the orbital plane. e is the Earth's rotation speed;

[0072] set up Figure 1 The subsatellite positions of sat1 and sat2 are respectively expressed as and The minimum number of hops between two satellite planes is:

[0073]

[0074] The minimum number of hops in a plane is:

[0075]

[0076] Since the network is a ring, the data packet can also reach the destination in the reverse direction. If the path in a given direction is too long, the data packet will go in the opposite direction. Therefore, ΔL0 and ΔU should be standardized to the interval [-π,π, let The minimum number of hops between satellites is:

[0077]

[0078] Due to the existence of ascending orbit and descending orbit, according to the orbit information of the source and destination satellites, for the four access modes of A2A, A2D, D2A, and D2D, the corresponding u and ζ(u) values ​​are shown in Table 1:

[0079] Table 1 Calculation formulas for key parameters

[0080]

[0081] The minimum number of hops between any two satellites is calculated with a very small error using the formula. Indicates the minimum number of hops between satellite a and satellite b.

[0082] In the embodiment of the present application, the satellite network is modeled as an undirected graph G=(N,L), where N represents a set of satellite nodes, L represents a set of links, and n represents a set of satellite nodes. i ∈N represents the i-th satellite, which has computing resources represents the number of VNFs of type f placed on it, l i,i' ∈L represents the connected satellite n i and n i' The link has bandwidth resources Under distributed control, the satellite network is divided into P×Q subnets, F represents a set of VNFs of different types, |F| represents the number of VNF types, and each f∈F represents a type of VNF. represents the computing resources required to build a VNF of type f, c f represents the processing capacity that f can provide, and U represents the set of all arriving SFC requests. Each u∈U is represented by a seven-tuple (F u ,C u ,S u ,D u ,B u ,b u ,T u ), Represents a list of VNFs in order, represents the i-th VNF in the SFC request, Represents VNF Is it of type f? express Whether to deploy on the node i , express and Whether the virtual link between them is deployed on link l i,i' superior, express The processing power required, S u and D u Indicates the source and destination of the SFC request, B u Indicates the bandwidth resources required by the request, T u represents the service time of the SFC request. Considering the mobility of the satellite, the original deployment scheme will cause the end-to-end route to become longer over time, and SFC migration becomes necessary. u is the bandwidth required for a VNF migration. express Is the migration path deployed on link l i,i' In the above, since a single SFC service depends on all VNFs, the service will be in a downtime state before all VNFs are migrated. Therefore, the embodiment of the present application adopts a method of migrating each VNF in parallel to minimize the downtime caused by the migration.

[0083] In SFC deployment, some constraints are set. Each VNF of an SFC can only be placed on one satellite at a time:

[0084]

[0085] Satellite nodes must have sufficient computing resources:

[0086]

[0087] The selected VNF instance must have sufficient processing power:

[0088]

[0089] The bandwidth used by the ISL cannot exceed the maximum bandwidth of the link:

[0090]

[0091] All constraints need to be satisfied in each snapshot, with the goal of reducing the consumption of node and link resources in the satellite network due to SFC deployment and migration, and also aiming to increase the acceptance rate and reduce the migration frequency;

[0092]

[0093] stC1-C4,C5:x,y,z,a∈{0,1}.

[0094] In an embodiment of the present application, the problem of SFC deployment in a satellite network can be systematically modeled and described through an integer linear programming model. In order to solve the problem, a heuristic algorithm is adopted. The heuristic algorithm has high computational efficiency and is much faster than ILP. Therefore, it is suitable for large-scale problems, and the algorithm implementation is relatively simple, easy to understand and apply. Although it cannot guarantee that the global optimal solution is found, through efficient heuristic design, the quality of the solution can be made as close to the exact solution as possible. Therefore, a heuristic algorithm is proposed.

[0095] Specifically, Figure 5 As shown, the method includes:

[0096] Step 1: Based on the SFC set to be deployed, each controller exchanges information to build a simplified topology;

[0097] Step 2, each controller executes the subnet selection algorithm in parallel;

[0098] Step 3: Each controller exchanges network information to obtain the VNF that needs to be deployed in the current controller management subnet;

[0099] Step 4: The current controller obtains all tasks to be executed, executes the deployment algorithm in the subnet, and if the execution is successful, it is added to the successful execution list. If the execution fails, the deployment is not performed, and a network transmission success or failure signaling is sent;

[0100] In step 5, each controller exchanges network information again, updates the SFC sets of successful and failed deployments according to the information transmitted by other subnets, puts the SFC sets of failed deployments into the SFC sets waiting to be deployed in the next execution process, and returns to step 1.

[0101] As a specific implementation, in step 1, Figure 2 As shown, in order to coordinate the distributed controllers to implement SFC deployment, the satellite network is divided into subnets, such as Figure 3 As shown in the figure, a simplified topology is constructed to guide the distributed controller to make decisions. The simplified topology is used to simplify the state within the subnet and emphasize the connection relationship between subnets. This simplified topology can abstract key nodes and links and ignore secondary nodes and links. The distributed controller is responsible for maintaining the satellite status information within the subnet. At the same time, it realizes cross-subnet collaborative deployment by exchanging a small amount of information and shares the simplified topology information with other controllers to ensure global optimization and coordination. Each controller makes independent deployment decisions based on the simplified topology and the real-time state within the subnet.

[0102] In the simplified topology, each subnet is represented by a simplified node. Suppose the constellation is divided into P×Q subnets, and (p,q) is used to represent the subnet number. The simplified topology is represented by in is a node set, represents a simplified node representing a subnetwork (p,q), and the simplified node has the attribute Indicates the total amount of computing resources remaining in the subnet, simplifying the node to have a location attribute That is, the latitude and longitude of the satellite at the center of each subnet, is a set of links, each Is a connection node and A simplified link with the property Contains horizontal and vertical jump counts, It is the node calculated by the minimum hop count model introduced earlier. and The minimum number of hops reflects the distance between two simplified nodes.

[0103] As a specific implementation method, in step 2, in the satellite network, the satellite moves very fast, resulting in the source and destination access subnets and the shortest path constantly changing. The subnet change process is as follows: Figure 4 As shown, assuming that a subnet switch occurs at snapshot time n, at snapshot time 1 and snapshot time n, the shortest paths with similar hop counts from source to destination are represented by dashed lines and dot-dash lines respectively. The two areas have an overlapping area, as shown by the subnet in the shaded area at snapshot time n. Deploying SFC in this overlapping area can reduce the length of the routing path and minimize the number of migrations.

[0104] Based on this, the embodiment of the present application proposes a simplified node proximity centrality in the time dimension to indicate the importance of the subnet in the deployment process;

[0105] The closeness centrality of a node is defined as: the reciprocal of the sum of the shortest paths from a node to other nodes. The larger the value, the closer the node is to other nodes.

[0106] In a snapshot, the greater the proximity centrality between the simplified node and the simplified nodes representing the source and destination subnets, the shorter the path connecting the source and destination subnets through the subnet. Assuming that there are P*Q subnets in total, the simplified topology constructed has a total of P*Q nodes representing each subnet. Assuming [t1,t n ] The simplified node representing the source subnet is The simplified nodes representing the destination subnet are

[0107] In t i Calculate P*Q simplified nodes to the source at any time and destination The closeness centrality is presented in the form of a weight matrix:

[0108]

[0109] Indicates t i Simplify the node (p,q) to and The closeness centrality, d(a,b) represents the number of hops between nodes a and b calculated based on the abstract link. The weight matrix should be centered on the rectangular range formed by the minimum number of hops through the subnet and show a decreasing trend around it, indicating that the subnet in the rectangle is in a more important position in the redeployment;

[0110] [t m ,t n ]The weight matrix within the time range is accumulated to obtain:

[0111]

[0112] Reflection m ,t n ] time to simplify the average distance of the node (p, q) to each access subnet. The larger the average distance value, the smaller the average distance. The weight matrix should be centered on the overlapping rectangular range formed by the minimum hop subnet within a period of time, and show a decreasing trend around it, indicating that the subnet within the overlapping range is in a more important position in the deployment. So far, the closeness center of the subnet is obtained.

[0113] Select the subnet to deploy SFC based on the subnet centrality. , which means that the cumulative weight matrix Take the subnet number with the maximum value, that is, the subnet number with the shortest average distance. If there are multiple subnet numbers with equal weights, take the subnet with the most remaining resources to achieve load balancing. Let t m Continuously calculate the service arrival time until:

[0114]

[0115] That is, through the subnet connect The number of hops is greater than At this time, the subnet that the average shortest path passes through is not within the rectangular range formed by the shortest path at the next moment. Continuing to accumulate the overlapping subnets will cause additional path loss, and the accumulation stops. If t n+1 The time service has left and the accumulation calculation will also stop;

[0116] The shortest path Subnets on The weights in the matrix are sorted in descending order to obtain the set Assume that the final selected deployment subnet is S. In order to make the remaining resources in the selected set S sufficient for SFC to request u to newly establish all VNFs and generate a smaller redundant path, the weights are ranked from large to small. The subnets in join S until:

[0117]

[0118] Among them, suppose there are N*M satellites in each subnet. is the average resource of each satellite node, l is the average number of satellites that pass through the subnet to link two adjacent subnets, and the calculation formula is as follows:

[0119] Assume {U1,U2,U3,U4} is the set of satellites on the four boundaries, P j ={(v j ,v j' )|v j ∈U j ,v j' ∈U j' ,j'≠j} is U j The path from the boundary satellite to other boundary satellites, len() represents the number of satellites on the path, |P j | indicates the number of paths in the set,

[0120]

[0121] It represents the remaining resources on the satellite that pass through a subnet on average without redundant paths. Because it is impossible to predict which boundary node will enter the subnet and which boundary node will leave the subnet, this is done to ensure that no matter which boundary node is accessed, even if there is no VNF required by the request in the subnet set S, the requested deployment can be completed through a newly established method and the generation of redundant paths is reduced.

[0122] As a specific implementation method, in step 3, when the subnet controller receives the signaling of deploying the sub-SFC, it needs to complete the placement of the VNF and the inter-VNF routing selection. The controller takes the central satellite on the border adjacent to the previous deployed subnet as the temporary source, and takes the central satellite on the border adjacent to the next subnet as the temporary destination node. If it is the first or last subnet, the satellite with the minimum number of hops from the source and destination nodes is selected as the temporary source and temporary destination node respectively. It is very important to consider the relationship between link and VNF usage when performing SFC deployment. Intuitively, when the link capacity is very limited, the number of servers is sufficient to support all traffic requirements. The requirements are routed on the shortest path, and idle virtual machines are started along the path to configure the required VNFs. On the contrary, when the link has sufficient capacity but only a few virtual machines are available, VNFs should be reused as much as possible, even if a longer path may be required to traverse these shared VNFs. How to balance the route length and the number of newly started VNFs is the focus of the embodiments of this application;

[0123] In order to give priority to using satellite nodes that have deployed corresponding VNFs without over-consuming link resources, the algorithm first searches for reusable VNFs within a smaller range of n hops. If this fails, it searches for nodes within the subnet where new VNFs can be established. In each search process, the nodes are ranked according to the weight cost (v i ) are sorted, and nodes that meet the constraints are selected from high to low weights. The weights comprehensively consider the remaining bandwidth of the links adjacent to the nodes, the remaining resources of the nodes, and the position of the nodes on the shortest path, so as to give priority to nodes with large remaining bandwidth and resources, close to the shortest path and close to the last deployed nodes. Finally, the purpose of comprehensively considering load balancing, reducing bandwidth and resource consumption and improving acceptance rate is achieved. The weight calculation formula for each satellite is given according to whether the VNF node is newly established or reused;

[0124] If v i For newly created VNF Satellite:

[0125]

[0126] If v i To reuse VNF in Satellite:

[0127]

[0128] Among them, I is related to v i The set of nodes with intersatellite links, r() represents the remaining resources, α>β, and p(a,b) represents the link length between a and b calculated by the minimum hop model;

[0129] When the service function chain is migrated, the method described in the embodiment of the present application will not perform migration every time a satellite switch occurs. On the contrary, when the controller recognizes that the number of hops in the current service routing path exceeds a certain threshold of the minimum number of hops, it will trigger the subnet selection and VNF deployment algorithm within the subnet to recalculate the deployment path and use the shortest path for VNF migration.

[0130] In order to verify the superior technical effect of the method described in the embodiment of the present application, SFCO-AMD, LASO-LAM, and FMC are used as benchmark solutions for performance evaluation, among which SFCO-AMD builds an abstract topology for multi-domain deployment, considers the distance from the node to the last node when deploying in each domain, LASO-LAM finds the shortest path through the Dijkstra algorithm, and finds the migration path through the Tabu search algorithm, and FMC proposes to use the Jaccard similarity coefficient as the weight for selecting a new migration path;

[0131] Table 2 summarizes the parameters used in the simulation. In the simulation, a LEO satellite network with 24 orbital planes arranged in a +Grid network topology is used. Each orbital plane has 66 satellites. In the dynamic scenario, the number of SFC requests arriving in each snapshot follows the Poisson distribution λ p = 40, for SFC requests, the nearest satellite is always selected as the access satellite, and the other simulation parameters and values ​​used are summarized in Table 2;

[0132] Table 2 Simulation parameters

[0133]

[0134]

[0135] By evaluating the bandwidth and node resource consumption status, the cost-saving performance of the algorithm deployment SFC can be fully reflected. Using different algorithms, the average bandwidth and node resource consumption of each SFC is as follows: Figure 6-7 As shown;

[0136] As the number of SFCs in the network increases, the average bandwidth consumed by each SFC increases. Since this experiment limits the selected subnet range to the shortest path, the additional bandwidth consumption is small. SFCO-AMD, LASO-LAM, and FMC migrate every time a switch occurs, resulting in a large number of migrations. Therefore, the migration bandwidth increases and the total bandwidth consumption increases. The reason why FMC has the highest consumption is that when selecting nodes, it does not consider whether the node is on the shortest path or close to the shortest path. According to the weight proposed by FMC, each time the VNF deployment location is determined, only the remaining bandwidth and node resources of the previous node's adjacent nodes are considered, and no "directionality" is given to the node;

[0137] By simulating dynamic scenarios, the performance of service function chain deployment solutions in real scenarios can be better reflected. Figure 8 and Fig. 9 Displays the consumption of network bandwidth resources and node resources of each SFC over time in each snapshot;

[0138] Since there are few SFCs in the network at the beginning, as the number of SFCs increases, the average resource consumption of each SFC shows an upward trend like the static scenario. As the number of SFCs in the network tends to be stable, the average bandwidth consumed by each SFC also tends to be stable. Since SFCO-AMD only considers the distance between the selected node and the previous node, LASO-LAM will migrate every time it switches, which increases the migration bandwidth and overall bandwidth consumption. FMC only considers the remaining bandwidth of adjacent nodes and also does not consider reducing the migration frequency to reduce the migration cost. LM-SFCP considers the position of the node on the shortest path and the characteristics of satellite movement, which greatly reduces the number of migrations and reduces the migration bandwidth.

[0139] Similar to the static environment, as the number of deployed VNFs in the network increases, the number of sharable VNFs increases, and the average node computing resources occupied by each SFC tends to decrease. SFCO-AMD prioritizes deployment on closer nodes, which may start more new VNFs. LASO-LAM and FMC consider the remaining node resources and link bandwidth, but do not consider using the deployed VNFs. LM-SFCP prioritizes the use of available VNFs, thereby reducing the number of activated VNFs and resource costs;

[0140] In dynamic scenarios, the original service solution will cause additional resource consumption as the satellite moves, which will bring a bad service experience to users. At this time, service migration is necessary. Although there are many studies aimed at reducing the impact of migration on users, migration will inevitably bring different degrees of downtime. Therefore, the number of migrations is an important indicator for evaluating algorithms. Fig.10 Shows the ratio of services that migrated to total services over time;

[0141] SFCO-AMD only uses the current time information to deploy VNFs on the shortest path. LASO-LAM focuses on minimizing the migration cost and performs migration whenever a satellite switch occurs. However, during the high-speed movement of the satellite, the original deployment plan may still be valid after the satellite switch and does not violate the delay constraint. FMC selects nodes with sufficient bandwidth to improve the migration success rate, but does not solve the problem of reducing the migration time. LM-SFCP focuses on the predictability of satellite movement and selects satellites that can provide long-term services with a shorter path, thereby reducing the migration time.

[0142] Fig.11 The acceptance rate of SFC requests at each moment is given. As time goes by, the number of SFCs in the network gradually increases, and the acceptance rate shows a downward trend. Since LM-SFCP uses resources more reasonably, the network consumes less resources and can accommodate more SFC requests, so the acceptance rate is the highest.

[0143] The present invention is not limited by the above embodiments. The above embodiments and descriptions are only for explaining the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention may be subject to various changes and improvements, which are within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the appended claims.

Claims

1. A method for deploying and migrating SFCs in a giant low-orbit satellite constellation with mobile awareness, characterized in that: include: Step 1: Based on the SFC set to be deployed, each controller exchanges information to build a simplified topology; Step 2, each controller executes the subnet selection algorithm in parallel; Step 3: Each controller exchanges network information to obtain the VNF that needs to be deployed in the current controller management subnet; Step 4: The current controller obtains all tasks to be executed, executes the deployment algorithm in the subnet, and if the execution is successful, it is added to the successful execution list. If the execution fails, the deployment is not performed, and a network transmission success or failure signaling is sent; In step 5, each controller exchanges network information again, updates the SFC sets of successful and failed deployments according to the information transmitted by other subnets, puts the SFC sets of failed deployments into the SFC sets waiting to be deployed in the next execution process, and returns to step 1.

2. The SFC deployment and migration method in a giant low-orbit satellite constellation with mobile awareness according to claim 1 is characterized in that: In step 2, according to The simplified node proximity centrality in the time dimension is used to indicate the importance of the subnet in the deployment process. The larger the reciprocal value of the sum of the shortest paths from a node to other nodes, the closer the node is to other nodes. The larger the proximity centrality between the simplified node and the simplified nodes representing the source and destination subnets, the shorter the path connecting the source and destination subnets through the subnets. Suppose there are P*Q subnets, and the simplified topology constructed has a total of P*Q nodes representing each subnet. Assume that [t1,t n ] The simplified node representing the source subnet is The simplified nodes representing the destination subnet are In t i Calculate P*Q simplified nodes to the source at any time and destination The closeness centrality is presented in the form of a weight matrix: Indicates t i Simplify the node (p,q) to and The closeness centrality of , d(a,b) represents the number of hops between nodes a and b calculated based on the abstract link; [t m ,t n ]The weight matrix within the time range is accumulated to obtain: Reflection m ,t n ] time to simplify the average distance between the node (p, q) and each access subnet. The larger the value, the smaller the average distance. Thus, the closeness centrality of the subnet is obtained.

3. The SFC deployment and migration method in a giant low-orbit satellite constellation with mobile awareness according to claim 2 is characterized in that: In step 2, the subnet where the SFC is to be deployed is selected based on the closeness centrality of the subnet; make Represents the cumulative weight matrix The subnet number with the largest value is taken, that is, the subnet number with the shortest average distance. If there are multiple subnet numbers with equal weights, the subnet with the most remaining resources is taken to achieve load balancing; Assume t m Continuously calculate the service arrival time until: Via subnet connect The number of hops is greater than At this time, the subnet that the average shortest path passes through is not within the rectangular range formed by the shortest path at the next moment. Continuing to accumulate the overlapping subnets will cause additional path loss, so the accumulation stops; if t n+1 The time service has left and the accumulation calculation will also stop; The shortest path The subnet on The weights in the matrix are sorted in descending order to obtain the set Assume that the final selected deployment subnet is S. In order to make the remaining resources in the selected set S sufficient for SFC to request u to newly establish all VNFs and generate a smaller redundant path, the weights are ranked from large to small. The subnets in join S until: Assume that there are N*M satellites in each subnet. is the average resource of each satellite node, l is the average number of satellites that pass through the subnet to link two adjacent subnets, let {U1, U2, U3, U4} be the set of satellites on the four boundaries, P j ={(v j ,v j' )|v j ∈U j ,v j' ∈U j' ,j'≠j} is U j The path from the boundary satellite to other boundary satellites, len() represents the number of satellites on the path, |P j | represents the number of paths in the set, and l is calculated as follows: Represents the remaining resources on the satellite that are passed through a subnet on average without redundant paths.

4. The SFC deployment and migration method in a giant low-orbit satellite constellation with mobile awareness according to claim 1, characterized in that: In step 3, the local subnet deploys the VNF, and the deployed VNF is added to the local tasks to be executed.

5. The SFC deployment and migration method in a giant low-orbit satellite constellation with mobility awareness according to claim 4 is characterized in that: In step 3, other subnets deploy VNFs and transmit deployment information through network signaling.

6. The SFC deployment and migration method in a giant low-orbit satellite constellation with mobility awareness according to claim 5, characterized in that: In step 3, when deploying VNFs in each subnet, it will first search for reusable VNFs in the n-hop neighbor satellites adjacent to the access satellite. If this fails, it will search for nodes within the subnet that can newly establish VNFs. During the search process, the nodes will be sorted according to the weight cost (v i ) are sorted, where If v i For newly created VNF Satellite: If v i To reuse VNF in Satellite: Among them, I is related to v i The set of nodes with inter-satellite links, r() represents the remaining resources, α>β, and p(a,b) represents the link length between a and b calculated by the minimum hop model.

7. The SFC deployment and migration method in a giant low-orbit satellite constellation with mobility awareness according to claim 1, characterized in that: In step 1, a simplified topology is constructed to guide the distributed controller to make decisions. The simplified topology is used to simplify the state within the subnet, emphasize the connection relationship between subnets, abstract key nodes and links, and ignore secondary nodes and links.

8. The SFC deployment and migration method in a giant low-orbit satellite constellation with mobility awareness according to claim 1, characterized in that: In step 3, when the subnet controller receives the signaling of the deployed sub-SFC, the VNF placement and inter-VNF routing are completed. The controller takes the central satellite on the border adjacent to the previous deployed subnet as the temporary source and the central satellite on the border adjacent to the next subnet as the temporary destination node.

9. The SFC deployment and migration method in a giant low-orbit satellite constellation with mobility awareness according to claim 8, characterized in that: In step 3, if it is the first or last subnet, the satellite with the minimum number of hops from the source and destination nodes is selected as the temporary source and temporary destination nodes respectively.

Images