Dynamic sfc mapping method and device for integrated satellite-ground network

Abstract

The application discloses a kind of dynamic SFC mapping methods and devices of star-ground integrated network, the method includes constructing star-ground integrated network, and searches every service stream all SFC mapping strategy in network;Obtain historical network traffic load and classify, to obtain multiple network traffic load;According to resource sharing rate value and network estimated income, search the optimal SFC mapping strategy set corresponding to each network traffic load, to obtain the final SFC mapping strategy set;The dynamic SFC mapping problem under scene is modeled as Markov decision process;Network state is acquired at the start time slot of each time period;The final SFC mapping strategy set constitutes action space, wherein, the mapping decision of each SFC mapping strategy set can change with the movement of satellite, to adapt to the dynamic characteristics of SGIN topology;Deep Q network is trained, so as to select SFC mapping strategy set according to network state at the start time slot of each time period, to maximize long-term network income.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mobile communication, in particular to a dynamic SFC mapping method of satellite-ground integrated network and a dynamic SFC mapping device of satellite-ground integrated network. BACKGROUND

[0002] 6G satellite-ground integrated network (SGIN) is considered as a typical 6G network architecture, which expands the coverage of wireless networks and is considered as an important development direction of 6G mobile communication networks; SGIN is composed of a ground network segment and a space network segment, and by utilizing the characteristics of rich computing resources of the ground network and wide communication coverage of the space network segment, it provides ubiquitous and high-quality network services for emerging applications.

[0003] In SGIN, a flexible and efficient resource management framework is needed to allocate heterogeneous network resources (such as computing and communication resources) of different network segments (i.e. satellite network segment and ground network segment); network function virtualization (NFV) and software defined network (SDN) technologies can efficiently allocate resources, realize the deep integration of ground network and satellite network; by using NFV technology, network functions (such as firewall and network address translation) are decoupled from hardware network devices with specific functions and softwareized as virtual network functions (VNFs), which are deployed on general-purpose commercial servers; in order to ensure efficient end-to-end data transmission, SDN technology configures packet forwarding rules on programmable switches through Openflow protocol, decouples the control plane from the data plane, and realizes centralized network control. Based on NFV and SDN technologies, end-to-end network services are provided in the form of service function chains (SFCs), where each service chain is composed of a series of VNFs connected by virtual links in a predefined order; these VNFs can be deployed and executed on heterogeneous network nodes (i.e. satellite and ground nodes).

[0004] SGIN is a heterogeneous network, which contains satellite network segment and ground network segment, and they have complementary network characteristics; the satellite network segment can realize global communication coverage in a low-cost manner, but its communication bandwidth and computing resources are limited; the ground network segment has sufficient communication and computing resources, but its coverage is limited. Considering the application scenario of dynamic traffic, in the integrated satellite-ground network, how to dynamically select the SFC mapping strategy, use the complementary characteristics of different network segments, balance the network resource and VNF migration cost and service benefit, and maximize the long-term benefit of the network is an important problem.

[0005] At present, some work has been done on the SFC mapping problem in the SGIN scenario, and under the stable network topology and service traffic, efficient SFC mapping can be realized; however, when the network state changes (for example, link connection or network traffic load changes), the SFC mapping decision needs to be dynamically determined, in addition, the VNF migration cost caused by the movement of the satellite and the change of the SFC mapping strategy is also considered. SUMMARY

[0006] The present application aims to at least solve one of the above technical problems in the art. To this end, one object of the present application is to propose a dynamic SFC mapping method for an integrated satellite-ground network, which can determine a set of SFC mapping strategies according to the changing network state, thereby maximizing the long-term network benefit.

[0007] A second object of the present application is to propose a dynamic SFC mapping device for an integrated satellite-ground network.

[0008] To achieve the above object, the first aspect of the present application provides a dynamic SFC mapping method for a satellite-ground integrated network, comprising the following steps: constructing a satellite-ground integrated network, and searching all SFC mapping strategies of each service flow in the satellite-ground integrated network by using a breadth-first search method, selecting one SFC mapping strategy of each service flow for combination, and constructing a SFC mapping strategy set of multiple service flows; obtaining historical network traffic load, and classifying the network traffic load by using a clustering algorithm to obtain multiple categories of network traffic load; searching the optimal SFC mapping strategy set corresponding to each category of network traffic load according to a resource sharing rate value and network estimated income by using a greedy algorithm to obtain a final SFC mapping strategy set; modeling a dynamic SFC mapping problem in a satellite-ground integrated network scenario as a Markov decision process, wherein the decision process is divided into multiple time periods, each time period contains multiple time slots, and the network topology of the satellite-ground integrated network remains unchanged in each time slot; a controller obtains the network state of the satellite-ground integrated network at the starting time slot of each time period, wherein the network state includes the estimated network traffic load of the current time period, the network resource utilization rate of the starting time slot of the current time period, the network topology, the VNF deployment situation, and the SFC mapping strategy set adopted in the last time period; the final SFC mapping strategy set searched under different network traffic loads is constructed into an action space, wherein the mapping decision of each SFC mapping strategy set changes with the movement of the satellite to adapt to the dynamic characteristics of the SGIN topology; and the deep Q network is trained so as to select the SFC mapping strategy set in the action space according to the obtained network state of the satellite-ground integrated network at the starting time slot of each time period.

[0009] According to the dynamic SFC mapping method of the satellite-ground integrated network provided in the embodiments of the present application, firstly, the satellite-ground integrated network is constructed, and the breadth-first search is used to search all SFC mapping strategies of each service flow in the satellite-ground integrated network, and one SFC mapping strategy of each service flow is selected for combination to construct an SFC mapping strategy set of multiple service flows; then, the historical network traffic load is obtained, and a clustering algorithm is used to classify the network traffic load to obtain multiple categories of network traffic load; next, a greedy algorithm is used to search the optimal SFC mapping strategy set corresponding to each category of network traffic load according to the resource sharing rate value and the network estimated income to obtain a final SFC mapping strategy set; then, the dynamic SFC mapping problem in the satellite-ground integrated network scenario is modeled as a Markov decision process, wherein the decision process is divided into multiple time periods, each time period contains multiple time slots, and the network topology of the satellite-ground integrated network remains unchanged in each time slot; then, the controller obtains the network state of the satellite-ground integrated network at the starting time slot of each time period, wherein the network state includes the estimated network traffic load of the current time period, the network resource utilization rate of the starting time slot of the current time period, the network topology, the VNF deployment situation and the SFC mapping strategy set adopted in the last time period; the final SFC mapping strategy set searched under different network traffic loads is used to form an action space, wherein the mapping decision of each SFC mapping strategy set changes with the movement of the satellite to adapt to the dynamic characteristics of the SGIN topology; finally, the deep Q network is trained to select the SFC mapping strategy set in the action space according to the obtained network state of the satellite-ground integrated network at the starting time slot of each time period; in this way, the SFC mapping strategy set can be determined according to the changing network state, so as to maximize the long-term network income.

[0010] In addition, the dynamic SFC mapping method of the satellite-ground integrated network provided in the above embodiments of the present application can also have the following additional technical features:

[0011] Optionally, the greedy algorithm is used to search the optimal SFC mapping strategy set corresponding to each category of network traffic load according to the resource sharing rate value and the network estimated income to obtain a final SFC mapping strategy set, including:

[0012] For each type of network traffic load, search the SFC mapping strategy set corresponding to the resource sharing rate value from large to small in turn, and calculate the network estimated revenue corresponding to each SFC mapping strategy set, wherein the resource sharing rate value is the ratio of the total amount of shared computing resources of each service flow to the total amount of computing resources of the network, and the network estimated revenue is equal to the service revenue obtained minus the communication resource cost and computing resource cost of the network consumed, and when the service demand of the user is met, the corresponding service revenue can be obtained; in the search process, the network estimated revenues of the adjacent SFC mapping strategy sets searched in turn are compared; if the network estimated revenue of the SFC mapping strategy set searched currently is greater than the network estimated revenue of the SFC mapping strategy set searched previously, the SFC mapping strategy set searched currently replaces the SFC mapping strategy set searched previously and is added to the action space; if the network estimated revenue of the SFC mapping strategy set searched currently is less than or equal to the network estimated revenue of the SFC mapping strategy set searched previously, or all SFC mapping strategy sets are traversed, the search is stopped, and the final SFC mapping strategy set is obtained.

[0013] Optionally, the network state is represented as:

[0014]

[0015] wherein, represents the predicted packet arrival rate of each service flow in the kth time period, represents that the estimated predicted packet arrival rate of the service flow l in the time period k obeys a Poisson process with parameters , and includes the network topology and VNF deployment situation at the start time slot of the kth time period, the data packet processing rate of the physical node and the transmission rate of the link, w (k-1) represents the SFC mapping strategy set adopted in the last time period.

[0016] Optionally, the network estimated revenue of the time slot t is calculated according to the following formula:

[0017] χ' t = -α1φ 1,t -α2φ 2,t +α4R t

[0018]

[0019]

[0020]

[0021] wherein, φ 1,t represents the network computing resource utilization rate, is a binary variable, which is equal to 1 if at least one VNF is mapped to node v at time slot t, otherwise 0. Otherwise denotes the computing resource used by a VNF instance, denotes the total computing resource in the network; φ 2,t denotes the communication resource utilization, is a binary variable, which is equal to 1 if at least one virtual link is mapped to physical link (v, u) at time slot t, otherwise 0. Otherwise denotes the communication resource used, denotes the total communication resource in the network; the service revenue of the network is denoted as R t , D l,t denotes the end-to-end packet delay of traffic flow l, D is the delay requirement of the service, λ l,t (1-D l,t / D) is the service revenue of traffic flow l at time slot t.

[0022] Optionally, the set of SFC mapping strategies in the action space is denoted as:

[0023]

[0024]

[0025]

[0026]

[0027]

[0028] where w m denotes a set of SFC mapping strategies; m denotes the sequence number of the set of mapping strategies in the action space, denotes the SFC mapping strategy of traffic flow l in w m ; v denotes the VNF mapping strategy of the satellite network segment, which is a binary variable, when denotes that a VNF is mapped to the accessed satellite node, otherwise denotes the VNF mapping strategy of the ground network segment; denotes the virtual link mapping strategy of the satellite network segment; denotes the virtual link mapping strategy of the ground network.

[0029] To achieve the above object, the second aspect embodiment of the present application proposes a dynamic SFC mapping device of a satellite-ground integrated network, comprising: a network construction module, a first acquisition module, a search module, a problem construction module, a second acquisition module and a mapping module, wherein the network construction module is configured to construct the satellite-ground integrated network, search all SFC mapping strategies of each service flow in the satellite-ground integrated network by using the breadth-first search method, select one SFC mapping strategy of each service flow for combination, and construct a SFC mapping strategy set of multiple service flows; the first acquisition module is configured to acquire historical network traffic load, and classify the network traffic load by using a clustering algorithm to obtain multiple categories of network traffic load; the search module is configured to search the optimal SFC mapping strategy set corresponding to each category of network traffic load according to the resource sharing rate value and the network estimated income by using a greedy algorithm, to obtain a final SFC mapping strategy set; the problem construction module is configured to model the dynamic SFC mapping problem in the satellite-ground integrated network scenario as a Markov decision process, wherein the decision process is divided into multiple time periods, each time period contains multiple time slots, and the network topology of the satellite-ground integrated network remains unchanged in each time slot; the second acquisition module is configured to acquire the network state of the satellite-ground integrated network at the starting time slot of each time period, wherein the network state includes the estimated network traffic load of the current time period, the network resource utilization rate of the starting time slot of the current time period, the network topology, the VNF deployment situation and the SFC mapping strategy set adopted in the last time period; the mapping module is configured to constitute an action space with the final SFC mapping strategy set, wherein the mapping decision of each SFC mapping strategy set changes with the movement of the satellite to adapt to the dynamic characteristics of the SGIN topology; the deep Q network is trained so as to select the SFC mapping strategy set in the action space according to the acquired network state of the satellite-ground integrated network at the starting time slot of each time period.

[0030] The dynamic SFC mapping device of the satellite-ground integrated network according to the embodiment of the application describes the relationship between the network state and the SFC mapping strategy set by describing the SFC mapping problem in the SGIN as a Markov Decision Process (MDP), and proposes a dynamic SFC mapping algorithm based on Deep Q-Learning (DQL), dynamically determines the SFC mapping strategy set for multiple service flows with time-varying traffic rates in a heterogeneous time-varying SGIN, and proposes a new parameter, Sharing Ratio (SR), to describe the sharing level of the computing resources of the SFC mapping strategy set. In order to efficiently allocate heterogeneous network resources, the optimal SFC mapping strategy set is selected for different network loads according to the network resource cost, service revenue and SR value to form an action space, and the SFC mapping strategy set is dynamically determined according to the constantly changing network state by training a deep Q network, so as to maximize the long-term network revenue.

[0031] In addition, the dynamic SFC mapping device of the satellite-ground integrated network according to the above-mentioned embodiment of the application can also have the following additional technical features:

[0032] Optionally, the search module is further configured to search, for each type of network traffic load, the SFC mapping strategy set corresponding to the resource sharing ratio value from large to small in sequence, and calculate the network estimated revenue corresponding to each SFC mapping strategy set, wherein the resource sharing ratio value is the ratio of the total amount of shared computing resources of each service flow to the total amount of computing resources of the network, and the network estimated revenue is equal to the obtained service revenue minus the consumed communication resource cost and computing resource cost of the network. When the service demand of the user is met, the corresponding service revenue can be obtained. In the search process, the network estimated revenues of the adjacent searched SFC mapping strategy sets are compared in sequence. If the network estimated revenue of the currently searched SFC mapping strategy set is greater than the network estimated revenue of the previously searched SFC mapping strategy set, the currently searched SFC mapping strategy set replaces the previously searched SFC mapping strategy set and is added to the action space. If the network estimated revenue of the currently searched SFC mapping strategy set is less than or equal to the network estimated revenue of the previously searched SFC mapping strategy set, or all SFC mapping strategy sets are traversed, the search is stopped, and the final SFC mapping strategy set is obtained.

[0033] Optionally, the network state is represented as:

[0034]

[0035] wherein, represents the predicted packet arrival rate of each service flow in the kth time period, denotes the estimated prediction packet arrival rate of traffic flow / at time period k, which is assumed to follow a Poisson process with parameter includes the network topology and VNF deployment at the beginning of the kth time period, the packet processing rate of physical nodes and the transmission rate of links, w (k-1) denotes the set of SFC mapping strategies taken in the last time period.

[0036] Optionally, the network estimated revenue at time slot t is calculated according to the following formula:

[0037] χ' t = -a1φ 1,t - a2φ 2,t + a4R t

[0038]

[0039]

[0040]

[0041] where φ 1,t denotes the network computing resource utilization, is a binary variable, at time slot t, if at least one VNF is mapped to node v, otherwise denotes the computing resource used by VNF instances, denotes the total computing resource in the network; φ 2,t denotes the communication resource utilization, is a binary variable, at time slot t, if at least one virtual link is mapped to physical link (v, u), otherwise denotes the communication resource used, denotes the total communication resource in the network; the service revenue of the network is denoted as R t , D l,t denotes the packet end-to-end delay of traffic flow / , D is the delay requirement of the service, λ l,t (1 - D l,t / D) is the service revenue of traffic flow / at time slot t.

[0042] Optionally, the set of SFC mapping strategies in the action space is denoted as:

[0043]

[0044]

[0045] ​

[0046]

[0047]

[0048] wherein w m represents a set of SFC mapping strategies; m represents the serial number of the mapping strategy set in the action space, represents the SFC mapping strategy of service flow l in w m ; represents the VNF mapping strategy of the satellite network segment, which is a binary variable, when represents that the VNF is mapped to the accessed satellite node, otherwise represents the VNF mapping strategy of the ground network segment; represents the virtual link mapping strategy of the satellite network segment; represents the virtual link mapping strategy of the ground network. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 is a flowchart of a dynamic SFC mapping method for a satellite-ground integrated network according to an embodiment of the present application;

[0050] Figure 2 is a schematic diagram of an architecture of a satellite-ground integrated network according to an embodiment of the present application;

[0051] Figure 3 is a schematic diagram of an architecture of a deep reinforcement learning-based SFC dynamic mapping algorithm according to an embodiment of the present application;

[0052] Figure 4 is a schematic diagram of a scenario of a satellite-ground integrated network according to an embodiment of the present application;

[0053] Figure 5 is a block diagram of a dynamic SFC mapping device for a satellite-ground integrated network according to an embodiment of the present application. DETAILED DESCRIPTION

[0054] Embodiments of the present application are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by reference to the accompanying drawings are exemplary and are intended to explain the present application, and cannot be understood as limiting the present application.

[0055] For a better understanding of the above technical solutions, the exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0056] For a better understanding of the above technical solutions, the exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0057] Figure 1 The flowchart of the dynamic SFC mapping method of the integrated satellite-ground network according to the embodiment of the present application is shown in Figure 1 The dynamic SFC mapping method of the integrated satellite-ground network includes the following steps:

[0058] S101, construct an integrated satellite-ground network, and search for all SFC mapping strategies of each service flow in the integrated satellite-ground network using a breadth-first method, select one SFC mapping strategy of each service flow for combination, and construct an SFC mapping strategy set of multiple service flows.

[0059] As a specific embodiment, the SGIN architecture based on SDN / NFV technology is shown in Figure 2 The satellite network segment is composed of satellites, user terminals and ground stations connected to each other, the user terminals realize bidirectional communication between users and satellites using satellite-to-ground links (SGLs), the ground stations realize bidirectional communication between ground infrastructure and satellites using SGLs, and the satellites communicate bidirectionally through inter-satellite links (ISLs); considering the positions of the satellites in the constellation, two types of ISLs are considered: intra-plane ISLs established by satellites in the same orbital plane and inter-plane ISLs established by satellites in adjacent orbital planes; each satellite establishes four ISLs with its adjacent satellites, including two intra-plane ISLs and two inter-plane ISLs; the ground network segment is established by interconnecting base stations and routers using wired links, wherein the base stations provide network access for local end users, and the routers are responsible for forwarding data packets using wired links; based on NFV / SDN technology, VNFs can be flexibly deployed on satellites, routers and end users, a network operation controller (NOC) can globally monitor network status and service demand, and programmable routing can be performed to improve routing efficiency of service flows. Figure 2 In the above embodiment, there can be multiple paths traversing a series of VNFs in a predetermined order to provide services for users in different locations (such asFigure 2 the blue arrow, the green arrow and the pink arrow in FIG. 6) ; since the communication coverage of the satellite network segment is wide, the satellite can route the traffic flow to any location where the ground station is deployed (such as Figure 2 the blue solid arrow and the green arrow in FIG. 6).

[0060] In addition, for the network model, in the SGIN, a Very-Small-Aperture-Terminal (VSAT) can communicate with the ground station through the satellite network segment; the access satellite of the VSAT is determined according to the longest link duration rule, and the communication path with the minimum hop number is selected for routing. In addition, the NOC is deployed in the ground network segment, and is responsible for routing, activating / deactivating VNF instances and migrating VNFs. The satellite regularly (such as every second) reports the updated resource status (for example, the use of communication links and computing resources on the satellite) and position to the ground station, and the ground station forwards the information to the NOC. In addition, when the satellite-ground link is switched or the NOC changes the SFC mapping strategy, the NOC forwards the new flow table to the satellite through the ground station to implement the new routing scheme. At the same time, the NOC can monitor the resource status of the ground network segment (for example, the use of wired links, BSs and routers), and sends the flow table to the routers and base stations through the Control-Data-Plane Interface (CDPI) to realize the routing of traffic flow between ground nodes. According to Figure 2 , the service establishment and maintenance process is described as follows: 1) a user subscribes to a service to a Business and Operation Support System (BOSS), the BOSS analyzes the service requirements of the user (such as the required VNF type, delay requirement, etc.), and forms an SFC; 2) the SFC is submitted to the NOC through an Application Programming Interface (API), and the NOC deploys the required VNFs to the routers and satellites according to the monitored SGIN resource status, network topology and network function type of the SFC; 3) the NOC calculates the flow table for SFC mapping, and delivers the flow table to the satellites and routers through the CDPI.

[0061] In S102, historical network traffic loads are obtained, and a clustering algorithm is used to classify the network traffic loads to obtain a plurality of categories of network traffic loads.

[0062] In S103, a greedy algorithm is used to search for an optimal SFC mapping strategy set corresponding to each category of network traffic load according to the resource sharing rate value and the network estimated revenue, to obtain a final SFC mapping strategy set.

[0063] It should be noted that the time correlation of the network traffic load is broken by selecting the optimal SFC mapping strategy set for different network traffic loads.

[0064] As an embodiment, a greedy algorithm is adopted to search the optimal SFC mapping strategy set corresponding to each type of network traffic load according to the resource sharing rate value and the network estimated income to obtain the final SFC mapping strategy set, including:

[0065] S1031, for each type of network traffic load, search the SFC mapping strategy set corresponding to the resource sharing rate value from large to small in turn, and calculate the network estimated income corresponding to each SFC mapping strategy set, wherein the resource sharing rate value is the ratio of the total amount of shared computing resources of each service flow to the total amount of computing resources of the network, and the network estimated income is equal to the service income obtained minus the communication resource cost and computing resource cost of the network consumed, and when the service demand of the user is met, the corresponding service income can be obtained;

[0066] S1032, compare the network estimated incomes of the adjacent SFC mapping strategy sets searched in turn in the search process;

[0067] S1033, if the network estimated income of the SFC mapping strategy set searched currently is greater than the network estimated income of the SFC mapping strategy set searched previously, replace the SFC mapping strategy set searched previously with the SFC mapping strategy set searched currently and add it to the action space;

[0068] S1034, if the network estimated income of the SFC mapping strategy set searched currently is less than or equal to the network estimated income of the SFC mapping strategy set searched previously, or all SFC mapping strategy sets are traversed, stop searching and obtain the final SFC mapping strategy set.

[0069] It should be noted that the optimal SFC mapping strategy set is selected for different network traffic loads to constitute the action space, so as to efficiently allocate the heterogeneous network resources.

[0070] S104, describe the dynamic SFC mapping problem in the star-ground integrated network scene as a Markov decision process to depict the relationship between the network state and the SFC mapping strategy, so as to solve the problem constructed subsequently by the deep Q network. Wherein, the decision process is divided into multiple time periods, each time period contains multiple time slots, and the network topology of the star-ground integrated network remains unchanged in each time slot.

[0071] S105, the controller acquires the network state of the integrated satellite-ground network at the beginning of each time slot, wherein the network state includes the estimated network traffic load of the current time period, the network resource utilization at the beginning of the current time period, the network topology, the VNF deployment, and the SFC mapping strategy set taken in the last time period.

[0072] S106, the final SFC mapping strategy set of each type of network traffic load searched is constituted as an action space, wherein the mapping decision of each SFC mapping strategy set changes with the movement of the satellite to adapt to the dynamic characteristics of the SGIN topology.

[0073] S107, the deep Q network is trained to select the SFC mapping strategy set in the action space according to the acquired network state of the integrated satellite-ground network at the beginning of each time period.

[0074] That is, the controller acquires the network state of the integrated satellite-ground network at the beginning of each time period so that the deep Q network dynamically determines the SFC mapping strategy set according to the constantly changing network state, wherein when the user switches to access the satellite within a time period, the controller retransmits the SFC mapping strategy set taken in the current time period to the satellite through which the service flows.

[0075] As a specific embodiment, the time is divided into T time slots, each time slot has a fixed and same length, denoted as: 1, 2, L, T; the connection between nodes remains unchanged at time slot t, and the network topology is static; the SGIN at time slot t is modeled as an undirected graph where V t represents the set of physical nodes in the satellite-ground network, E t represents the set of physical links in the satellite-ground network; where V G is the set of ground nodes of the SGIN, including base stations, routers, VSATs and ground stations, is the set of satellites covering the ground network segment at time slot t; due to the mobility of the satellite, the satellites covering the ground network segment will change over time; where E G is the set of wired links connecting the ground nodes, is the set of SGLs at time slot t, is the set of ISLs at time slot t.

[0076] For the service model, the VNF required by the service is represented as: where f i (i = 1, 2, L, n) represents the VNF of function type i. The binary variable represents the VNF fi Whether it is deployed on node v is indicated as follows:

[0077]

[0078] To simplify resource management, a node can only support one type of VNF, represented as:

[0079]

[0080] Here, we consider that the services provided require four types of VNFs, namely VNF1, VNF2, VNF3, and VNF4. VNF1 is deployed at the user end, VNF2 is deployed on all satellites as a lightweight basic network function, and VNF3 and VNF can be deployed on the ground network segment according to the node's computing resources, link bandwidth resources, network topology, and network function type.

[0081] Data packets from users accessing the same base station or VSAT form a service flow; assuming there are L service flows in SGIN, represented as follows: Business Flow Through the service function chain π l Access services, service function chain π l Composed of a series of VNFs, in, Represents the service function chain π l The set of VNFs, where j represents the VNF in π. l In the sequence number, i represents the function type of the VNF. For the service function chain π l The abstraction of the physical path between two consecutive VNFs is called a virtual link. To VNF The virtual link is represented as

[0082] The end-to-end latency requirement for the service is D (milliseconds), and the average packet size is σ (bits); the packet arrival process of service flow l in time slot t is modeled with parameter λ. l,t The Poisson process indicates that the arrival process of data packets from different service flows is independent and identically distributed.

[0083] For the computation and communication model, deployed on node v∈V t of The data packet processing rate (packets / second) is expressed as... The packet processing delay is expressed as:

[0084]

[0085] Physical link (v,u)∈E tThe transmission rate (bits per second) is expressed as B. (v,u) The packet transmission delay is expressed as:

[0086]

[0087] To fulfill a service request, the service flow starting from the user terminal needs to traverse the SFCπ sequentially. l The VNF in the process ultimately reaches the destination node, including VNF mapping and virtual link mapping. Binary variables. This indicates that in time slot t, Whether it is mapped to node v is indicated as:

[0088]

[0089] It can only be mapped to the node where the corresponding VNF instance is installed, as shown below:

[0090]

[0091] VNF It can only be mapped to one physical node, that is...

[0092]

[0093] binary variables This indicates that in time slot t, the virtual link Whether it is mapped to a physical link (v,u) is expressed as:

[0094]

[0095] For the cost model, network resource costs include communication resource costs and computing resource costs. It is a binary variable that, in time slot t, if at least one VNF maps to node v, then otherwise binary variables Indicates whether a VNF instance on node v is in use, expressed as:

[0096]

[0097] in, For indicator functions, it is represented as:

[0098]

[0099] The cost of computing resources is defined as the utilization rate of computing resources in the network, i.e.

[0100]

[0101] where, denotes the computing resource used by VNF instances, denotes the total computing resource in the network. is a binary variable, at time slot t, if there is at least one virtual link mapped to physical link (v, u), otherwise denotes

[0102]

[0103] The communication resource cost is defined as the communication resource utilization in the network, i.e.

[0104]

[0105] where, denotes the communication resource used, denotes the total communication resource in the network.

[0106] The VNF migration cost is considered, for service flow l, by using integer variable z l,t denotes the number of VNFs remapped for service flow l at time slot t, denoted as:

[0107]

[0108] The VNF migration cost is defined as the number of VNFs migrated at time slot t, denoted as:

[0109]

[0110] Consider a SGIN based on SDN / NFV technology; in the SGIN, VNFs can be flexibly deployed on satellites and ground nodes. The SFC mapping problem is to determine a new SFC mapping policy for service flows according to previous SFC mapping policies and current network status. The objective is to maximize the long-term network revenue by balancing network resource cost, VNF migration cost and service revenue, under the conditions of satisfying service request constraints, capacity constraints, delay constraints and flow conservation constraints.

[0111] where, the capacity constraint is:

[0112] The arrival rate of data packets that need to be processed by node v E V t should not exceed the data packet processing rate of the node, denoted as:

[0113]

[0114] The arrival rate of data packets that need to be processed by node v E V tThe arrival rate of data packets of the service flow l cannot exceed the transmittable data packet rate of the link, denoted as:

[0115]

[0116] wherein the delay constraint is:

[0117] The end-to-end delay D of data packets of the service flow l l,t cannot exceed the delay requirement D of the service, denoted as:

[0118]

[0119] The end-to-end delay D of data packets l,t includes processing delay, transmission delay, queuing delay and propagation delay; wherein the queuing delay is related to the arrival rate of data packets and the processing rate of nodes, and the propagation delay is related to the communication distance of SGLs and ISLs.

[0120] wherein the flow conservation constraint is:

[0121] For the service flow l, the flow conservation constraint is given by the following formula. If v = v l,s , i.e. v is the source node, then the number of outgoing service flows is one more than the number of incoming service flows. If v = v l,d , i.e. v is the destination node, then the number of incoming service flows is one more than the number of outgoing service flows. If v ∈ V t \{v l,s ,v d}, then the number of incoming service flows is equal to the number of outgoing service flows.

[0122]

[0123] wherein the optimization problem is:

[0124] The service revenue of the network at time slot t is denoted as R t , and the end-to-end delay is one of the important factors affecting the service experience. Therefore, R t is denoted as:

[0125]

[0126] D is the delay requirement of the service, and λ l,t (1-D l,t / D) is the service revenue of the service flow l at time slot t.

[0127] From the perspective of the operator, the optimization goal is to maximize the service performance. At the same time, the physical resource cost and the VNF migration cost should be minimized. Therefore, the network revenue at time slot t is denoted as:

[0128] χ t= -a1f 1,t - a2f 2,t - a3f 3,t + a4R t (21)

[0129] where a1, a2, a3 and a4 are weights.

[0130] To maximize the long-term network revenue, the dynamic SFC mapping problem in SGIN can be formulated as:

[0131]

[0132] s.t. (6)-(7), (16)-(18), (19) (22)

[0133]

[0134] In the above, the set of SFC mapping strategies for the integrated satellite-ground network is the joint SFC mapping strategy for multiple service flows; in order to save computing resources, VNF instances can be shared among multiple service flows. SR is defined as the ratio of shared computing resources to total computing resources, used to evaluate the level of computing resource sharing of the set of SFC mapping strategies, denoted as:

[0135]

[0136] where H1 is the total amount of computing resources of the network, and H2 is the total amount of computing resources shared by each service flow.

[0137] By sharing VNF instances, computing resources can be saved, but service performance may be reduced. This is because multiple service flows sharing VNF instances will cause an increase in queue delay, and the route of data packets to the shared VNF instance may be longer, thereby increasing transmission delay. Therefore, SR can reflect the trade-off between network resource cost and service revenue.

[0138] A DRL-based dynamic SFC mapping algorithm for the 6G SGIN scenario is proposed, which determines a new set of SFC mapping strategies based on network resources, network traffic load, network topology and previous sets of SFC mapping strategies. The goal is to balance network resource cost, VNF migration cost and service revenue, and maximize long-term network revenue.

[0139] Optimization period The time is divided into K time periods, denoted as: 1, 2, L, K, each time period contains N time slots. The Kth time period is denoted as [N(k-1)+1, N(k-1)+2, L, Nk]. The NOC observes the network state at each time slot and determines a set of SFC mapping policies at the beginning of each time period. The decision of the NOC only depends on the latest state, thus the SFC mapping process can be formulated as an MDP. The key elements of the MDP are as follows:

[0140] State space: represents the predicted packet arrival rate of each traffic flow in the kth time period, represents that the estimated packet arrival rate of traffic flow l in time period k obeys a Poisson process with parameters In the kth time period, the observed network state is denoted as:

[0141]

[0142] where, includes network topology parameters VNF deployment information packet processing rate of physical nodes and transmission rate B of links (v,u) , w (k-1) represents the set of SFC mapping policies taken in the last time period.

[0143] Action space: The action space contains a series of actions, where w m is a set of SFC mapping policies, denoted as:

[0144]

[0145] where m represents the serial number of this set of mapping policies in the action space, represents the SFC mapping policy of traffic flow l in w m , denoted as:

[0146]

[0147] where is the VNF mapping policy of the satellite network segment, which is a binary variable, when represents that VNF 2 is mapped to the accessed satellite, otherwise is the VNF mapping policy of the ground network segment, denoted as:

[0148]

[0149] is the virtual link mapping strategy of satellite network segment, denoted as:

[0150]

[0151] is the virtual link mapping strategy of ground network, denoted as:

[0152]

[0153] Due to the mobility of the satellite, in different time slots of a time period, the mapping strategy of the same service flow is different The specific VNF mapping and virtual link mapping decisions taken may be different. Figure 4 The example of shows the mapping decisions taken by a mapping strategy in different time slots of a time period, in which the user in Wuwei accesses the satellite through VSAT for service access, and needs VNF1, VNF2, VNF3 and VNF4 for data processing. For the same mapping decision, i.e., mapping VNF1 at the user end, mapping VNF2 at the server located at the access satellite, using the minimum hop algorithm to route the service flow to the ground station located in Lanzhou, and mapping VNF3 and VNF4 on the servers deployed in Zunyi and Guiyang respectively. From Figure 4 (a) and (b) know that due to the mobility of the satellite, the access satellite of the user changes, and the specific mapping decision is different. Therefore, in order to make the mapping strategy feasible when the satellite moves, for a service flow l, a mapping strategy consists of N mapping decisions.

[0154] The establishment process of the action space is: (1) using the k-means clustering method, clustering the historical records of network load; (2) selecting an optimal SFC mapping strategy set for each cluster to establish the action space The clustering result of the historical records of network load is denoted as:

[0155]

[0156] wherein, is a cluster, consisting of historical records of network load in different time slots.

[0157] An optimal SFC mapping set needs to be selected for each cluster to construct the action space It should meet the service request constraints, capacity constraints, delay constraints and flow conservation constraints. In addition, the balance between network resource supply cost and service income must also be considered. The estimated income of the SFC mapping strategy set in time slot t can be denoted as:

[0158] χ' t = -α1φ 1,t -α2φ2,t + a4R t (31)

[0159] χ' t The VNF migration cost is not included because the time correlation of network traffic load is broken when selecting the optimal SFC mapping policy set for different network traffic loads.

[0160] The process of searching the optimal SFC mapping policy set under different network traffic loads includes: first, searching all SFC mapping policies for each service flow according to the breadth-first search method. For each network load cluster, search the SFC mapping policy set with the maximum SR value and calculate the network estimated revenue of the SFC mapping policy set. This SFC mapping policy set saves network resource cost, but the service revenue may not be optimal; then, search for a new SFC mapping policy set by reducing the SR value and calculate its network estimated revenue; compare the network estimated revenues of the newly searched SFC mapping policy set and the previously searched SFC mapping policy set, and select the set with the highest network estimated revenue to add to the action space . The search process continues until the network estimated revenue of the newly searched SFC mapping policy set no longer increases compared to the network estimated revenue of the previously searched SFC mapping policy set or all SFC mapping policy sets have been traversed.

[0161] The reward obtained in the time period k is represented as:

[0162]

[0163] The proposed deep reinforcement learning-based dynamic SFC mapping (DDSM) is as follows Figure 3 . The NOC receives the network state s (k) , and the host network and the target network estimate the Q value Q(s (k) ,w; θ) and the target Q value of each SFC mapping policy set in the action space, respectively, where θ and θ' represent the parameters of the host network and the target network, respectively. Then, the NOC selects the SFC mapping policy set through the ε-greedy algorithm to obtain the reward r (k) and the next state s (k+1) . An experience [s (k) ,w (k) ,r (k) ,s (k+1) ] contains the current network state, the SFC mapping set, the network revenue, and the next network state, which is stored in the experience replay pool E. Randomly sampling G samples from the experience replay pool E forms a mini-batch subset is represented as:

[0164]

[0165] The parameter theta of the main network is updated based on a minimum loss function, denoted as:

[0166]

[0167] Wherein, mu (k) is a target value, denoted as:

[0168]

[0169] Wherein, delta2 is a discount factor in [0, 1]. The target network updates its network by copying the parameters of the main network every C steps.

[0170] In summary, according to the dynamic SFC mapping method of the integrated satellite-ground network according to the embodiment of the application, the SFC mapping strategy set can be determined according to the changing network state, so as to balance the network resource cost, VNF migration cost and service performance benefit, and maximize the long-term network benefit. The joint SFC mapping strategy of multiple service flows is denoted as the SFC mapping strategy set, and a new parameter, the sharing rate, is proposed to describe the sharing level of the computing resources of the SFC mapping strategy set. In order to efficiently allocate heterogeneous network resources, according to the network resource cost, service benefit and SR value, the optimal SFC mapping strategy set is searched for different network loads based on the greedy algorithm. In order to depict the relationship between the dynamic network state (including the time-varying service traffic and network topology) and the SFC mapping strategy set, the dynamic SFC mapping in the integrated satellite-ground network scenario is established as a Markov process, wherein the decision process is divided into multiple time periods, each time period contains multiple time slots, and the network topology of the integrated satellite-ground network remains unchanged in each time slot; each service function chain mapping strategy contains a series of service function chain mapping decisions to ensure the feasibility of the service function chain mapping strategy when the satellite moves; then, the controller obtains the network state of the integrated satellite-ground network at the starting time slot of each time period, wherein the network state includes the estimated network traffic load of the current time period, the network resource utilization rate at the starting time slot of the current time period, the network topology, the VNF deployment situation and the SFC mapping strategy set adopted in the last time period; finally, the searched SFC mapping strategy sets under different network loads are constructed into an action space, and the deep Q network is trained, so as to select the corresponding SFC mapping strategy set in the action space according to the obtained network state of the integrated satellite-ground network at the starting time slot of each time period.

[0171] In order to realize the above-mentioned embodiments, the embodiment of the application further proposes a dynamic SFC mapping device of an integrated satellite-ground network, such as Figure 5As shown, the dynamic SFC mapping device of the satellite-ground integrated network includes a network construction module 10, a first acquisition module 20, a search module 30, a problem construction module 40, a second acquisition module 50, and a mapping module 60.

[0172] The network construction module 10 is configured to construct the satellite-ground integrated network and search all SFC mapping strategies of each service flow in the satellite-ground integrated network by using a breadth-first search method to construct a set of SFC mapping strategies for multiple service flows. The first acquisition module 20 is configured to acquire historical network traffic loads and classify the network traffic loads by using a clustering algorithm to obtain multiple categories of network traffic loads. The search module 30 is configured to search for an optimal set of SFC mapping strategies corresponding to each category of network traffic load according to a resource sharing rate value and a network estimated income by using a greedy algorithm to obtain a final set of SFC mapping strategies. The problem construction module 40 is configured to model a dynamic SFC mapping problem in a satellite-ground integrated network scenario as a Markov decision process, wherein the decision process is divided into multiple time periods, each time period includes multiple time slots, and the network topology of the satellite-ground integrated network remains unchanged in each time slot. The second acquisition module 50 is configured to acquire a network state of the satellite-ground integrated network at the starting time slot of each time period, wherein the network state includes an estimated network traffic load of the current time period, a network resource utilization rate of the starting time slot of the current time period, a network topology, a VNF deployment situation, and a set of SFC mapping strategies adopted in the previous time period. The mapping module 60 is configured to constitute an action space with the final set of SFC mapping strategies, wherein the mapping decision of each set of SFC mapping strategies changes with the movement of the satellite to adapt to the dynamic characteristics of the SGIN topology. A deep Q network is trained to select a set of SFC mapping strategies in the action space according to the acquired network state of the satellite-ground integrated network at the starting time slot of each time period.

[0173] As an embodiment, the search module 30 is further configured to search, for each type of network traffic load, a set of SFC mapping strategies corresponding to resource sharing rate values from large to small in sequence, and calculate a network estimated revenue corresponding to each set of SFC mapping strategies, wherein the resource sharing rate value is a ratio of a total amount of computing resources shared by each service flow and a total amount of computing resources of the network, and the network estimated revenue is equal to a service revenue obtained minus a communication resource cost and a computing resource cost of the network consumed, and the corresponding service revenue can be obtained when the service demand of the user is met; in the search process, the network estimated revenues of adjacent sets of SFC mapping strategies searched in sequence are compared; if the network estimated revenue of the current set of SFC mapping strategies searched is greater than the network estimated revenue of the previously searched set of SFC mapping strategies, the current set of SFC mapping strategies searched is used to replace the previously searched set of SFC mapping strategies and added to the action space; if the network estimated revenue of the current set of SFC mapping strategies searched is less than or equal to the network estimated revenue of the previously searched set of SFC mapping strategies, or all sets of SFC mapping strategies are traversed, the search is stopped to obtain a final set of SFC mapping strategies.

[0174] As an embodiment, the network state is represented as:

[0175]

[0176] wherein, represents a predicted packet arrival rate of each service flow in the kth time period, represents that the predicted packet arrival rate of the service flow l in the time period k obeys a Poisson process with parameters , and includes the network topology and VNF deployment at the start of the kth time period, the packet processing rate of the physical node and the transmission rate of the link, w (k-1) represents the set of SFC mapping strategies adopted in the previous time period.

[0177] As an embodiment, the network estimated revenue of the time period t is calculated according to the following formula:

[0178] χ' t = -α1φ 1,t -α2φ 2,t +α4R t

[0179]

[0180]

[0181]

[0182] wherein, φ1,t denotes the network computing resource utilization, is a binary variable that equals to 1 if at least one VNF is mapped to node v at time slot t, and 0 otherwise. Otherwise denotes the computing resource used by a VNF instance, denotes the total computing resource in the network; φ 2,t denotes the communication resource utilization, is a binary variable that equals to 1 if at least one virtual link is mapped to physical link (v, u) at time slot t, and 0 otherwise. Otherwise denotes the communication resource used, denotes the total communication resource in the network; R t , D l,t denotes the end-to-end packet delay of traffic flow l, D is the delay requirement of the service, λ l,t (1-D l,t / D) is the service revenue of traffic flow l at time slot t.

[0183] As an embodiment, the set of SFC mapping strategies in the action space is denoted as:

[0184]

[0185]

[0186]

[0187]

[0188]

[0189] where w m denotes a set of SFC mapping strategies; m denotes the sequence number of this set of mapping strategies in the action space, denotes the SFC mapping strategy of traffic flow l in w m ; v denotes the VNF mapping strategy of the satellite network segment, which is a binary variable that equals to 1 if denotes that a VNF is mapped to the accessed satellite node, and 0 otherwise denotes the VNF mapping strategy of the ground network segment; denotes the virtual link mapping strategy of the satellite network segment; denotes the virtual link mapping strategy of the ground network.

[0190] It should be noted that the above description of Figure 1The example of the dynamic SFC mapping method of the satellite-ground integrated network is also applicable to the dynamic SFC mapping device of the satellite-ground integrated network, and details are not repeated here.

[0191] In summary, according to the dynamic SFC mapping device of the satellite-ground integrated network of the embodiment of the present application, the SFC mapping problem in the SGIN is described as a Markov decision process to depict the relationship between the network state and the SFC mapping strategy, and a dynamic SFC mapping algorithm based on deep Q learning is proposed, in the heterogeneous time-varying SGIN, a set of SFC mapping strategies is dynamically determined for multiple service flows with time-varying traffic rates, a new parameter SR is proposed to describe the sharing level of the computing resources of the set of SFC mapping strategies, in order to efficiently allocate heterogeneous network resources, the optimal set of SFC mapping strategies is selected for different network loads according to the network resource cost, service income and SR value to form an action space, the mapping decision of each set of SFC mapping strategies changes with the movement of the satellite to adapt to the dynamic characteristics of the SGIN topology, through training the deep Q network, the set of SFC mapping strategies is dynamically determined according to the constantly changing network state, so as to maximize the long-term network income.

[0192] Those skilled in the art will appreciate that embodiments of the present application can be provided as methods, systems, or computer program products. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage media, etc.) having computer usable program code embodied in the medium.

[0193] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices produce a device that implements the flowcharts and / or block diagrams. Figure 1 The device that implements the function specified in one flow or multiple flows and / or blocks. Figure 1 The device that implements the function specified in one flow or multiple flows and / or blocks.

[0194] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flow Figure 1 flow or flows and / or blocks Figure 1 function specified in the flow or flows and / or blocks.

[0195] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 flow or flows and / or blocks ​ function specified in the flow or flows and / or blocks.

[0196] It is noted that any references made herein to elements or

[0197] While preferred embodiments of the application have been described, it should be apparent that further modifications and improvements can occur to those skilled in the art once instructed by the foregoing disclosure. Accordingly, it is intended that the spirit and scope of the application be limited only by the appended claims and equivalents thereof.

[0198] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.

[0199] In the description of the present application, it is to be understood that the terms "first", "second", "third" and the like, merely identify features being described but do not imply those features are either the first, second or third in use or order, unless otherwise specifically indicated. Thus, features identified as a "first", "second", "third" or the like can implicitly or explicitly include one or more of the features identified with that particular designation. In the description of the present application, the meaning of "a", "an" and "the" includes plural references unless the context clearly indicates otherwise.

[0200] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0201] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "over," and "on top" of the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0202] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms should not be construed as necessarily referring to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0203] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A dynamic SFC mapping method for a space-ground integrated network, characterized in that, Includes the following steps: Construct a space-ground integrated network, and use a breadth-first search method to search for all SFC mapping strategies for each service flow in the space-ground integrated network. Select one SFC mapping strategy for each service flow and combine them to construct a set of SFC mapping strategies for multiple service flows. Historical network traffic load is obtained, and the network traffic load is classified using a clustering algorithm to obtain multiple types of network traffic load. A greedy algorithm is used to search for the optimal SFC mapping strategy set corresponding to each type of network traffic load based on the resource sharing rate value and the estimated network revenue, so as to obtain the final SFC mapping strategy set. The dynamic SFC mapping problem in the space-ground integrated network scenario is modeled as a Markov decision process, in which the decision process is divided into multiple time periods, each time period contains multiple time slots, and the network topology of the space-ground integrated network remains unchanged in each time slot. The controller acquires the network status of the space-ground integrated network at the beginning of each time slot. The network status includes the estimated network traffic load for the current time slot, the network resource utilization rate of the beginning time slot of the current time slot, the network topology, the VNF deployment status, and the set of SFC mapping strategies adopted in the previous time slot. The action space consists of the final SFC mapping policy set for each type of network traffic load found. The mapping decision of each SFC mapping policy set changes as the satellite moves to adapt to the dynamic characteristics of the SGIN topology. A deep Q-network is trained so that, at the beginning of each time slot, a set of SFC mapping strategies is selected in the action space based on the network state of the satellite-ground integrated network.

2. The dynamic SFC mapping method for a space-ground integrated network as described in claim 1, characterized in that, A greedy algorithm is used to search for the optimal SFC mapping strategy set corresponding to each type of network traffic load based on the resource sharing rate value and the estimated network revenue, to obtain the final SFC mapping strategy set, including: For each type of network traffic load, the SFC mapping strategy set corresponding to the resource sharing rate value in descending order is searched, and the estimated network revenue corresponding to each SFC mapping strategy set is calculated. The resource sharing rate value is the ratio of the total computing resources shared by each service flow to the total computing resources of the network. The estimated network revenue is equal to the service revenue obtained minus the communication resource cost and computing resource cost consumed by the network. When the user's service needs are met, the corresponding service revenue can be obtained. During the search process, the network estimated gains are compared sequentially with the adjacent sets of SFC mapping strategies found in the search. If the network estimated return of the currently searched SFC mapping strategy set is greater than the network estimated return of the previously searched SFC mapping strategy set, then the currently searched SFC mapping strategy set is used to replace the previously searched SFC mapping strategy set and added to the action space. If the network estimated return of the currently searched SFC mapping strategy set is less than or equal to the network estimated return of the previously searched SFC mapping strategy set, or if all SFC mapping strategy sets have been traversed, then the search stops and the final SFC mapping strategy set is obtained.

3. The dynamic SFC mapping method for a space-ground integrated network as described in claim 2, characterized in that, Network status is represented as: in, This represents the predicted packet arrival rate for each service flow in the k-th time period. This indicates that the predicted packet arrival rate of the estimated business flow l in time period k follows the parameter . The Poisson process, This includes the network topology and VNF deployment at the start of the k-th time slot, the packet processing rate of physical nodes, and the transmission rate of links, w (k-1) This represents the set of SFC mapping strategies adopted in the previous time period.

4. The dynamic SFC mapping method for a space-ground integrated network as described in claim 3, characterized in that, Calculate the estimated network revenue for time slot t using the following formula: x' t =-α1φ 1,t -a2f 2,t +α4R t Where, φ 1,t Indicates the utilization rate of network computing resources. It is a binary variable that, in time slot t, if at least one VNF maps to node v, then otherwise This indicates the computing resources used by the VNF instance. φ represents the total computing resources in the network. 2,t Indicates the utilization rate of communication resources. It is a binary variable. In time slot t, if at least one virtual link maps to the physical link (v,u), then... otherwise Indicates the communication resources used. R represents the total communication resources in the network; the service revenue of the network is represented by R. t D l,t Let λ represent the end-to-end latency of data packets in service flow l, where D is the latency requirement of the service, and λ is the latency of the data packets. l,t (1-D l,t / D) represents the service revenue of business flow l in time slot t.

5. The dynamic SFC mapping method for a space-ground integrated network as described in claim 4, characterized in that, The set of SFC mapping strategies in the action space is represented as follows: Among them, w m This represents a set of SFC mapping strategies; m represents the sequence number of this set of mapping strategies in the action space. Indicates in w m SFC mapping strategy for business flow l; This represents the VNF mapping strategy for a satellite network segment; it is a binary variable. This indicates that the VNF is mapped to the access satellite node; otherwise... This indicates the VNF mapping strategy for the terrestrial network segment; This indicates the virtual link mapping strategy for satellite network segments; This represents the virtual link mapping strategy for terrestrial networks.

6. A dynamic SFC mapping device for a space-ground integrated network, characterized in that, include: The network construction module is used to build an integrated space-ground network. It uses a breadth-first search method to search for all SFC mapping strategies for each service flow in the integrated space-ground network, selects one SFC mapping strategy for each service flow and combines them to build a set of SFC mapping strategies for multiple service flows. The first acquisition module is used to acquire historical network traffic load and classify the network traffic load using a clustering algorithm to obtain multiple types of network traffic load. The search module is used to search for the optimal SFC mapping strategy set corresponding to each type of network traffic load using a greedy algorithm based on the resource sharing rate value and the estimated network revenue, so as to obtain the final SFC mapping strategy set. The problem construction module is used to model the dynamic SFC mapping problem in the space-ground integrated network scenario as a Markov decision process. The decision process is divided into multiple time periods, each containing multiple time slots. The network topology of the space-ground integrated network remains unchanged in each time slot. The second acquisition module is used to acquire the network status of the space-ground integrated network at the beginning of each time slot. The network status includes the estimated network traffic load for the current time slot, the network resource utilization rate of the beginning of the current time slot, the network topology, the VNF deployment status, and the set of SFC mapping strategies adopted in the previous time slot. The mapping module is used to construct the action space from the final SFC mapping strategy set for each type of network traffic load found. The mapping decision of each SFC mapping strategy set changes as the satellite moves to adapt to the dynamic characteristics of the SGIN topology. The deep Q network is trained so that at the beginning of each time slot, the SFC mapping strategy set is selected in the action space according to the network state of the space-ground integrated network.

7. The dynamic SFC mapping device for a space-ground integrated network as described in claim 6, characterized in that, The search module is also used for, For each type of network traffic load, the SFC mapping strategy set corresponding to the resource sharing rate value in descending order is searched, and the estimated network revenue corresponding to each SFC mapping strategy set is calculated. The resource sharing rate value is the ratio of the total computing resources shared by each service flow to the total computing resources of the network. The estimated network revenue is equal to the service revenue obtained minus the communication resource cost and computing resource cost consumed by the network. When the user's service needs are met, the corresponding service revenue can be obtained. During the search process, the network estimated gains are compared sequentially with the adjacent sets of SFC mapping strategies found in the search. If the network estimated benefit of the currently searched SFC mapping strategy set is greater than the network estimated benefit of the previously searched SFC mapping strategy set, then the currently searched SFC mapping strategy set is added to the action space instead of the previously searched SFC mapping strategy set. If the network estimated return of the currently searched SFC mapping strategy set is less than or equal to the network estimated return of the previously searched SFC mapping strategy set, or if all SFC mapping strategy sets have been traversed, then stop the search and obtain the final SFC mapping strategy set.

8. The dynamic SFC mapping device for a space-ground integrated network as described in claim 7, characterized in that, Network status is represented as: in, This represents the predicted packet arrival rate for each service flow in the k-th time period. This indicates that the predicted packet arrival rate of the estimated business flow l in time period k follows the parameter . The Poisson process, This includes the network topology and VNF deployment at the start of the k-th time slot, the packet processing rate of physical nodes, and the transmission rate of links, w (k-1) This represents the set of SFC mapping strategies adopted in the previous time period.

9. The dynamic SFC mapping device for a space-ground integrated network as described in claim 8, characterized in that, Calculate the estimated network revenue for time slot t using the following formula: x' t =-α1φ 1,t -a2f 2,t +α4R t Where, φ 1,t Indicates the utilization rate of network computing resources. It is a binary variable that, in time slot t, if at least one VNF maps to node v, then otherwise This indicates the computing resources used by the VNF instance. φ represents the total computing resources in the network. 2,t Indicates the utilization rate of communication resources. It is a binary variable. In time slot t, if at least one virtual link maps to the physical link (v,u), then... otherwise Indicates the communication resources used. R represents the total communication resources in the network; the service revenue of the network is represented by R. t D l,t Let λ represent the end-to-end latency of data packets in service flow l, where D is the latency requirement of the service, and λ is the latency of the data packets. l,t (1-D l,t / D) represents the service revenue of business flow l in time slot t.

10. The dynamic SFC mapping device for a space-ground integrated network as described in claim 9, characterized in that, The set of SFC mapping strategies in the action space is represented as follows: Among them, w m This represents a set of SFC mapping strategies; m represents the sequence number of this set of mapping strategies in the action space. Indicates in w m SFC mapping strategy for business flow l; This represents the VNF mapping strategy for a satellite network segment; it is a binary variable. This indicates that the VNF is mapped to the access satellite node; otherwise... This indicates the VNF mapping strategy for the terrestrial network segment; This indicates the virtual link mapping strategy for satellite network segments; This represents the virtual link mapping strategy for terrestrial networks.

Images