A method for deploying service function chains in an integrated air-space-ground scenario

Abstract

The present invention discloses a service function chain deployment method in an air-ground-ground integration scenario. The present invention first divides the traffic into elephant flow and mouse flow according to the size of the traffic data rate, and constructs a corresponding auxiliary logic function diagram according to different flow types. Then, in order to avoid the situation where the resource constraints cannot be met due to repeated selection of the same physical node, the A‑Star algorithm is run on the graph of the elephant flow to obtain a deployment decision, and the Dijkstra algorithm is run on the graph of the mouse flow to obtain a deployment decision. Finally, if the deployment decision fails, the penalty value of the node in the graph is adjusted according to the feedback reason for the failure, and the A‑Star algorithm or the Dijkstra algorithm is re-run to obtain a new deployment decision. The present invention improves the deployment success rate of the service function chain and reduces the end-to-end delay.

Description

Technical Field

[0001] The present invention relates to the field of service function chains, and specifically to a service function chain deployment method in an air-ground-space integration scenario. Background Art

[0002] The advent of the 5G era has promoted the development of new technologies. Among them, network slicing is one of the key technologies of 5G. Due to the three major design capabilities of 5G, namely ultra-high speed, extremely low latency and massive connections, three major application scenarios have emerged: enhanced mobile broadband, ultra-reliable low-latency communication and massive machine communication. Different application scenarios have very different requirements for service capabilities. Therefore, 5G networks need to provide differentiated services for different application scenarios through network slicing technology. 5G end-to-end network slicing refers to the flexible allocation of network resources, on-demand networking, and virtualization of multiple logical subnets with different characteristics and isolation from each other based on the 5G network. Each end-to-end network slice is composed of wireless network, transmission network and core network sub-slices, and is uniformly managed through the end-to-end slice management system. In order to realize network slicing, network function virtualization technology needs to be applied.

[0003] Through network function virtualization technology, network operators can separate network functions from physical hardware. Physical hardware has computing, storage and other resources, which can be abstracted into virtual resource pools for operators to use and allocate. By running virtual machines that can perform specific network functions on general physical hardware, operators can allocate resources more flexibly and reduce hardware expenses and operating costs.

[0004] The Service Function Chain (SFC) represents a network service request as a set of ordered connected virtual network functions (VNFs), such as firewall-network address translation-traffic monitoring. A data packet completes a network service request by traversing a set of network functions in an ordered manner.

[0005] With the advent of the 5G era and the rapid development of the communications field in the 6G era, relying solely on ground networks can no longer meet the growing communication needs. Today, industry experts have proposed a communication scenario for an integrated air-ground-air network. Among them, the air-based network consists of high-altitude communication platforms, unmanned aerial vehicle self-organizing networks, etc., the space-based network consists of various satellite systems to form a space-based backbone network and a space-based access network, and the ground-based network is mainly composed of ground Internet and mobile communication networks. Based on the ground-based network, the space-based network and the air-based network serve as supplements and extensions, providing a ubiquitous, intelligent, collaborative, and efficient information security infrastructure for various network applications in a wide area.

[0006] At present, some researchers have proposed SFC deployment schemes in different scenarios, and focus on different optimization goals, such as minimizing deployment overhead, minimizing energy consumption, maximizing operator benefits, etc. For example, genetic algorithms and ant colony algorithms are used to generate deployment schemes. However, most studies only consider the deployment of SFC in ground-based networks, and the optimization goal is single, which cannot effectively improve multiple performance indicators. Summary of the invention

[0007] In view of the above-mentioned deficiencies in the prior art, the present invention provides a service function chain deployment method in an air-ground integrated scenario, which solves the problems of scenario inapplicability and single optimization target in the service function chain deployment problem.

[0008] To achieve the above object, the present invention adopts the following technical solution:

[0009] A service function chain deployment method in an air-ground integration scenario comprises the following steps:

[0010] 1) Read the network status and the service request arriving at the current moment, initialize the satellite weight λ, and determine the type of the traffic as elephant flow or mouse flow according to the data rate requirement of the service request;

[0011] 2) Based on the traffic type determined in step 1), construct an auxiliary logic function diagram corresponding to the request;

[0012] 3) If the request type is elephant flow, set the function to calculate the h value;

[0013] 4) If the request type is elephant flow, execute the A-Star algorithm to obtain the shortest path from the source node to the destination node on the auxiliary logic function graph; if the request type is mouse flow, execute the Dijkstra algorithm to obtain the shortest path from the source node to the destination node on the auxiliary logic function graph;

[0014] 5) Deploy in the physical network according to the deployment decision obtained by the shortest path. If all constraints are met, the deployment is successful and proceed to step 6), otherwise proceed to step 7);

[0015] 6) Allocate the node resources and bandwidth resources required for the request and update the network status;

[0016] 7) According to the unsatisfied constraints, update the penalty value of the node in the auxiliary logic function diagram and the weight of the satellite. If the satellite weight is greater than 0, return to step 2), otherwise reject the service request.

[0017] Furthermore, in step 2), the specific process of constructing the auxiliary logic function diagram is as follows:

[0018] 1) Find a candidate node set that can provide the required resources for each VNF in the SFC; the candidate node set of VNF i constitutes the i+1th column of nodes in the auxiliary logic function graph; in the constructed graph, there is only one node in the first column, which is the source node of the SFC, and there is only one node in the last column, which is the destination node of the SFC;

[0019] 2) Except for the last column of nodes, the nodes in each column are connected to the nodes in the next column through directed edges. The nodes in the current column are the arc tails, and the nodes in the next column are the arc heads. Each connection has a weight, the weight cost n′m′ The calculation is as follows:

[0020]

[0021] Where m′ and n′ represent nodes in the auxiliary logic function diagram, and n represents a node in the physical topology. and They represent the upper limit of CPU resources and RAM resources of node n respectively. and They represent the upper limit of CPU resources and the upper limit of RAM resources of the physical node mapped by node m′, respectively. and They represent the ratio of the available CPU resources to the upper limit of the CPU resources and the ratio of the available RAM resources to the upper limit of the RAM resources of the physical node mapped by the node m′, respectively. and They represent the CPU resources and RAM resources required for initialization of the instance corresponding to VNFi. and They represent the CPU resources and RAM resources required by VNF i in an SFC to use the instance, respectively. n′m′ represents the shortest link delay between two nodes obtained by the Dijkstra algorithm; if the current column is the candidate node column of the last VNF ​​of the SFC, the connection weight between the node in the current column and the destination node only considers the link delay, that is,

[0022] 3) In addition to calculating the weight of the connection, the penalty value is also calculated for the node. The penalty value is initially set to 0. When it is found that the nodes n′ and m′ at both ends of the connection map to the same physical node, that is, the two adjacent VNFs have the same candidate node, the penalty value of the node m′ with the later column number is updated:

[0023]

[0024] Among them, i represents the current column number, j represents the previous column number, and k represents the adjustable coefficient; if the request is an elephant flow, the auxiliary logic function diagram is constructed; if the request is a mouse flow, the weight of each connection is added to the penalty value of the node at the end of the connection, that is, the connection weight calculation formula of the auxiliary logic function diagram of the mouse flow is:

[0025]

[0026] 4) If a candidate node is a satellite node, the weights of all links connected to the node must be multiplied by a satellite weight λ.

[0027] Furthermore, in step 3), the specific calculation process of the h value is:

[0028] Get all nodes in the closed list that have been added to the shortest path. If the physical node mapped by the currently calculated node n' is already included in the closed list, then add the node penalty value of the node in the closed list to the h value; then determine whether the node m' at the other end of the line starting from n' is mapped to the same physical node as n'. If the nodes m' and n' at the other end of all lines starting from n' are not mapped to the same physical node, the h value calculation is completed; conversely, if there is a m' and n' mapping to the same physical node, then calculate the proportion of the weight of the line in the sum of the weights of all lines starting from n':

[0029]

[0030] Among them, Ω i represents the candidate node set corresponding to VNF i; finally, the h value of node n′ in the current i-th column The calculation is as follows:

[0031]

[0032] Further, in step 7), according to the unsatisfied constraints, the method for updating the penalty values ​​of the nodes and the weights of the satellites in the auxiliary logic function diagram is as follows:

[0033] (1) If the deployment fails due to insufficient node resources, the penalty value of the node that is used most times in the current deployment decision is added to its usage times, and the satellite weight λ is subtracted from δ1;

[0034] (2) If the deployment fails due to insufficient link bandwidth, a node is randomly selected in the current deployment decision, the penalty value of the node is added to its usage times, and the satellite weight λ is subtracted from δ1;

[0035] (3) If the deployment fails due to exceeding the specified delay, the satellite weight λ is reduced by δ2;

[0036] Among them, 0<δ1<δ2<λ.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] Widely applicable scenarios. Traditional service function chain deployment methods are mostly designed for pure ground networks, such as data center networks, edge networks, etc. The present invention expands the scenario to an integrated air-ground network, and is not limited by the topology scale. It also takes into account the sharing of virtual network functions and the repeatable deployment of a single node.

[0039] High deployment success rate. The present invention fully considers all potential deployable nodes and modifies the deployment plan through a reasonable and effective failure feedback mechanism, so that a decision that can be successfully deployed can be obtained through a relatively small number of modifications.

[0040] The present invention reasonably uses fewer candidate nodes to complete the deployment of a service function chain, thereby reducing the propagation delay of the link and improving the service quality requirements of users. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of an application scenario of the present invention;

[0042] Figure 2 It is the overall flow chart of the algorithm of the present invention;

[0043] Figure 3 A schematic diagram for constructing an auxiliary logic function diagram for the present invention;

[0044] Figure 4 is the successful acceptance rate of the service function chain in the simulation scenario;

[0045] Figure 5 is the average delay of service function chains of different lengths in the simulation scenario. DETAILED DESCRIPTION

[0046] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0047] See also Figure 1 , a service function chain deployment method in an air-ground integration scenario, comprising the following steps:

[0048] 1) Read the network status and the service request that arrives at the current moment. Initialize the satellite weight λ and determine the type of traffic as elephant flow or mouse flow based on the data rate requirement of the service request.

[0049] Get the current network topology G = (N, L). Since the air-ground integrated scenario includes ground nodes and satellite nodes, the set of ground nodes is represented as N in this scenario. G , the set of satellite nodes is represented as N S , the set of ground links is denoted as L G , the set of intersatellite links is denoted as L S , the set of satellite-to-ground links is denoted as L SG , therefore, N = N G ∪N S , L=L G ∪L S ∪L SG Due to the mobility of satellites, the network topology will change at different times, but in a short time interval, the topology remains unchanged. The time it takes to calculate the deployment decision and allocate resources for an SFC is very short, so when deploying an SFC, the network topology is static.

[0050] Get the service request q arriving at the current time, which can be expressed as Among them, s q and d q Respectively represent the source node and destination node of the request, b q Indicates the data rate requirement, D q Indicates the maximum tolerable delay, represents the ordered VNF requests, Δt q Indicates the duration of the service function chain.

[0051] According to the data rate threshold θ set by the operator, the traffic type of the currently received service request is determined. If the data rate of the request is b q If it is greater than θ, it is considered to be an elephant flow; otherwise, it is considered to be a mouse flow.

[0052] 2) Based on the traffic type determined in step 1), an auxiliary logic function diagram corresponding to the request is constructed.

[0053] See also Figure 2 , the construction process of the auxiliary logic function diagram is:

[0054] 2.1) Find a set of candidate nodes that can provide the required resources for each VNF in the SFC. The set of candidate nodes of VNF i constitutes the i+1th column of nodes in the auxiliary logic function graph. In the constructed graph, there is only one node in the first column, which is the source node of the SFC, and there is only one node in the last column, which is the destination node of the SFC. Therefore, the graph contains a total of Column node. Indicates the number of VNFs included in request q.

[0055] 2.2) Except for the last column of nodes, the nodes in each column are connected to the nodes in the next column through directed edges. The nodes in the current column are the arc tails, and the nodes in the next column are the arc heads. Each connection has a weight, the weight cost n′m′ The calculation is as follows:

[0056]

[0057] Where m′ and n′ represent nodes in the auxiliary logic function diagram, and n represents a node in the physical topology. and They represent the upper limit of CPU resources and RAM resources of node n respectively. and They represent the upper limit of CPU resources and the upper limit of RAM resources of the physical node mapped by node m′, respectively. and They represent the ratio of the available CPU resources to the upper limit of the CPU resources and the ratio of the available RAM resources to the upper limit of the RAM resources of the physical node mapped by the node m′, respectively. and They represent the CPU resources and RAM resources required for initialization of the instance corresponding to VNFi. and They represent the CPU resources and RAM resources required by VNF i in an SFC to use the instance, respectively. n′m′ represents the shortest link delay between two nodes obtained by Dijkstra algorithm. If m′ and n′ map to the same physical node, the delay overhead is 0. When , the connection weight between the node in the current column and the destination node only considers the link delay, that is,

[0058] 2.3) In addition to calculating the weight of the connection, the penalty value is also calculated for the node, and the penalty value is initially set to 0. When it is found that the nodes n′ and m′ at both ends of the connection map to the same physical node, that is, the two adjacent VNFs have the same candidate node, the penalty value of the node m′ with the later column number is updated:

[0059]

[0060] Where i represents the current column number, j represents the previous column number, and k represents the adjustable coefficient. If q is an elephant flow, the auxiliary logic function diagram is constructed. If q is a mouse flow, the weight of each connection is added to the penalty value of the node at the end of the connection, that is, the connection weight calculation formula of the auxiliary logic function diagram of the mouse flow is:

[0061]

[0062] 2.4) If a candidate node is a satellite node, the weights of all links connected to the node must be multiplied by a satellite weight λ.

[0063] 3) If the request is for an elephant flow, set the function to calculate the h value.

[0064] Get all nodes in the closed list that have been added to the shortest path. If the physical node mapped by the currently calculated node n' is already included in the closed list, then add the node penalty value of the node in the closed list to the h value. After that, determine whether the node m' at the other end of the line starting from n' is mapped to the same physical node as n'. If the nodes m' and n' at the other end of all lines starting from n' are not mapped to the same physical node, the h value calculation is complete. On the contrary, if there is a situation where m' and n' map to the same physical node, then calculate the proportion of the weight of the line in the sum of the weights of all lines starting from n':

[0065]

[0066] Among them, Ω i represents the candidate node set corresponding to VNF i. Finally, the h value of node n′ in the current i-th column is The calculation is as follows:

[0067]

[0068] 4) If the request type is elephant flow, execute the A-Star algorithm to obtain the shortest path from the source node to the destination node on the auxiliary logic function graph; if the request type is mouse flow, execute the Dijkstra algorithm to obtain the shortest path from the source node to the destination node on the auxiliary logic function graph.

[0069] The ordered nodes included in the shortest path except the source node and the destination node are the deployment nodes corresponding to each VNF obtained.

[0070] 5) Deploy in the physical network according to the deployment decision obtained by the shortest path. If all constraints are met, the deployment is successful and proceed to step 6), otherwise proceed to step 7).

[0071] 6) Allocate the node resources and bandwidth resources required for the request and update the network status.

[0072] 7) Update the penalty value of the node and the weight of the satellite in the auxiliary logic function diagram according to the unsatisfied constraints. If the satellite weight λ is greater than 0, return to step 2), otherwise reject the service request.

[0073] Choose different update methods for different deployment failure reasons:

[0074] (1) If deployment fails due to insufficient node resources, the penalty value of the node that is used most times in the current deployment decision is added to its usage times, and the satellite weight λ is subtracted from δ1.

[0075] (2) If the deployment fails due to insufficient link bandwidth, a node is randomly selected in the current deployment decision, the penalty value of the node is added to its usage times, and the satellite weight λ is subtracted from δ1.

[0076] (3) If the deployment fails due to exceeding the specified delay, the satellite weight λ is reduced by δ2.

[0077] Among them, 0<δ1<δ2<λ.

[0078] The parameter settings of the embodiment are shown in Table 1.

[0079] Table 1

[0080]

[0081] The experiment sets the simulation time to 8000 seconds. During the simulation time, the satellite topology will change. Figure 4 As shown, the successful acceptance rate of the method proposed by the present invention is 1.20% to 8.74% higher than that of the genetic algorithm.

[0082] The SFC length indicates the number of VNF requests contained in an SFC. Figure 5 As shown, under different service function chain lengths, the deployment decision delay obtained by the method proposed in the present invention is lower than that of the genetic algorithm, and when the number of VNFs included in the service function chain increases, the delay increase of the present invention is significantly lower than the delay increase of the genetic algorithm.

Claims

1. A service function chain deployment method in an air-ground integration scenario, characterized in that: The following steps are involved: 1) Read the network status and the service request arriving at the current moment; initialize the satellite weight λ, and determine the type of service request traffic as elephant flow or mouse flow according to the data rate requirement of the service request; 2) Based on the traffic type determined in step 1), an auxiliary logic function diagram corresponding to the service request is constructed; wherein, in step 2), the specific process of constructing the auxiliary logic function diagram is as follows: 2.1) Find a candidate node set that can provide the required resources for each VNF in the SFC; the candidate node set of VNF i constitutes the nodes in the i+1th column of the auxiliary logic function graph; in the constructed graph, there is only one node in the first column, which is the source node of the SFC, and there is only one node in the last column, which is the destination node of the SFC; 2.2) Except for the last column of nodes, the nodes in each column are connected to the nodes in the next column through directed edges. The nodes in the current column are arc tails, and the nodes in the next column are arc heads. Each connection has a weight, cost n′m′ The calculation is as follows: Where m′ and n′ represent nodes in the auxiliary logic function diagram, and n represents a node in the physical topology. and They represent the upper limit of CPU resources and RAM resources of node n respectively. and They represent the upper limit of CPU resources and the upper limit of RAM resources of the physical node mapped by node m′, respectively. and They represent the ratio of the available CPU resources to the upper limit of the CPU resources and the ratio of the available RAM resources to the upper limit of the RAM resources of the physical node mapped by the node m′, respectively. and They represent the CPU resources and RAM resources required for initialization of the instance corresponding to VNF i, respectively. and They represent the CPU resources and RAM resources required by VNF i in an SFC to use the instance, respectively. n′m′ represents the shortest link delay between two nodes obtained by the Dijkstra algorithm; if the current column is the candidate node column of the last VNF ​​of the SFC, the connection weight between the node in the current column and the destination node only considers the link delay, that is, 2.3) In addition to calculating the weight of the connection, the penalty value is also calculated for the node. The penalty value is initially set to 0. When it is found that the nodes n′ and m′ at both ends of the connection map to the same physical node, that is, the two adjacent VNFs have the same candidate node, the penalty value of the node m′ with the later column number is updated: Among them, i represents the current column number, J represents the previous column number, and k represents the adjustable coefficient; if the request is an elephant flow, the auxiliary logic function diagram is constructed; if the request is a mouse flow, the weight of each connection is added to the penalty value of the node at the end of the connection, that is, the connection weight calculation formula of the auxiliary logic function diagram of the mouse flow is: 2.4) If a candidate node is a satellite node, the weights of all links connected to the node must be multiplied by a satellite weight λ; 3) If the request type is elephant flow, set the function for calculating the h value; in step 3), the specific calculation process of the h value is: Get all nodes in the closed list that have been added to the shortest path. If the physical node mapped by the currently calculated node n' is already included in the closed list, then add the node penalty value of the node in the closed list to the h value; then determine whether the node m' at the other end of the line starting from n' is mapped to the same physical node as n'. If the nodes m' and n' at the other end of all lines starting from n' are not mapped to the same physical node, the h value calculation is completed; conversely, if there is a m' and n' mapping to the same physical node, then calculate the proportion of the weight of the line in the sum of the weights of all lines starting from n': Among them, Ω i represents the candidate node set corresponding to VNF i; finally, the h value of node n′ in the current i-th column The calculation is as follows: 4) If the request type is elephant flow, execute the A-Star algorithm to obtain the shortest path from the source node to the destination node on the auxiliary logic function graph; if the request type is mouse flow, execute the Dijkstra algorithm to obtain the shortest path from the source node to the destination node on the auxiliary logic function graph; 5) Deploy in the physical network according to the deployment decision obtained by the shortest path. If all constraints are met, the deployment is successful and proceed to step 6), otherwise proceed to step 7); 6) Allocate the node resources and bandwidth resources required for the request and update the network status; 7) According to the unsatisfied constraints, update the penalty values ​​of the nodes in the auxiliary logic function graph and the weights of the satellites; if the satellite weight λ is greater than 0, return to step 2), otherwise reject the service request.

2. The method for deploying a service function chain in an air-ground integrated scenario according to claim 1, characterized in that: In step 7), according to the unsatisfied constraints, the method for updating the penalty values ​​of the nodes and the weights of the satellites in the auxiliary logic function diagram is as follows: (1) If the deployment fails due to insufficient node resources, the penalty value of the node that is used most times in the current deployment decision is added to its usage times, and the satellite weight λ is subtracted from δ1; (2) If the deployment fails due to insufficient link bandwidth, a node is randomly selected in the current deployment decision, the penalty value of the node is added to its usage times, and the satellite weight λ is subtracted from δ1; (3) If the deployment fails due to exceeding the specified delay, the satellite weight λ is reduced by δ2; Among them, 0<δ1<δ2<λ.

Images