CN113572686B - Heaven and earth integrated self-adaptive dynamic QoS routing method based on SDN - Google Patents

Heaven and earth integrated self-adaptive dynamic QoS routing method based on SDN Download PDF

Info

Publication number
CN113572686B
CN113572686B CN202110812489.2A CN202110812489A CN113572686B CN 113572686 B CN113572686 B CN 113572686B CN 202110812489 A CN202110812489 A CN 202110812489A CN 113572686 B CN113572686 B CN 113572686B
Authority
CN
China
Prior art keywords
satellite
node
algorithm
nodes
network
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202110812489.2A
Other languages
Chinese (zh)
Other versions
CN113572686A (en
Inventor
杨力
潘成胜
魏德宾
戚耀文
李涵睿
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Dalian University
Original Assignee
Dalian University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Dalian University filed Critical Dalian University
Priority to CN202110812489.2A priority Critical patent/CN113572686B/en
Publication of CN113572686A publication Critical patent/CN113572686A/en
Application granted granted Critical
Publication of CN113572686B publication Critical patent/CN113572686B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00Routing or path finding of packets in data switching networks
    • H04L45/46Cluster building
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04BTRANSMISSION
    • H04B7/00Radio transmission systems, i.e. using radiation field
    • H04B7/14Relay systems
    • H04B7/15Active relay systems
    • H04B7/185Space-based or airborne stations; Stations for satellite systems
    • H04B7/1851Systems using a satellite or space-based relay
    • H04B7/18519Operations control, administration or maintenance
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/12Discovery or management of network topologies
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14Network analysis or design
    • H04L41/145Network analysis or design involving simulating, designing, planning or modelling of a network
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02DCLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00Reducing energy consumption in communication networks
    • Y02D30/70Reducing energy consumption in communication networks in wireless communication networks

Abstract

The invention discloses a space-ground integrated self-adaptive dynamic QoS routing method based on an SDN, which comprises the following steps: establishing a hierarchical clustering network model based on the SDN; establishing network resource mapping; and establishing a multi-constraint QOS self-adaptive routing algorithm SDN-AD. The invention can effectively reduce the control overhead and improve the transmission efficiency by reducing the long-distance transmission and the shortest-distance clustering of the control packet. The invention formulates the multi-constraint QoS problem into an optimization problem taking the minimum transmission cost as a target, can effectively calculate the transmission cost of links of different classes, and shields the difference between satellites and ground networks of different levels. The invention can better adapt to network change, thereby providing services meeting different service quality requirements. The invention solves the optimization problem and realizes the self-adaptive routing. The invention has better performance in controlling overhead, network service quality and algorithm convergence speed.

Description

Heaven and earth integrated self-adaptive dynamic QoS routing method based on SDN
Technical Field
The invention relates to a space-ground integrated network, in particular to a space-ground integrated self-Adaptive Dynamic Routing method (SDN-AD for short) based on a Distributed SDN.
Background
The Space-Ground Integrated Network (SGIN) integrates a satellite with a Ground Network, has The characteristics of wide coverage area, access according to needs, no restriction of geographic conditions on communication and The like, can support diversified communication requirements, and is a new trend of future Network development. With the rapid development of the world-wide integrated network, the requirements of various communication services in the network are greatly changed, and clear quality guarantee requirements are provided in the aspects of bandwidth, time delay, packet loss rate and the like. In order to meet the service requirements, a network with guaranteed quality of coverage determination, connection determination and service quality adaptation needs to be provided.
The satellite network is used as a key part in a heaven-earth integrated network, and the space information resources can be effectively integrated and bear a large amount of deterministic services due to the flexible networking characteristic of the satellite network. But due to the characteristic of high-speed movement of the satellite network, the satellite network has the problems of large space-time scale of the network and limited space resources. In order to reasonably plan Network resources and meet the requirement of service quality guarantee, researchers introduce Software-Defined networking (SDN) into a world-wide integrated Network according to a numerical control separation idea. A logically centralized loosely coupled control plane and a physically distributed data plane are implemented, the centralized management mode of which provides the possibility for the implementation of routing algorithms that take into account a variety of link states. However, a large amount of control signaling generated by the intelligent network system occupies the originally scarce inter-satellite network resources, and is difficult to adapt to the space-ground integrated intelligent network environment.
Therefore, a more reasonable configuration model needs to be designed for the characteristics of the SDN architecture to better implement Quality of service (QoS) routing. Most of the existing SDN architectures only deploy a controller on a geosynchronous Orbit satellite (GEO) or the ground, and no matter the controller is placed on the GEO satellite or the ground, the distance from the controlled node is long, and a signaling sent by the controller needs to be transmitted for a long distance, which may generate a high overhead cost. The method aims to solve the problem that the SDN framework is applied to the large overhead of the control signaling of the heaven-earth integrated network. The method comprises the steps that controllers are respectively deployed in a GEO Satellite, a ground Satellite gateway and a ground data center through a comprehensive Satellite Communication SDN (SERvICE Defined Framework for Integrated Space-Terrestrial Satellite Communication), so that the agility of network services is effectively improved, the calculation burden is dispersed, and the control signaling overhead cannot be effectively reduced.
In order to meet the problem of service quality guarantee under the condition of large space-time scale of a space-ground integrated network, scholars conduct a great deal of Routing strategy research considering link state characteristics, Software Defined Routing Algorithm (SDRA) improves Distributed Routing Algorithm (DRA), congestion control is achieved through inter-satellite link delay weight, and transmission quality between low-orbit satellites is effectively guaranteed. The routing strategy supported by the service quality realizes a plurality of flow maximization paths and acceptable transmission delay. But they all only consider a single quality of service requirement, affecting the overall utilization of the network.
Therefore, a Low Earth Orbit satellite (LEO) with multi-objective decision is provided, and a network routing algorithm can calculate a path meeting the QoS requirement. A virtual topology Delay Tolerant Network (DTN) routing algorithm based on queue scheduling can perform queue scheduling according to different data types, and elastic load balanced routing is achieved. A multi-constraint route optimization algorithm based on a directed acyclic graph introduces a self-adaptive link comprehensive cost function, but does not calculate a threshold value of a constraint condition.
The software defines the satellite network architecture and the intelligent QoS routing scheme, and intelligently provides fine-grained QoS guarantee. The routing scheme of energy perception can carry out routing according to different energy levels of the satellite, and the characteristics of different services are considered, so that load balance is realized. Based on various QoS demand routing algorithms (MQoO) adopting Lagrangian relaxation method under SDN satellite network architecture, the approximate algorithm can obtain relatively accurate calculation results, but cannot guarantee calculation time. The Routing algorithm (SRADR, Status and recommendation Adaptive QoS Dynamic Routing) is suitable for satellite network state and Reputation Adaptive service quality, Routing discovery and maintenance are carried out by utilizing the ant colony algorithm, the self-adaptation of the service quality is realized, and the convergence speed of the algorithm is low.
Disclosure of Invention
In order to solve the problems in the prior art, the invention aims to design a distributed SDN-based heaven-earth integrated adaptive dynamic routing method which can reduce SDN architecture control signaling overhead, reasonably distribute network resources and adaptively guarantee different service requirements of multiple services.
In order to achieve the purpose, the technical scheme of the invention is as follows: a space-ground integrated self-adaptive dynamic routing method based on an SDN comprises the following steps:
A. establishing layered clustering network model based on SDN
According to the characteristics that the coverage area of the GEO satellite is large and the ground is relatively static, the GEO satellite is set as a main controller. And according to the characteristics of high movement speed and small communication time delay of the LEO satellite, the LEO satellite is set as a slave controller.
A1, determining a network model
The SDN-based space-ground integrated network is composed of LEO satellite nodes, GEO satellite nodes and ground nodes. Representing the topological graph of the SDN-based heaven-earth integrated network as an undirected weighted topological graph G (V, E) and storing the undirected weighted topological graph G (V, E) in a GEO satellite main controller, wherein V (V) is { V { (V) } V G ,v L ,v T Represents a finite set of all nodes in the network, v G ,v L ,v T And respectively represent node sets of GEO satellite nodes, LEO satellite nodes and ground nodes. E ═ IOL ═ ISL ≡ UDL } represents the set of all links in the network, IOL represents an interplanetary link, ISL represents an intersatellite link, and UDL represents a satellite-to-ground link.
G for nth GEO satellite n Expressed as, 1. ltoreq. N. ltoreq.N G Let N be in order for the GEO satellite to cover the global area G Not less than 3. Dividing the whole satellite network into N according to the coverage range of the GEO satellite G A region, i.e.
Figure BDA0003168740510000031
L for the h LEO satellite h Indicates that the total number of LEO satellites is M L ×N L ,M L Representing the number of satellite orbital planes, N L Representing the number of satellites in the orbital plane of the satellite.
According to the virtual node idea, the earth surface is divided into different areas according to the longitude and latitude, and a logical address is set for each area. The logical address within each region remains the same, but the LEO satellite membership therein changes. When a LEO satellite member leaves a logical area, the next LEO satellite enters the logical area to replace its position. Coverage clustering for GEO satellites is determined according to the following conditions.
Figure BDA0003168740510000041
Figure BDA0003168740510000042
Figure BDA0003168740510000043
In the formula (I), the compound is shown in the specification,
Figure BDA0003168740510000044
representing LEO satellites L h And GEO satellite G n The half central angle of (c).
Figure BDA0003168740510000045
Represents G n Maximum half-center angle covered for the LEO satellite orbital plane.
Figure BDA0003168740510000046
Is L h The radial distance to the earth's center. G' n Represents G n And mapping nodes in the orbital plane of the LEO satellite. L G' n L h L represents G' n And L h The distance between them.
Figure BDA0003168740510000047
Is G n Radial distance to the earth's center.
Figure BDA0003168740510000048
Representing the minimum elevation angle of the GEO satellite relative to the orbital plane to the LEO satellite.
The LEO satellite satisfying the formula (3) is determined according to | G' n L h And | minimum principle, only marking in the cluster of the GEO satellite. Since the LEO satellite coverage areas correspond one-to-one to the ground areas, LEO satellite members and ground station members within the GEO satellite coverage area are determined. The overall network topology under the SDN-based hierarchical clustering network model can be determined only by establishing a controller cluster configuration strategy to obtain a satellite dynamic connection relation.
A2 configuration controller cluster
X, Y, Z are respectively set as the earth fixed coordinate system of a single LEO satellite, the origin of the earth fixed coordinate system is the geocentric, the XOY plane is coincided with the equatorial plane of the earth, the OX axis is consistent with the Greenwich meridian direction, and the OZ axis is coincided with the polar axis of the earth. LEO satellite L is obtained from equations (4) to (9) without considering the effects of other celestial and satellite vehicles on LEO satellites j And LEO satellite L k Actual distance between
Figure BDA0003168740510000049
Figure BDA00031687405100000410
Ω G =ωΔt (5)
X=H[cosμcos(Ω-Ω G )-sinμsin(Ω-Ω G )cosσ] (6)
Y=H[cosμsin(Ω-Ω G )-sinμcos(Ω-Ω G )cosσ] (7)
Z=H(sinμsinσ) (8)
Figure BDA00031687405100000411
In the formula, mu o Is the argument of the satellite in the near-to-the-earth,
Figure BDA00031687405100000412
the linear velocity is delta t is a minimum time, omega is the rotation angular velocity of the earth, omega is the ascent point right ascension of the satellite, and sigma is the orbit inclination of the satellite.
For each divided region V i Designing a slave controller configuration clustering strategy of the LEO satellite as follows: taking the number a of slave controllers of the LEO satellite as V i The cluster number in the slave controller set is denoted as LC ═ c 1 ,c 2 ,...,c a },c i The administered cluster is expressed as
Figure BDA0003168740510000051
All data are divided into a clusters, and a cluster centers are selected. Calculating the node of each LEO satellite to the clustering center c according to the formula (9) i Is a distance of
Figure BDA0003168740510000052
Will meet
Figure BDA0003168740510000053
Required nodes to be grouped into clusters
Figure BDA0003168740510000054
In which L is req Is the farthest reachable distance of the controller. And selecting a clustering center in the cluster as an LEO satellite slave controller. As shown in equation (10), an intra-cluster adjacency matrix R with b × b weight as distance is established * Representing nodes in a cluster as logical addresses
Figure BDA0003168740510000055
The set of links within a cluster is denoted as R.
Figure BDA0003168740510000056
And the LEO satellite slave controller receives the state information of the local subnet, calculates a routing table in the cluster, and finally, the GEO satellite master controller is responsible for collecting the aggregation resource information of all edge nodes to realize the routing calculation among the clusters.
B. Establishing network resource mappings
B1, establishing a transmission cost model
By combining the characteristics of wide communication range, prolonged transmission time and high packet loss rate of a GEO/LEO double-layer satellite network, selecting link delay, packet loss rate, link residual bandwidth and node load which are acquired in real time as measurement parameters through a distributed SDN model, and establishing a transmission cost model, wherein the method comprises the following steps:
calculating the link delay: delay, which is the time required for data traffic to travel from a source node to a destination node, is one of the main performance metrics that measure QOS. Between any two nodes m, nLink l m,n Is expressed by the following equation:
Figure BDA0003168740510000057
in the formula (I), the compound is shown in the specification,
Figure BDA0003168740510000061
in order to achieve a delay in the transmission,
Figure BDA0003168740510000062
is the propagation delay.
Calculating the residual bandwidth: the remaining bandwidth refers to the amount of data transmission in the communication in which the link is unoccupied. When transmitting service data, two ports connected with two nodes at two ends of link, the byte sending rate of one port is equal to the byte receiving rate of the other port, so link l m,n The remaining bandwidth of (c) is only represented by the port data of node m:
Figure BDA0003168740510000063
in the formula, curr speed Indicates the bandwidth of the designated port of the node, in bytes(m,p) Represents the byte reception rate, out, of the p-port of node m bytes(m,p) Representing the byte transmission rate of the p-port of node m.
Calculating the packet loss rate: packet loss rate
Figure BDA0003168740510000064
Refers to the ratio of data packets not received to total transmitted data packets in a communication.
Figure BDA0003168740510000065
In the formula, rx packet(i,j) Indicates the total number of packets, tx, sent by node i packet(i,j) Indicating the total number of packets received by node j.
Calculating a transmission cost:
Figure BDA0003168740510000066
Representing service data packets b i On the link l m,n The transmission cost of (c).
Figure BDA0003168740510000067
In the formula (I), the compound is shown in the specification,
Figure BDA0003168740510000068
representing data traffic b i The more bandwidth is required for the link, the more bandwidth is required to be available, and the link/is selected m,n The greater the probability of (c). The packet loss rate is a multiplicative parameter and is converted into an additive parameter by taking a logarithm. q. q.s m Is the number of packets, Q, of the transmit queue in node m m Is the total length of the sending queue of node m, higher
Figure BDA0003168740510000069
Indicating a higher workload in queue m, with a larger Q m Node m of (a) forwards the packet more efficiently. Therefore, the temperature of the molten metal is controlled,
Figure BDA00031687405100000610
the capacity and the workload of the node are effectively described. Finally, the effective transmission capacity between nodes m and n is limited by the nodes with higher relative workloads. Delta delay Representing the maximum delay threshold of the path, delta band Represents the minimum residual bandwidth threshold, δ, of the path packetloss The maximum packet loss rate threshold is represented, and different thresholds are self-adaptive to different service types.
Thus, according to the intra-cluster adjacency matrix R * Establishing a time delay matrix of nodes in the cluster
Figure BDA0003168740510000071
Bandwidth matrix
Figure BDA0003168740510000072
And packet loss rate matrix
Figure BDA0003168740510000073
B2, aggregating cluster resources
And B, establishing an aggregation cluster resource set view for the main controller based on the SDN-based hierarchical clustering network model established in the step A. Setting the edge node set as U and the adjacent matrix as U, calculating the accumulated time delay of any two edge nodes i and j through a cluster resource aggregation algorithm in order to effectively transmit data between different clusters
Figure BDA0003168740510000074
Actual packet loss rate
Figure BDA0003168740510000075
And effective bandwidth
Figure BDA0003168740510000076
The transmission capability of a cluster is represented by the bandwidth, packet loss rate and time delay between edge nodes connecting two or more clusters.
The cluster resource aggregation algorithm comprises the following specific steps:
and Step1, initializing the time delay, the packet loss rate and the bandwidth of any pair of edge nodes.
Step2, the cumulative time delay of any two edge nodes is calculated.
And Step3, calculating the accumulated packet loss rate of any two edge nodes.
Step4 the minimum bandwidth between any two edge nodes is calculated and the effective bandwidth is constructed using the MIN-MAX principle.
And Step5, calculating the accumulated time delay, the actual packet loss rate and the effective bandwidth of any two edge nodes.
Step6 construction of time delay matrix between clusters
Figure BDA0003168740510000077
Packet loss rate matrix
Figure BDA0003168740510000078
Bandwidth matrix
Figure BDA0003168740510000079
C. Establishing multi-constraint QOS self-adaptive dynamic routing algorithm SDN-AD
In order to realize the self-adaptive dynamic routing algorithm based on the multi-constraint QoS, a multi-constraint QoS problem is formulated into an optimization problem with the aim of the lowest transmission cost. In order to provide personalized services and better adapt to network changes, adam (adaptive motion estimation) optimization algorithms are used to implement the adaptation of constraint thresholds. In order to better adapt to the high dynamic characteristics of SIGN and improve the convergence speed of the algorithm, an improved bee colony algorithm based on logistic regression and tabu search is used to solve the optimal solution problem. Adaptive routing is achieved by finding a path that satisfies constraints and has minimal transmission costs within and between clusters, the specific steps being as follows:
c1, problem definition
In order to find a path with minimum transmission cost from a source node s to a destination node d under the condition of satisfying the threshold constraints of time delay, residual bandwidth and packet loss rate
Figure BDA0003168740510000081
Converting the routing problem into a multi-constraint target optimization problem, wherein a mathematical model of the optimization problem is expressed in the form of:
Figure BDA0003168740510000082
Subjectto:
Figure BDA0003168740510000083
Figure BDA0003168740510000084
Figure BDA0003168740510000085
Figure BDA0003168740510000086
Figure BDA0003168740510000087
Q={b 1 ,b 2 ,...,b i } (21)
equation (15) defines the objective function to select the path with the minimum total transmission cost.
Figure BDA0003168740510000088
Is a service data service b i The link that is to be passed through,
Figure BDA0003168740510000089
denotes b i The path from source node s to destination node d,
Figure BDA00031687405100000810
representing a set of paths from a source node to a destination node.
Figure BDA00031687405100000811
The transmission cost representing a reachable path from the source node to the destination node is an accumulation of the transmission cost of each link on the path.
Considering three QoS constraint parameters of additive measurement parameter namely time delay, concave constraint namely bandwidth and multiplicative constraint namely packet loss rate, equations (16) - (18) formulate constraint conditions of paths from a source node s to a destination node d according to QoS route measurement performance indexes, and x in equation (19) k Indicating whether there is a reachable path between the source node and the destination node. Equation (20) requires coverage aggregation
Figure BDA0003168740510000093
Formula (21) represents the set of all data traffic.
C2, multi-priority threshold Adma solution
In order to better utilize network resources in a world-wide integrated network, support different requirements of different services, provide personalized services, and consider the priorities of the different services in the routing calculation. Unpacking a network protocol according to the programmable characteristic of the SDN, identifying different data services, and giving the following different priorities according to different requirements of the services on QoS:
and the priority level 1 is real-time, jitter-sensitive and high-interactivity service.
And the priority 2 is transaction data handling and interactive service.
And 3, priority level 3, namely services and video streams only requiring low packet loss rate.
According to the characteristics of the QoS requirement of the service, setting the following different objective calculation functions for the delay, bandwidth and packet loss rate thresholds of three different priority services:
Figure BDA0003168740510000091
Figure BDA0003168740510000092
Figure BDA0003168740510000101
in the formula, Pri represents the priority type of the service, so that the three types of services select different routing lines according to requirements, and the higher the priority is, the faster the processing is obtained. The type 1 priority service requires lower time delay, and allows the packet loss rate to be relatively higher. The class 3 priority service requires a lower packet loss rate and allows a relatively higher time delay.
In order to solve the self-adaptive threshold values under different service priorities, f (theta) is set as an objective function through an Adam algorithm, and theta is epsilon { b ∈ [ ] m,n, d m,n ,pl m,n Let g be the parameter to be solved t Representing each falling gradient, i.e. calculating f over a time step t t (θ) partial derivative for θ:
t=t+1。 (25)
Figure BDA0003168740510000102
m t =β 1 ·m t-1 +(1-β 1 )·g t (27)
Figure BDA0003168740510000103
Figure BDA0003168740510000104
Figure BDA0003168740510000105
Figure BDA0003168740510000106
Figure BDA0003168740510000107
in the formula, m t And v t Are respectively g t The first and second order moments of (1) are equivalent to the gradient g t And
Figure BDA0003168740510000108
is desired to be
Figure BDA0003168740510000109
And
Figure BDA00031687405100001010
and
Figure BDA00031687405100001011
are respectively m t And
Figure BDA00031687405100001012
since the initial value is 0, in order to reduce the influence of the bias toward 0, the difference between the expected value and the true second-order matrix is corrected by the formula (32), and the initialization offset is corrected. Alpha is the learning rate, used to control the stride, beta 1 、β 2 The first and second order moment attenuation coefficients. Alpha, beta 1 、β 2 And both epsilon and epsilon are super parameters without adjustment.
Therefore, the threshold is obtained through the algorithm and is used as a constraint condition of the routing algorithm, and the optimal path meeting different threshold constraints is selected through the routing algorithm.
C3, routing algorithm SDN-AD based on improved artificial bee colony algorithm (IABC)
In order to solve the multi-constraint target optimization problem, the traditional artificial bee colony algorithm, namely the ABC algorithm, is improved. In order to avoid the blindness of honey source exploitation, a logistic chaotic map is adopted for the generation of new honey sources, and a solution set which is closer to an optimal solution is generated. In order to realize the restriction on the path, a tabu search table is introduced according to a tabu search thought. And (4) recording the nodes optimized each time into a taboo table 1 until the destination node is found, and obtaining a path from the original node to the destination node. The paths that do not satisfy the constraints are logged in the tabu table 2 and are not accessed again next time.
In order to realize the multi-constraint QoS routing algorithm based on the IABC algorithm, the multi-constraint optimization model is used as the IABC algorithm model, three performance indexes of time delay, bandwidth and packet loss rate are taken as parameters to be absorbed into the same objective function, and the transmission cost is taken as an optimization target. And finally, selecting a path between the source node s and the destination node d, wherein the path meets the constraint condition and has the minimum transmission cost. The SDN-AD algorithm is calculated according to the following stages:
c31, initial stage
Setting the number of employed bees, follower bees and reconnaissance beesCommission maximum iteration number limit and algorithm maximum loop iteration number MCN. Using intra-cluster adjacency matrices R * To complete initialization of the honey source. A search space range is determined, i.e., the source node to all connectable next hop nodes. Randomly generating an initial solution x in a search space i SN, i.e. the initialization link l i SN is the number of honey sources. Each solution x i It is equivalent to a honey source and is a D-dimensional vector, D being the dimension of the problem.
C32 bee hiring stage
Each hiring bee generates a new solution, new honey source, from the following equation:
Figure BDA0003168740510000121
in the formula (I), the compound is shown in the specification,
Figure BDA0003168740510000122
for the chaotic sequence obtained by logistic mapping, the calculation formula is as follows:
Figure BDA0003168740510000123
wherein μ is a control parameter, n is 1, 2.
Figure BDA0003168740510000124
And taking the minimum path link price as the fitness, evaluating the quality of a new honey source according to the fitness when the new honey source is generated, selecting by adopting a greedy selection method, hiring the bees to select the new honey source if the new honey source is superior to the old honey source, and otherwise, keeping the old honey source, recording the fitness of the new honey source and handing the new honey source to the follower bees. The fitness formula is shown below, f i Function values representing solutions:
Figure BDA0003168740510000125
Figure BDA0003168740510000126
by the strategy, the search is expanded in the early stage of evolution and has the capability of jumping out of the local optimal solution, the search precision is effectively improved in the late stage of evolution, and the convergence is accelerated.
C33 bee-following stage
After the following bees obtain the honey source information transmitted by the employed bees, the selection probability calculation is carried out according to the fitness value of the feasible solution. The probability calculation formula is as follows:
Figure BDA0003168740510000127
the more suitable, i.e., less costly, link is easier to pick and search for. And obtaining a new honey source near the honey source by the formula (33), and keeping the honey source with better quality through comparing the quality of the new honey source with the quality of the old honey source.
To implement the problem model described in step C1, the search path is stored and the constraint condition determination is performed on the searched path, and two tabu tables are added based on the tabu search.
Each bee has its own memory, which stores the optimal solution nodes that the bee has visited in Tabu table 1, i.e. Tabu1, and these nodes will not be visited any more in subsequent searches. When the destination node is included in Tabu1, the bee gets a feasible path from the source node to the destination node
Figure BDA0003168740510000131
D, B, Pl obtained according to step B and equations (16) - (18) and delta delay 、δ band 、δ packetloss And judging whether the feasible path meets the constraint condition, if not, recording a Tabu table 2, namely Tabu2, and avoiding the algorithm from searching the path again.
C34, scouting bee stage
If the honey source is not updated after the search within the limit maximum iteration threshold, the honey source is declared exhausted, the employed bee is converted into a scout bee, and a new honey source is searched again in the solution space.
Compared with the prior art, the invention has the following beneficial effects:
1. in order to adapt to the problems of limited network resources and various service requirements in SIGN. The invention firstly provides a hierarchical clustering network model based on a distributed SDN, which can effectively reduce control overhead and improve transmission efficiency by reducing long-distance transmission and shortest-distance clustering of control packets.
Firstly, in order to deploy a relatively stable control scheme in a high-dynamic satellite network environment, frequent switching of control deployment is reduced. The invention is designed as a main controller according to the characteristics of large coverage area and relative stillness to the ground of the GEO satellite. According to the characteristics that the LEO satellite is high in movement speed and small in communication time delay, the LEO satellite is set as a slave controller.
Secondly, the invention carries out the deployment setting of the master controller and the slave controller. On one hand, the GEO satellite serving as the master controller only performs control transmission with the LEO satellite slave controller, and transmission of long-distance interplanetary links of control data packets is reduced. The network model of the hierarchical clustering provided by the invention is arranged on a path
Figure BDA0003168740510000132
A node
Figure BDA0003168740510000133
And high orbit satellite G n The distance over which the control information is transmitted is
Figure BDA0003168740510000134
While passing through the LEO satellite L as a slave controller h Spanning
Figure BDA0003168740510000141
At cluster time, the distance for transmitting control information is
Figure BDA0003168740510000142
Due to the fact that
Figure BDA0003168740510000143
The transmission distance of the control packet is effectively reduced. On the other hand, the slave controller only controls the nodes in the cluster to which the slave controller belongs, receives the local subnet state information, and calculates the routing table in the cluster, so that the transmission efficiency is improved, and more agile control is realized. And finally, the GEO satellite main controller is responsible for collecting the aggregation resource information of all edge nodes to realize the routing calculation among the clusters.
2. To maximize network throughput while ensuring quality of service, a multi-constraint QoS problem is formulated as an optimization problem targeting a minimum transmission cost. Compared with the traditional transmission cost considering a single factor, the transmission cost function considering the time delay, the bandwidth, the packet loss rate and the node load is provided. Based on the characteristics of large difference of transmission delay and transmission quality of inter-satellite links, inter-satellite links and satellite-ground links, the method can effectively calculate the transmission cost of different types of links and shield the difference between satellites and ground networks of different levels.
Meanwhile, in order to adapt to the high dynamic characteristic of the world-ground integrated network and adaptively update the threshold according to the priorities of different services on the premise of not consuming a large amount of computing resources, an adaptive threshold based on Adam is provided as a constraint condition. The Adam algorithm is a first-order optimization algorithm with random gradient descent, and can effectively solve the maximum value or the minimum value of an objective function. The algorithm combines the advantages of the AdaGrad (adaptive gradient algorithm) algorithm and the RMSProp (root mean square propagation) algorithm. The Adam algorithm firstly keeps the difference between the current update gradient and the last update gradient within a small range when calculating the first moment of the gradient, and realizes smooth gradient transition, thereby adapting to an unstable target function. Then, a second order matrix of gradients is set, and the learning rate is updated by calculating the average of the current gradient square and the past gradient square. Therefore, different thresholds can be self-adapted according to different priorities, and the algorithm is simple and efficient and does not occupy a large amount of computing resources and storage resources. Compared with the traditional manual setting method, the method can better adapt to network change, thereby providing services meeting different service quality requirements.
Finally, in order to improve the convergence speed of the algorithm, an improved bee colony algorithm based on logistic chaotic mapping and tabu search is provided. The method adopts logistic chaotic mapping to improve the initialization of the honey source, and the initialization is equal to the traditional ABC in value of [ -1, 1 [)]By comparing with random numbers of
Figure BDA0003168740510000151
The method can generate a solution set which is more close to the optimal solution and is uniformly distributed, and blindness of honey source exploitation is effectively avoided. And realize clustering through a tabu search table
Figure BDA0003168740510000152
The path setting in (1). Similarly, the main controller also uses algorithm 1 to perform cross-domain routing, and the only difference is that in the routing view of the main controller, each control domain is abstracted into a virtual node by using a cluster resource aggregation algorithm, that is, the inter-cluster adjacency matrix U is used. The optimization problem is solved, and the self-adaptive routing is realized.
The calculation result shows that compared with the prior art, the method has better performance in controlling the overhead, the network service quality and the algorithm convergence speed. In conclusion, the heaven-earth integrated network routing strategy under the SDN framework has good application prospect.
Drawings
Fig. 1 is a flow chart of the service forwarding of the present invention.
Fig. 2 is a route overhead curve.
Fig. 3 is a delay bound threshold throughput curve.
Fig. 4 is a bandwidth constrained threshold throughput curve.
Fig. 5 is a packet loss rate constraint threshold throughput curve.
Fig. 6 is an algorithm convergence curve.
Fig. 7 is an average end-to-end delay curve.
Fig. 8 is a throughput curve.
Fig. 9 is a packet loss rate curve.
Detailed Description
The invention is further described below with reference to the figures and examples.
The embodiment is as follows: the invention uses STK and MATLAB to carry out combined simulation, STK software can simulate a dynamic satellite network scene, three high-orbit satellites capable of covering the whole world and 30 iridium star bases are used as low-orbit satellites, and 20 non-uniformly distributed ground nodes are selected as ground stations. Updating the satellite coordinates once every 1s, calculating the coordinates according to the satellite positions, outputting position and orbit parameter information, generating a dynamic topology matrix, importing the generated dynamic topology into MATLAB, and then using the MATLAB to establish inter-domain paths in the domain and carry out dynamic routing strategy simulation based on the path transmission cost. The simulation data packet is sent at the ground station, transmitted through the satellite node and finally received by the ground station. The simulated satellite parameters and network parameter settings are shown in tables 1-2:
table 1: satellite parameter setting
Figure BDA0003168740510000161
Table 2: network parameter setting
Figure BDA0003168740510000162
Performance test of hierarchical clustering network model
In order to verify the performance of the hierarchical clustering network model, three SDN architectures are constructed for simulation, and in order to reduce other interference factors, the three SDN architectures are constructed on the same topology.
The differences between the controllers are the number and the positions of the controllers, namely a model with a high orbit satellite as a controller, a network model with the high orbit satellite and a ground control node as a combined controller and the hierarchical clustering network model. And setting the routing cost as the sum of the signaling transmission delay and the calculation resource overhead.
Setting total node number of single path
Figure BDA0003168740510000171
Wherein the number of nodes in a cluster is n, the number of clusters is m, the control signaling interaction is completed in one communication, and the total transmission distance of the original SDN architecture is obtained
Figure BDA0003168740510000172
Figure BDA0003168740510000173
Transmission distance of layered clustering architecture
Figure BDA0003168740510000174
It can be seen that when the number of cross-domains is m>1, the total transmission distance of the layered clustering framework is always smaller than that of the original SDN framework, so that the transmission delay of the control signaling is effectively reduced. Computing cost O (P) of global routing table for each communication 2 ) Compared with the hierarchical clustering architecture, the routing table scale used by intra-domain calculation is greatly simplified, and the calculation overhead is O (n) i 2 +m 2 ) A significant reduction is obtained.
Fig. 2 shows the optimization of the hierarchical clustering algorithm to the routing cost. With the increase of the number of nodes, most of the routes under the model of the invention are short-distance control packet transmission between the satellites and the earth, and the long-distance transmission of two interstellar links is increased only when the clusters are crossed. The routing overhead under the openSAN architecture continues to increase as GEO satellite controllers send long-range control packets to more and more nodes. Routing overhead under the SERvICE architecture can fluctuate greatly along with inter-satellite control packets. Simulation results prove that the more the number of the nodes is, the better the optimization effect of the model is.
Two, hierarchical clustering network model performance test
The multi-priority threshold solving scheme provided by the invention can find balance between the algorithm convergence speed and the overall network performance through threshold setting on the premise of ensuring the service quality by accurately setting the threshold for the fine-grained service quality requirements of different services. If the threshold is set too low, the routing success rate is low, and the convergence speed of the algorithm is too low or cannot be converged, so that the service quality is influenced; the threshold value is set too high, which causes local congestion of the service, and network resources cannot be fully used, thereby affecting network performance.
The threshold setting scheme is checked using the network throughput index. The network throughput is the maximum number of tasks that the network can process in a unit time, and the higher the throughput is, the faster the service completion speed is, and the more sufficient the network resources are utilized. Setting the time delay threshold value selection range of the routing algorithm to be 1-200, the bandwidth threshold value selection range to be 0-600 and the packet loss rate threshold value selection range to be 0.001-0.05. The total of 1000 services with three priorities are injected into the network at the same rate, the fluctuation of the throughput with the time delay, bandwidth and packet loss rate threshold is as shown in fig. 3-5, as can be seen from fig. 3-5, when the threshold is set too low, the routing success rate is low, the maximum network service is affected, and when the threshold is set too high, the routing scheme is selected and concentrated on a certain link, so that only the local network load is increased, and the optimal network performance still cannot be achieved. Meanwhile, the multi-priority threshold solving scheme used by the invention is always at or near the optimal threshold, and the necessity of setting the threshold and the effectiveness of the solving method are proved.
Performance test of IABC algorithm
The heuristic algorithm of the multi-constraint problem is difficult to prove the optimization performance from the theoretical perspective, and the optimization effect of the test function is further adopted to verify the algorithm performance. Under the condition of keeping other conditions consistent, all the simulation experiments are carried out on the QoS requests generated in the built satellite network model environment, QoS request parameters are randomly generated according to the table 3, and statistics is carried out on the algorithm, the original IABC algorithm and the literature [20 ]]The MABC algorithm converges up to 2000 iterations. Setting phi m When W is 0.4, W is 0.2 xd, colony size is 20, and Limit value is 100.
As can be seen from the convergence graph of FIG. 6, the algorithm of the present invention has a low convergence rate in the early stage compared to IABC and MABC, but the convergence rate is increased in the later stage of the search with the increase of the number of iterations, because the IABC algorithm has the performance of jumping out of the local optimal solution, a balance is found between the development and the exploitation, that is, the present invention improves the functions of the search formula and the probability selection formula. The algorithm of the invention has reached convergence when iterating 800 times, and the time for completing 2000 iterations is 11.5 s. The ABC algorithm and the MABC algorithm require more iterations to achieve convergence, and require longer running time to complete 2000 iterations.
Fourth, SDN-AD routing algorithm performance test
In order to evaluate the performance of the SDN-AD routing algorithm provided by the invention, the invention realizes the following three related algorithms, namely a classical distributed satellite network routing algorithm DRA and an MQoO algorithm for solving the problem of multi-constraint QoS by an approximate algorithm Lagrange relaxation method; the software defined routing algorithm SDRA. And the verification is performed from the following indexes, QoS quality, network throughput, ISL average flow and routing overhead. In order to simulate real world integration network application, 100 data packets with different service priorities are randomly generated, the average data packet size is set to be 10Mbit and is randomly distributed in a three-layer network, and the transmission rate of the data packets is different from 1Mbps to 4.5Mbps per second.
(1) Average end-to-end delay:
the end-to-end delay is an important index for measuring the QoS quality. Fig. 7 shows the variation of the average end-to-end delay of four different algorithms when the traffic flow is from 1Mbps to 4 Mbps. Since the DRA algorithm adopts a distributed routing scheme, the local optimization is easy to be involved, and the average time delay of the DRA algorithm is higher. When the global traffic is small, various algorithms can be well exerted, the SDN-AD algorithm has no obvious advantages, and the average end-to-end time delay is slightly higher than that of the SDRA algorithm and the MQoO algorithm. With the increase of network load, other algorithms begin to be congested, and as the SDN-AD algorithm sets a threshold according to the priority of a data flow when designing a QoS transmission cost model, a global minimum delay path can be found all the time, and a relatively excellent performance can be maintained all the time. As can be seen from fig. 7, when the transmission rate reaches the maximum, the SDN-AD algorithm can still maintain a lower average end-to-end delay, which is 17.2% higher than the mqao algorithm, and still maintain a better performance under a higher load.
(2) Network throughput:
fig. 8 shows the throughput variation of traffic flow from 1Mbps to 4Mbps, and when the traffic flow is less than 2Mbps, the load capacity of each algorithm on the network throughput is basically the same. As traffic volume increases, SDN-AD throughput increases more rapidly and consistently with higher throughput. When the network throughput reaches 30MBPS, the SDRA algorithm and the MQoO algorithm are close to saturation, the SDN-AD routing algorithm can timely collect the network QoS state based on a layered and domain-divided network model, in the link cost optimization target, the available bandwidth of a link, the working load of nodes on a path and the difference between inter-satellite links and ground links are considered, and the link with less workload is preferentially set as a transmission path, so that the throughput is close to saturation later. It can be seen that the average throughput of the SDN-AD algorithm is increased by 11.1% compared to the SDRA, 13.7% compared to the mqao, and 22.4% compared to the DRA algorithm.
(3) Average system packet loss rate:
fig. 9 shows the change of the packet loss rate when the traffic flow is from 1Mbps to 4 Mbps. The average system packet loss rate reflects the reliability of the network transmission scheme and reveals its adaptability to the environment. As the transmission rate increases, the link gradually becomes congested, and when the traffic flow is greater than 3.5Mbps, the packet loss rate starts to increase significantly. And the SDN-AD route evaluates the link cost and limits the adaptive threshold of the packet loss rate, so that the SDN-AD route can better adapt to the increase of global flow and process the global flow compared with a dijisra algorithm, and the workload can be balanced in a multilayer satellite network. As shown in the figure, the SDN-AD algorithm data packet loss rate can be stably kept to be minimum, and is 7.9% lower than that of the MQoO algorithm and about 25% lower than that of the SDRA and DRA algorithms.
The cluster resource aggregation algorithm in step B of the invention comprises the following steps:
Input:R,u,d m,n ,pl m,n ,b m,n
output: time delay matrix, bandwidth matrix and packet loss rate matrix of edge node
Input:dl,bl,pl,Q m ,Q n ,s,d
Output:
Figure BDA0003168740510000211
Figure BDA00031687405100002111
The present invention is not limited to the embodiment, and any equivalent idea or change within the technical scope of the present invention is to be regarded as the protection scope of the present invention.

Claims (1)

1. A space-ground integrated self-adaptive dynamic routing method based on SDN is characterized in that: the method comprises the following steps:
A. establishing layered clustering network model based on SDN
According to the characteristics that the coverage area of the GEO satellite is large and the ground is relatively static, the GEO satellite is set as a main controller; according to the characteristics of high movement speed and small communication time delay of the LEO satellite, the LEO satellite is set as a slave controller;
a1, determining a network model
The method comprises the steps that an SDN-based heaven-earth integrated network is arranged and composed of LEO satellite nodes, GEO satellite nodes and ground nodes; representing a topological graph of the SDN-based heaven-earth integrated network as an undirected weighted topological graph G (V, E) and storing the undirected weighted topological graph in a GEO satellite main controller, wherein V { V ═ V } G ,v L ,v T Represents a finite set of all nodes in the network, v G ,v L ,v T Respectively representing node sets of GEO satellite nodes, LEO satellite nodes and ground nodes; e ═ (IOL ═ ISL @ UDL } represents the set of all links in the network, IOL represents the interplanetary link, ISL represents the interstellar link, UDL represents the satellite-to-ground link;
g for nth GEO satellite n Expressed as, 1. ltoreq. N. ltoreq.N G Let N be in order for the GEO satellite to cover the global area G Not less than 3; dividing the whole satellite network into N according to the coverage range of the GEO satellite G A region, i.e.
Figure FDA0003750098190000011
Figure FDA0003750098190000012
L for the h LEO satellite h Indicates that the total number of LEO satellites is M L ×N L ,M L Representing the number of satellite orbital planes, N L Representing the number of satellites in the satellite orbital plane;
dividing the earth surface into different areas according to longitude and latitude according to the virtual node idea, and setting a logical address for each area; the logical address in each region remains unchanged, but the LEO satellite membership therein changes; when a member of one LEO satellite leaves a logical area, the next LEO satellite enters the logical area to replace its location; determining coverage clustering of the GEO satellite according to the following conditions;
Figure FDA0003750098190000013
Figure FDA0003750098190000021
Figure FDA0003750098190000022
in the formula (I), the compound is shown in the specification,
Figure FDA0003750098190000023
representing LEO satellites L h And GEO satellite G n A half center angle of (d);
Figure FDA0003750098190000024
represents G n Maximum half-center angle covered by the orbital plane of the LEO satellite;
Figure FDA0003750098190000025
is L h A radial distance to the geocenter; g' n Represents G n Mapping nodes on an LEO satellite orbit plane; l G' n L h L represents G' n And L h The distance between them;
Figure FDA0003750098190000026
is G n A radial distance to the geocenter;
Figure FDA0003750098190000027
represents the minimum elevation angle of the GEO satellite relative to the orbital plane of the LEO satellite;
the LEO satellite satisfying the formula (3) is determined according to | G' n L h The minimum principle is uniquely marked in a cluster of the GEO satellite; because LEO satellite coverage areas correspond to the ground areas one by one, LEO satellite members and ground station members in the GEO satellite coverage area are determined; the overall network topology under the SDN-based hierarchical clustering network model can be determined only by establishing a controller cluster configuration strategy to obtain a satellite dynamic connection relation;
a2 configuration controller cluster
X, Y, Z are respectively set as the earth fixed coordinate systems of the single LEO satellite, the origin of the earth fixed coordinate system is the geocentric, the XOY plane is coincident with the equatorial plane of the earth, the OX axis is consistent with the Greenwich meridian direction, and the OZ axis is coincident with the polar axis of the earth; LEO satellite L is obtained from equations (4) to (9) without considering the effects of other celestial and satellite vehicles on LEO satellites j And LEO satellite L k Actual distance therebetween
Figure FDA0003750098190000028
μ=μ o +θΔt (4)
Ω G =ωΔt (5)
X=H[cosμcos(Ω-Ω G )-sinμsin(Ω-Ω G )cosσ] (6)
Y=H[cosμsin(Ω-Ω G )-sinμcos(Ω-Ω G )cosσ] (7)
Z=H(sinμsinσ) (8)
Figure FDA0003750098190000029
In the formula, mu o The satellite near-location amplitude angle, theta is a linear velocity, delta t is a minimum time, omega is an earth rotation angular velocity, omega is a satellite ascent point right ascension and sigma is a satellite orbit inclination angle;
for each divided region V i Designing a slave controller configuration clustering strategy of the LEO satellite as follows: taking the number a of slave controllers of the LEO satellite as V i The cluster number in the slave controller set is expressed as LC ═ c 1 ,c 2 ,...,c a },c i The administered cluster is expressed as
Figure FDA0003750098190000031
Dividing all data into a clusters, and selecting a cluster centers; according to the formula (9), calculating the LEO satellite nodes to the clustering center c i Is a distance of
Figure FDA0003750098190000032
Will be in line with
Figure FDA0003750098190000033
Required nodes to be grouped into clusters
Figure FDA0003750098190000034
In which L is req Is the furthest reachable distance of the controller; selecting a clustering center in the cluster as an LEO satellite slave controller; as shown in equation (10), an intra-cluster adjacency matrix R with b × b weight as distance is established * Representing nodes in a cluster as logical addresses
Figure FDA0003750098190000035
The intra-cluster link set is denoted as R;
Figure FDA0003750098190000036
the LEO satellite slave controller receives the state information of the local subnet, calculates a routing table in the cluster, and finally, the GEO satellite master controller is responsible for collecting the aggregation resource information of all edge nodes to realize the routing calculation between the clusters;
B. establishing network resource mappings
B1, establishing a transmission cost model
By combining the characteristics of wide communication range, prolonged transmission time and high packet loss rate of a GEO/LEO double-layer satellite network, selecting link delay, packet loss rate, link residual bandwidth and node load which are acquired in real time as measurement parameters through a distributed SDN model, and establishing a transmission cost model, wherein the steps are as follows:
calculating the link delay: delay refers to the time required for a data service from a source node to a destination node, and is one of main performance indexes for measuring QOS; link l between any two nodes m, n m,n Is expressed by the following equation:
Figure FDA0003750098190000037
in the formula (I), the compound is shown in the specification,
Figure FDA0003750098190000038
in order to achieve a delay in the transmission,
Figure FDA0003750098190000039
is the propagation delay;
calculating the residual bandwidth: the residual bandwidth refers to the data transmission quantity of unoccupied links in communication; when transmitting service data, two ports connected with two nodes at two ends of the link, the byte sending rate of one port is equal to the byte receiving rate of the other port, therefore, the link l m,n The remaining bandwidth of (c) is only represented by the port data of node m:
Figure FDA0003750098190000041
in the formula, curr speed Indicates the bandwidth of the designated port of the node, in bytes(m,p) Represents the byte reception rate, out, of the p-port of node m bytes(m,p) Represents the byte transmission rate of the p port of node m;
calculating the packet loss rate: packet loss rate
Figure FDA0003750098190000042
The ratio of data packets not received in the communication to the total transmitted data packets;
Figure FDA0003750098190000043
in the formula, rz packet(i,j) Indicates the total number of packets, tx, sent by node i packet(i,j) Representing the total number of data packets received by node j;
calculating the transmission cost:
Figure FDA0003750098190000044
representing service data packets b i On the link l m,n The cost of transmission over;
Figure FDA0003750098190000045
in the formula (I), the compound is shown in the specification,
Figure FDA0003750098190000046
representing data traffic b i The more bandwidth is required for the link, the more bandwidth is required to be available, and the link/is selected m,n The greater the probability of (c); the packet loss rate is a multiplicative parameter and is converted into an additive parameter by taking a logarithm; q. q.s m Is the number of packets, Q, of the transmit queue in node m m Is the total length of the sending queue of node m, higher
Figure FDA0003750098190000047
Indicating a higher workload in queue m, with a larger Q m The node m of (1) forwards the data packet more effectively; therefore, the temperature of the molten metal is controlled,
Figure FDA0003750098190000048
the capacity and the workload of the node are effectively described; finally, the effective transmission capacity between nodes m and n is limited by the nodes with higher relative workload; delta delay Representing the maximum delay threshold of the path, delta band Represents the minimum residual bandwidth threshold, δ, of the path packetloss The maximum packet loss rate threshold is represented, and different thresholds are self-adapted to different service types;
thus, according to the intra-cluster adjacency matrix R * Establishing a time delay matrix of nodes in the cluster
Figure FDA0003750098190000051
Bandwidth matrix
Figure FDA0003750098190000052
And packet loss rate matrix
Figure FDA0003750098190000053
B2, aggregating cluster resources
Establishing an aggregation cluster resource set view for the main controller based on the SDN-based hierarchical clustering network model established in the step A; setting the edge node set as U and the adjacent matrix as U, calculating the accumulated time delay of any two edge nodes i and j through a cluster resource aggregation algorithm in order to effectively transmit data between different clusters
Figure FDA0003750098190000054
Actual packet loss rate
Figure FDA0003750098190000055
And effective bandwidth
Figure FDA0003750098190000056
The transmission capability of the cluster is represented by bandwidth, packet loss rate and time delay among edge nodes connecting two or more clusters;
the cluster resource aggregation algorithm comprises the following specific steps:
step1: initializing the time delay, packet loss rate and bandwidth of any pair of edge nodes;
step2: calculating the accumulated time delay of any two edge nodes;
step3: calculating the accumulated packet loss rates of any two edge nodes;
step4: calculating the minimum bandwidth between any two edge nodes, and constructing an effective bandwidth by using an MIN-MAX principle;
step5: calculating the accumulated time delay, the actual packet loss rate and the effective bandwidth of any two edge nodes;
step6: constructing a time delay matrix D ═ D (D) between clusters i,j ) |U| And the packet loss rate matrix Pl ═ Pl (Pl) m,n ) |U| And the bandwidth matrix B ═ B i,j ) |U|
C. Establishing multi-constraint QOS self-adaptive dynamic routing algorithm SDN-AD
In order to realize a self-adaptive dynamic routing algorithm based on multi-constraint QoS, a multi-constraint QoS problem is formulated into an optimization problem with the target of the lowest transmission cost; in order to provide personalized services and better adapt to network changes, Adam (adaptive motion estimation) optimization algorithm is used for realizing the adaptation of constraint threshold; in order to better adapt to the high dynamic characteristic of SIGN and improve the convergence speed of the algorithm, an improved bee colony algorithm based on logistic regression and tabu search is used to solve the problem of optimal solution; adaptive routing is achieved by finding a path that satisfies constraints and has minimal transmission costs within and between clusters, the specific steps being as follows:
c1, problem definition
In order to find a path with minimum transmission cost from a source node s to a destination node d under the condition of meeting the threshold constraints of time delay, residual bandwidth and packet loss rate
Figure FDA0003750098190000061
Converting the routing problem into a multi-constraint target optimization problem, wherein a mathematical model of the optimization problem is expressed in the form of:
Figure FDA0003750098190000062
Subject to:
Figure FDA0003750098190000063
Figure FDA0003750098190000064
Figure FDA0003750098190000065
Figure FDA0003750098190000066
Figure FDA0003750098190000067
Figure FDA0003750098190000068
equation (15) defines the objective function to select the path with the minimum total transmission cost;
Figure FDA0003750098190000069
is a service data service b i The link that is to be passed through,
Figure FDA00037500981900000610
is shown by b i The path from source node s to destination node d,
Figure FDA00037500981900000611
representing a path set from a source node to a destination node;
Figure FDA00037500981900000612
the transmission cost of an reachable path from a source node to a destination node is represented and is the accumulation of the transmission cost of each link on the path;
considering three QoS constraint parameters of additive measurement parameter namely time delay, concave constraint namely bandwidth and multiplicative constraint namely packet loss rate, equations (16) - (18) formulate constraint conditions of paths from a source node s to a destination node d according to QoS route measurement performance indexes, and x in equation (19) k Indicating whether an reachable path exists between the source node and the destination node; equation (20) requires coverage aggregation
Figure FDA0003750098190000071
Formula (21) represents the set of all data traffic;
c2, Multi-priority threshold Adam solution
In order to better utilize network resources in a world-wide integrated network, support different requirements of different services and provide personalized services, the priorities of the different services should be considered in routing calculation; unpacking a network protocol according to the programmable characteristic of the SDN, identifying different data services, and giving the following different priorities according to different requirements of the services on QoS:
priority 1: real-time, jitter sensitive, high-interactivity services;
priority 2: transacting data and interactive services;
priority 3: only services and video streams with low packet loss rate are required;
according to the characteristics of the QoS requirement of the service, setting the following different objective calculation functions for the delay, bandwidth and packet loss rate thresholds of three different priority services:
Figure FDA0003750098190000072
Figure FDA0003750098190000073
Figure FDA0003750098190000074
in the formula, Pri represents the priority type of the service, so that different routing lines are selected by the three types of services according to requirements, and the higher the priority is, the faster the processing is obtained; the type 1 priority service requires lower time delay, and allows the packet loss rate to be relatively higher; the 3-type priority service requires lower packet loss rate and relatively higher allowable time delay;
in order to solve the self-adaptive threshold values under different service priorities, f (theta) is set as an objective function through an Adam algorithm, and theta is in an element of { b ∈ m,n ,d m,n ,pl m,n Let g be the parameter to be solved t Representing each falling gradient, i.e. calculating f over a time step t t (θ) partial derivative for θ:
t=t+1; (25)
Figure FDA0003750098190000081
m t =β 1 ·m t-1 +(1-β 1 )·g t (27)
Figure FDA0003750098190000082
Figure FDA0003750098190000083
Figure FDA0003750098190000084
Figure FDA0003750098190000085
Figure FDA0003750098190000086
in the formula, m t And v t Are respectively g t The first and second order moments of (1) are equivalent to the gradient g t And
Figure FDA0003750098190000087
is desired to be
Figure FDA0003750098190000088
And
Figure FDA0003750098190000089
Figure FDA00037500981900000810
and
Figure FDA00037500981900000811
are respectively m t And
Figure FDA00037500981900000812
because the initial value is 0, in order to reduce the influence of the bias toward 0, the difference between the expected value and the real second-order matrix is corrected by the formula (32), and the initialization deviation is corrected; alpha is the learning rate, used to control the stride, beta 1 、β 2 The first and second order moment attenuation coefficients; alpha, beta 1 、β 2 And the epsilon is super parameters without adjustment; beta is a 1 t 、β 2 t 、β 2 t-i Respectively represents beta 1 To the power of t, beta 2 To the power of t, beta 2 To the t-i power of;
therefore, the threshold is obtained through the algorithm and is used as a constraint condition of the routing algorithm, and the optimal path meeting different threshold constraints is selected through the routing algorithm;
c3, routing algorithm SDN-AD based on improved artificial bee colony algorithm (IABC)
In order to solve the multi-constraint target optimization problem, the traditional artificial bee colony algorithm, namely an ABC algorithm, is improved; in order to avoid the blindness of honey source exploitation, a logistic chaotic map is adopted for the generation of new honey sources to generate a solution set which is closer to an optimal solution; in order to realize the restriction on the path, a tabu search table is introduced according to a tabu search thought; recording the nodes optimized each time into a taboo table 1 until a destination node is found, and obtaining a path from the original node to the destination node; recording the paths which do not meet the constraint into a taboo table 2, and not accessing the paths next time;
in order to realize the multi-constraint QoS routing algorithm based on the IABC algorithm, the multi-constraint optimization model is an IABC algorithm model, three performance indexes of time delay, bandwidth and packet loss rate are taken as parameters and absorbed into the same objective function, and the transmission cost is an optimization target; finally, a path between the source node s and the destination node d is selected, the constraint condition is met, and the transmission cost is the minimum; the SDN-AD algorithm is calculated according to the following stages:
c31, initial stage
Setting the number of hiring bees, follower bees and reconnaissance bees, the maximum iteration number limit of hiring bees and the maximum cycle iteration number MCN of the algorithm; using intra-cluster adjacency matrices R * To complete initialization of the honey source; determining a search space range from a source node to all connectable next-hop nodes; randomly generating an initial solution x in a search space i 1,2, SN, i.e. the initialization link l i SN is the number of honey sources; each solution x i It is equivalent to a honey source and is a vector with D dimension, D is the dimension of the problem;
c32 bee hiring stage
Each hiring bee generates a new solution, i.e., a new honey source, from the following formula:
Figure FDA0003750098190000091
in the formula (I), the compound is shown in the specification,
Figure FDA0003750098190000101
for the chaotic sequence obtained by logistic mapping, the calculation formula is as follows:
Figure FDA0003750098190000102
where μ is a control parameter, n 1, 2.
Figure FDA0003750098190000103
Taking the minimum link price as fitness, when a new honey source is generated, evaluating the quality of the new honey source according to the fitness, selecting by adopting a greedy selection method, if the fitness is better than that of the old honey source, hiring bees to select the new honey source, and otherwise, keeping the old honey source, recording the fitness of the new honey source, and handing the fitness to a follower bee; the fitness formula is shown below, f i Function values representing solutions:
Figure FDA0003750098190000104
Figure FDA0003750098190000105
by the strategy, the search is expanded in the early stage of evolution and has the capability of jumping out of a local optimal solution, the search precision is effectively improved in the late stage of evolution, and the convergence is accelerated;
c33 bee-following stage
After the following bees obtain honey source information transmitted by the employed bees, carrying out probability calculation according to fitness values of feasible solutions; the probability calculation formula is as follows:
Figure FDA0003750098190000106
the higher the fitness, namely the lower the cost, the more easily the link is selected, and the deep search is carried out on the link; then obtaining a new honey source near the honey source by the formula (33), and keeping the honey source with better quality through comparing the advantages and disadvantages of the new and old honey sources;
in order to realize the problem model in the step C1, storing the searched path and judging the constraint condition of the searched path, and adding two tabu tables on the basis of reference tabu search;
each bee has own memory, the optimal solution nodes which the bee has visited are stored in the memory by using a Tabu table 1, namely Tabu1, and the nodes cannot be visited any more in the subsequent search; when the destination node is included in Tabu1, the bee gets a feasible path from the source node to the destination node
Figure FDA0003750098190000111
D, B, Pl obtained according to step B and equations (16) - (18) and delta delay 、δ band 、δ packetloss Judging whether the feasible path meets the constraint condition, if not, recording a Tabu table 2, namely Tabu2, and avoiding the algorithm from searching the path again;
c34, scouting bee stage
If the honey source is not updated after the search within the limit maximum iteration threshold, the honey source is declared exhausted, the employed bee is converted into a scout bee, and a new honey source is searched again in the solution space.
CN202110812489.2A 2021-07-19 2021-07-19 Heaven and earth integrated self-adaptive dynamic QoS routing method based on SDN Active CN113572686B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110812489.2A CN113572686B (en) 2021-07-19 2021-07-19 Heaven and earth integrated self-adaptive dynamic QoS routing method based on SDN

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110812489.2A CN113572686B (en) 2021-07-19 2021-07-19 Heaven and earth integrated self-adaptive dynamic QoS routing method based on SDN

Publications (2)

Publication Number Publication Date
CN113572686A CN113572686A (en) 2021-10-29
CN113572686B true CN113572686B (en) 2022-09-02

Family

ID=78165496

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110812489.2A Active CN113572686B (en) 2021-07-19 2021-07-19 Heaven and earth integrated self-adaptive dynamic QoS routing method based on SDN

Country Status (1)

Country Link
CN (1) CN113572686B (en)

Families Citing this family (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114268575B (en) * 2021-12-23 2023-07-14 苏州全时空信息技术有限公司 Self-adaptive three-dimensional transmission method and system in space-earth integrated information network
CN114301791A (en) * 2021-12-29 2022-04-08 中国电信股份有限公司 Data transmission method and device, storage medium and electronic equipment
CN114499639B (en) * 2022-01-19 2024-02-09 马爱 Ant colony optimization routing method with multiple QoS constraints in low-orbit satellite network
CN114745321B (en) * 2022-03-21 2023-06-16 南京邮电大学 QoS perception routing method of satellite network based on SDN
CN114938235B (en) * 2022-04-08 2023-05-30 北京邮电大学 Satellite network data transmission method and device integrating multipath and network coding
CN115276755B (en) * 2022-06-20 2023-05-02 哈尔滨工业大学(威海) Inter-satellite link and power distribution method for satellite network communication
CN114978979B (en) * 2022-07-27 2022-10-14 网络通信与安全紫金山实验室 Route generation method, apparatus and storage medium
CN115632701B (en) * 2022-12-21 2023-04-18 中南大学 Low-earth-orbit satellite network data scheduling method, system, equipment and storage medium
CN116016278A (en) * 2022-12-22 2023-04-25 四川九州电子科技股份有限公司 Dynamic adjustment method for EDCA parameters
CN116545853B (en) * 2023-07-04 2023-10-13 南京理工大学 Integrated network multi-objective optimized resource management method based on quantum particle swarm
CN116600352B (en) * 2023-07-19 2023-09-15 北京最终前沿深空科技有限公司 Space-earth integrated QoS consistency processing method, qoS convergent and QoS orchestrator
CN117544220A (en) * 2023-11-08 2024-02-09 中国人民解放军军事科学院系统工程研究院 Routing control method and device for high-low orbit satellite communication network
CN117675679A (en) * 2023-11-24 2024-03-08 航天恒星科技有限公司 Satellite time sensitive route scheduling method based on domain optimization

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108307435A (en) * 2018-01-29 2018-07-20 大连大学 A kind of multitask route selection method based on SDSIN
CN110493131A (en) * 2019-09-24 2019-11-22 大连大学 A kind of design method of Information Network routing policy under SDN framework

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110730131B (en) * 2019-10-22 2020-07-17 电子科技大学 SDN satellite network multi-QoS constraint routing method based on improved ant colony

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108307435A (en) * 2018-01-29 2018-07-20 大连大学 A kind of multitask route selection method based on SDSIN
CN110493131A (en) * 2019-09-24 2019-11-22 大连大学 A kind of design method of Information Network routing policy under SDN framework

Also Published As

Publication number Publication date
CN113572686A (en) 2021-10-29

Similar Documents

Publication Publication Date Title
CN113572686B (en) Heaven and earth integrated self-adaptive dynamic QoS routing method based on SDN
CN111148161A (en) Method and system for balancing load route between low-orbit satellite constellation satellites
CN108566663A (en) SDWSN energy consumption balance routing algorithms based on disturbance particle group optimizing
CN109981438A (en) A kind of satellite network load-balancing method towards SDN and NFV cooperation deployment frame
Tang Dynamically adaptive cooperation transmission among satellite-ground integrated networks
CN105282038B (en) For the distributed group of stars group optimization method based on stability analysis in mobile satellite network
CN108307435A (en) A kind of multitask route selection method based on SDSIN
CN113099505B (en) Air-space-ground integrated network routing method
Tang et al. Federated learning for intelligent transmission with space-air-ground integrated network (SAGIN) toward 6G
CN103888976A (en) Link selection method combined with network scheduling and routing
CN105472484A (en) Wave channel balancing route wavelength allocation method of power backbone optical transport network
Wang et al. A load balanced routing algorithm based on congestion prediction for LEO satellite networks
Liu et al. Routing for predictable multi-layered satellite networks
Han et al. Dynamic routing for software-defined LEO satellite networks based on ISL attributes
Chen et al. Time-varying resource graph based resource model for space-terrestrial integrated networks
Han et al. Space edge cloud enabling service migration for on-orbit service
Shi et al. Multi-QoS adaptive routing algorithm based on SDN for satellite network
Meng et al. Intelligent routing orchestration for ultra-low latency transport networks
CN114268575B (en) Self-adaptive three-dimensional transmission method and system in space-earth integrated information network
CN114513241B (en) SDN-based high-performance QoS guaranteed low-orbit satellite inter-satellite routing method
Yi et al. Route strategy of satellite network in GNSS based on topology evolution law
CN108429577A (en) A kind of satellite QoS routing algorithms based on PROMETHEE methods
Zhang et al. Dynamical controller placement among SDN space-terrestrial integrated networks
Shi et al. Reinforcement learning routing in space-air-ground integrated networks
Chen et al. Time-varying resource graph based processing on the way for space-terrestrial integrated vehicle networks

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant