CN108307435A - A kind of multitask route selection method based on SDSIN - Google Patents
A kind of multitask route selection method based on SDSIN Download PDFInfo
- Publication number
- CN108307435A CN108307435A CN201810083451.4A CN201810083451A CN108307435A CN 108307435 A CN108307435 A CN 108307435A CN 201810083451 A CN201810083451 A CN 201810083451A CN 108307435 A CN108307435 A CN 108307435A
- Authority
- CN
- China
- Prior art keywords
- satellite
- network
- link
- sdsin
- multitask
- 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.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/0268—Traffic management, e.g. flow control or congestion control using specific QoS parameters for wireless networks, e.g. QoS class identifier [QCI] or guaranteed bit rate [GBR]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/14—Relay systems
- H04B7/15—Active relay systems
- H04B7/185—Space-based or airborne stations; Stations for satellite systems
- H04B7/1853—Satellite systems for providing telephony service to a mobile station, i.e. mobile satellite service
- H04B7/18558—Arrangements for managing communications, i.e. for setting up, maintaining or releasing a call between stations
- H04B7/1856—Arrangements for managing communications, i.e. for setting up, maintaining or releasing a call between stations for call routing
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L45/00—Routing or path finding of packets in data switching networks
- H04L45/302—Route determination based on requested QoS
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
- H04L47/12—Avoiding congestion; Recovering from congestion
- H04L47/125—Avoiding congestion; Recovering from congestion by balancing the load, e.g. traffic engineering
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/08—Load balancing or load distribution
- H04W28/082—Load balancing or load distribution among bearers or channels
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W40/00—Communication routing or communication path finding
- H04W40/02—Communication route or path selection, e.g. power-based or shortest path routing
- H04W40/04—Communication route or path selection, e.g. power-based or shortest path routing based on wireless node resources
- H04W40/10—Communication route or path selection, e.g. power-based or shortest path routing based on wireless node resources based on available power or energy
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W40/00—Communication routing or communication path finding
- H04W40/02—Communication route or path selection, e.g. power-based or shortest path routing
- H04W40/12—Communication route or path selection, e.g. power-based or shortest path routing based on transmission quality or channel quality
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W84/00—Network topologies
- H04W84/02—Hierarchically pre-organised networks, e.g. paging networks, cellular networks, WLAN [Wireless Local Area Network] or WLL [Wireless Local Loop]
- H04W84/04—Large scale networks; Deep hierarchical networks
- H04W84/06—Airborne or Satellite Networks
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Astronomy & Astrophysics (AREA)
- General Physics & Mathematics (AREA)
- Aviation & Aerospace Engineering (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a kind of multitask route selection method based on SDSIN, includes the following steps：Establish rational network model；Space tasks grade classification；Based on SDSIN multitask route selection method founding mathematical models；Link parameter decision；Right value update.The present invention is based on the algorithms of machine learning, the Multiple factors such as link state parameter and space tasks are fully considered, realize optimal route selection, improve the utilization rate of Internet resources, it avoids local link and the state of congestion occurs, the QoS demand of all kinds of space tasks is effectively met simultaneously, the present invention fully considers propagation delay time, the link characteristics such as packet loss and remaining bandwidth, training set is trained according to above three parameter, update link parameter weights, effectively multitask routing is collected, improve the utilization rate of underlying resource, realize the load balancing of the whole network.
Description
Technical field
The present invention relates to satellite network, especially a kind of software definition Information Network (Software Defined
Spatial Information Network, abbreviation SDSIN).
Background technology
It is right with various new tasks, the appearance of new opplication (oceangoing voyage, emergency management and rescue, navigator fix, space flight measurement and control etc.)
Wireless network proposes numerous different new demands.In order to solve, wireless network node type is various, capacity variance is big, technology body
System differs, communicating link data propagation delay time and error rate height, network topology dynamic change, business be with multifarious Service Quality
The problems such as amount demand, researcher use for reference software defined network (Software Defined Network, SDN) technology, shield
Agreement isomerism so that network is able to rapid deployment.However acquiescence uses Shortest path routing strategy, the plan in SDN network
Shortest path routing is carried out only for hop count when omiting routing, does not consider service quality (the Quality of of space tasks
Service, QoS) demand, such as remaining bandwidth, propagation delay time, packet loss.At the same time, this routing mode may be such that
Certain link is repeatedly chosen, and causes the link load excessive, causes link obstructions, and then influence the QoS demand of space tasks.
Therefore, route selection method needs to fully consider the QoS demand of different spaces task, and mission requirements can be met most by therefrom selecting
Shortest path.
Currently, in relation to considering that the method for the network parameters such as link bandwidth, time delay, handling capacity has in SDN routing policies：
A kind of VL-EDF (Virtual Latency Earliest Deadline First Algorithm, VL-EDF)
Queue scheduling algorithm, it can serve the data of delay sensitive well；It also proposed the insensitive data flow tune of time delay simultaneously
Algorithm LBFS (Latcncy Based Flow Scheduling Algorithm) is spent, switching node port can be carried out
Monitoring, when port flow load is more than given threshold, by the insensitive data flow handoff of time delay in respective links to underloading chain
On the road, the problem of but being the failure to fully consider link bandwidth and handling capacity；Have again research and propose it is a kind of based on data center
ECMP algorithms, the algorithm improve the load balancing degrees of link in network according to the realization dynamic routing of the bandwidth of link, but not
There is the service attribute in view of data flow, selects the remaining bandwidth of link as routing standard, cannot relatively accurately reflect link
Time delay, packet loss conditions.
Other than the angle from link state accounts for SDN routing policies, also angle pair of the researcher from QoS
Routing policy is analyzed, and wherein typical method has：Fine granularity processing is carried out to data stream.Researcher proposes a kind of end and arrives
Data flow in network is divided into high-priority traffic and background stream by the flow mode at end, the model, with link when
Prolong, standard of the bandwidth as routing, according to the time delay of link, selects path to Bandwidth Dynamic, the high data flow of priority is excellent
Advanced walking along the street is by ensureing its QoS attribute.But this mode only ensures a kind of QoS attributes of data flow, leads to other
The QoS attributes of data flow are uncontrollable, still can't resolve the problem that link load is unbalanced, network resource utilization is low.
Invention content
To solve the above problem of the existing technology, the present invention to design it is a kind of can fully consider link state parameter,
Realize that the Multi-path route selection method based on SDSIN of optimal route selection avoids part to improve network resource utilization
Link congestion realizes the load balancing of network.
To achieve the goals above, mentality of designing of the invention is：Using a kind of spatial information net based on software definition
Network (Software Defined Spatial Information Network, SDSIN) framework, actual deployment scheme is such as
Under：
SDN core concepts are applied to Information Network, by by satellite data plane and control plane separation, making to defend
Star only needs to implement simply to forward, in the deployment of actual network, space-based backbone network that data plane is made of repeater satellite
The access net composition constituted with user's satellite；Control plane high concentration is carried by management and control centre punch one.SDN controllers possess
Global view inquires link state and network flow etc. by southbound interface.
Under SDSIN frameworks, by the way that Information Network Satellite node resource and inter-satellite link resource are abstracted,
The flexible shared of resource can be achieved, the dynamic control of Information Network can be realized by the controller of centralization.
A kind of multitask route selection method based on SDSIN, includes the following steps：
A, the node that satellite network is redefined according to the thought of SDN establishes rational network model
Under SDSIN frameworks, though satellite topology is that high dynamic changes, since satellite is transported according to definitive orbit
Capable, movement is that have predictability in periodically variable.For this feature, Route Selection takes topology to control
Mode is to shield its high dynamic variation characteristic.SDSIN network dynamic topologies are discretized into n timeslice according to system cycle T,
It is denoted as T1,T2,…,Tn, topological novariable is thought in timeslice, and finally each timeslice is routed.
A1, network model is defined
Wireless network topology is indicated with G (V, E, C), wherein V={ v1,v2,...,vmIndicate telephone net node collection
It closes.E={ E1,E2,…EnThe set that SDSIN Intersatellite Links form is represented, C indicates the set of controller node, wherein controlling
Device is placed on high rail satellite GEO.
A2, link connection matrix is defined
Software definition satellite network model is G (V, E, C), is by a series of high rail satellite GEO, middle rail satellite MEO, low rail
What satellite LEO and inter-satellite link formed, link connection matrix is expressed as：
Wherein aijIt is expressed as node i and the link connection relationship of node j, wherein：
A3, link state parameter is defined
SDN controllers possess global view, and link state and network flow are inquired by southbound interface.For inter-satellite link
SubsetThere are corresponding time delay, remaining bandwidth, packet loss information, uses d respectivelyij、bij、sijIt indicates.It can
P is shared with the collection in path to indicate, for any subpath piUnder ∈ P,
Time delay function representation is：
Remaining bandwidth function representation is：
Packet loss function representation is：
Link state parameter is expressed as：
F=[f1(pi),f2(pi),f3(pi)] (6)
A4, available path collection is defined
With (vs,vd) indicate network model in source node and destination node node pair, then
One is represented from source node vsTo destination node vdPath.It is assumed that sharing m path set after mission dispatching, it is expressed as：P=
{P1,P2,…,Pm, wherein PiIndicate i-th available path collection.In network model G (V, E, C), available path is expressed as P'
={ p1,p2,…,pq, wherein 1≤q≤m.
B, space tasks grade classification
The premise of multitask Route Selection is divided to space tasks grade.Link shape is fully considered in network model
Task is divided into 7 classes by state parameter on demand, i.e.,：Time delay bandwidth packet loss sensitivity, only delay sensitive, only bandwidth sensitive, only packet loss
Rate sensitivity, time delay bandwidth sensitive, time delay packet loss are sensitive and bandwidth packet loss is sensitive.
At SDSIN, 7 kinds of tasks in above-mentioned network model are reduced to 4 kinds of typical space tasks, i.e.,：It is emergent to appoint
Business, communication task, number/figure pass task and TT＆C task, for 4 kinds of different tasks, according to its task feature, by task task
Divide 4 grades, i.e. Task={ Task1,Task2,Task3,Task4, above-mentioned 4 kinds of tasks are corresponded to respectively.
C, SDSIN multitask route selection method founding mathematical models are based on
Multitask routing problem based on SDSIN refers in available path collection P'={ p1,p2,…,pq}(1≤q≤
M) in, the QoS demand of link state parameter f and different spaces task is fully considered, how to be obtained optimal.Time delay and packet loss are all
With service quality in inverse ratio, that is, the smaller the better, and it is to be the bigger the better that bandwidth is proportional with service quality, so routing problem
It is：
Wherein,Value indicate in link pqUplink parameter fiPriority size, wiIndicate i-th of link
The preferred number of parameter, andR(Pq) indicate path bandwidth utilization rate.
Multitask route selection method uses a kind of method of machine learning, includes the following steps：
C1, communication source node v in SDSIN networks is determinedsWith destination node vd, for the available path P'=in network
{p1,p2,…,pq(1≤q≤m), according to supervised learning principle, obtain a large amount of training set D, be denoted as D={ (x1,y1),
(x2,y2),…,(xm,ym), wherein (xm,ym)What is indicated is the input and output result of m group training.
C2, the training set obtained for step C1, are trained training set, fresh information gain function Gain (D, f)
Numerical value, the priority of link state parameter is reselected, new W=(w are obtained1,w2,w3)T, as link state
The priority criteria of parameter.
C3, the newer W=(w of foundation step C21,w2,w3)TNumerical value is to link state parameter f={ f1(pq),f2(pq),f3
(pq) be updated, obtain optimized link state parameter attribute collection f*=arg max Gain (D, f)；
C4, step tri- steps of C1-C3 are repeated, continues to update weights W=(w1,w2,w3)T, until loss function E (d) ＜
When ε, iteration stopping, right value update terminates, and has selected optimal path.D ∈ D are the subsets of training sample.
D, link parameter decision
Route selection method based on multitask should carry out weight division, with letter to different link state parameters first
Gain function is ceased to reflect Attribute transposition.
It is as follows to define information gain function：
Assuming that the ratio in training sample D shared by kth class sample is pk(k=1,2 ..., | y |), then the comentropy of D isIt is assumed that having u possible value { f to link state parameter f1,f2,…,fu, use link
State parameter f divides training sample D, will produce u branch, wherein u-th of branch contain it is all in link shape in D
Value is f on state parameter fuSample, be denoted as Du。
Weight is assigned to branchInformation gain is expressed as：
Information gain is bigger, it is meant that it is better to be divided obtained effect using link state parameter f, is increased by information
The sizes values of benefit select optimized link state parameter f*=arg max Gain (D, f).Input data is pre-processed, is needed
Information gain function is transformed, be shown below：
Wherein, Gain (D, f, t) is the information gain after training sample D is divided based on division points t bis-.It is obtained according to information gain
Go out most suitable link state parameter f*=arg max Gain (D, f).
E, right value update
After the completion of training, suitable link parameter requirement is had adjusted according to information gain, is needed below to W=(w1,w2,
w3)TCarry out right value update.
E1, definition activation primitive are as follows：
It enables
WhereinInitialization W=(w first1,w2,w3)T,wi∈ (0,1], to input set
When being trained, one threshold θ of setting is needed to be set to 1 if F >=θ, otherwise is -1.Activation primitive characterization is under present weight
Whether correct categorization of perception can be exported.Weight update is expressed as：
Wherein,Y is the correct classification of input sample, and y' is calculated according to activation primitive
The classification come, η is learning rate, and 0.3 is taken in machine learning field.
E2, definition loss function are as follows：
Wherein d ∈ D are the subset of training sample, tdFor target reality output, odIt being exported for algorithm, n indicates iterations,
As loss function E (d) ＜ ε, iteration stopping, right value update terminates, and has obtained best initial weights W=(w1,w2,w3)T,wi∈(0,
1], according to different weight distributions, it is an optimal communication path to look for most suitable end-to-end path and combine.
Further, the SDSIN frameworks include applying plane, control plane and data Forwarding plane, and described answers
It is communicated by Internet ground networks by ground control centre with plane；The control plane is by three high rail satellite
GEO is formed, and control plane is transmitted by northbound interface agreement into row information with application plane；The data plane is by different
Local area network is constituted, and data plane and control plane carry out data interaction by OpenFlow agreements.
Further, the access that the space-based backbone network and user's satellite that the data plane is made of repeater satellite are constituted
Net composition.
Further, the local area network is loop network, and the loop network is by multiple middle rail satellite MEO and low
Rail satellite LEO is connected into a closed ring.
Further, the local area network is star network, and the star network is by multiple middle rail satellite MEO and low
Rail satellite LEO is constituted；One of satellite is in network center position, remaining satellite individually with the satellite communication at center.
Further, the local area network is mesh network, and the mesh network is by multiple middle rail satellite MEO and low
Rail satellite LEO is constituted, and the topology of mesh network is arbitrary, and each satellite regards Topology connection situation and other connected satellite nodes
It is communicated.
Compared with prior art, the invention has the advantages that：
1, the core technology application of the invention by software defined network (Software Defined Networking, SDN)
Onto Information Network.The core concept of software defined network is to be divided network-based control plane and data forwarding plane
From to simplify the structure of the network equipment, making satellite only need to implement simply to forward and hardware configuration function, thus solve
The drawback that satellite node design is complicated, cost is high.By the way that Information Network Satellite node resource and inter-satellite link are provided
Source carries out abstract, it can be achieved that the flexible of resource is shared, and the dynamic control of Information Network can be realized by the controller of centralization
System.By establishing rational mathematical model, under SDSIN frameworks, the characteristics of for satellite topology high dynamic cyclically-varying, adopt
Take the mode that topology controls to shield its high dynamic variation characteristic.By SDSIN network dynamic topologies according to system cycle T discretization
At n timeslice, it is denoted as T1,T2,…,Tn, topological novariable is thought in piece, finally each timeslice is routed, so
Routing iinformation is fed back into SDN controllers again afterwards, realizes the whole network information mutual communication, accelerates network deployment.
2, the present invention is based on SDSIN networks is tied the QoS of flow feature and space tasks using the method for machine learning
Altogether, in the case where network link capacity is limited, it is proposed that a kind of multitask route selection method based on SDSIN, this
Algorithm of the invention based on machine learning, has fully considered the Multiple factors such as link state parameter and space tasks, realizes optimal
Path selection improves the utilization rate of Internet resources, avoids local link and the state of congestion occurs, while effectively meeting
The QoS demand of all kinds of space tasks, the present invention fully consider the link characteristics such as propagation delay time, packet loss and remaining bandwidth, according to
Above three parameter is trained training set, updates link parameter weights, effectively collects multitask routing, improves
The utilization rate of underlying resource, realizes the load balancing of the whole network.
Description of the drawings
Fig. 1 is software definition Information Network schematic diagram.
Fig. 2 is to be based on SDSIN multitask route selection method flow charts.
Specific implementation mode
The present invention is further described through below in conjunction with the accompanying drawings.
The present invention has initially set up software definition Information Network framework, as shown in Figure 1, application software of the present invention defines
The network architecture is divided into using plane, control plane and data plane, wherein using plane by application plane control centrally through
Internet ground networks are communicated, and control plane is controlled with three GEO satellites, and control plane passes through with using plane
Northbound interface agreement is transmitted into row information, and data plane can be made of different local area network, such as loop network, Star network
Network and netted network, data plane and control plane carry out data interaction by OpenFlow agreements, by equalling satellite data
Face and control plane separation, make satellite only need to implement simply to forward, and in the deployment of actual network, data plane can be in
After the access net composition that space-based backbone network and user's satellite that satellite is constituted are constituted；Control plane high concentration, can be by management and control
The heart is unified to be carried.SDN controllers possess global view, can inquire link state and network flow etc. by southbound interface.Upper
It states under framework, it is abstract, it can be achieved that resource by carrying out Information Network Satellite node resource and inter-satellite link resource
It is flexibly shared, the dynamic control of Information Network can be realized by the controller of centralization；As shown in Fig. 2, being to be based on SDSIN
Multitask route selection method flow chart, the route selection method have fully considered that link state parameter and space tasks etc. are multiple
Factor realizes optimal route selection, improves the utilization rate of Internet resources, avoids local link and the state of congestion occurs,
The QoS demand for effectively meeting all kinds of space tasks simultaneously, realizes the load balancing of network.
The present invention is directed to current route selection method underaction, it is difficult to combine the QoS of flow feature and space tasks
The problem of getting up proposes the multitask route selection method based on SDSIN, this method fully considered link state parameter and
The Multiple factors such as space tasks, realize optimal route selection, improve the utilization rate of Internet resources, avoid local link and go out
The state of existing congestion, while the QoS demand of all kinds of space tasks is effectively met, realize the load balancing of network.Having
When body is implemented, the node of satellite network is first redefined according to the thought of SDN, establishes rational network model；Then according to reality
Border situation is by space tasks by grade classification at four classes；It is established according to the SDSIN frameworks selected and is route based on SDSIN multitasks
Selection method mathematical model designs the multitask route selection algorithm based on SDSIN, carries out decision for link parameter, in turn
Update weights, to adapt to the demand of satellite network, the algorithm fully considered link state parameter and space tasks etc. it is multiple because
Element realizes optimal route selection, improves the utilization rate of Internet resources, avoids local link and the state of congestion occurs, together
When effectively meet the QoS demands of all kinds of space tasks, realize the load balancing of network.
The present invention is not limited to the present embodiment, any equivalent concepts in the technical scope of present disclosure or changes
Become, is classified as protection scope of the present invention.
Claims (6)
1. a kind of multitask route selection method based on SDSIN, it is characterised in that：Include the following steps：
A, the node that satellite network is redefined according to the thought of SDN establishes rational network model
Under SDSIN frameworks, though satellite topology is that high dynamic changes, since satellite is run according to definitive orbit
, movement is that have predictability in periodically variable；For this feature, Route Selection takes the side that topology controls
Formula is to shield its high dynamic variation characteristic；SDSIN network dynamic topologies are discretized into n timeslice according to system cycle T, are remembered
For T1,T2,…,Tn, topological novariable is thought in timeslice, and finally each timeslice is routed；
A1, network model is defined
Wireless network topology is indicated with G (V, E, C), wherein V={ v1,v2,...,vmIndicate telephone net node set；E
={ E1,E2,…EnThe set that SDSIN Intersatellite Links form is represented, C indicates that the set of controller node, wherein controller are put
It sets on high rail satellite GEO；
A2, link connection matrix is defined
Software definition satellite network model is G (V, E, C), is by a series of high rail satellite GEO, middle rail satellite MEO, low orbit satellite
What LEO and inter-satellite link formed, link connection matrix is expressed as：
Wherein aijIt is expressed as node i and the link connection relationship of node j, wherein：
A3, link state parameter is defined
SDN controllers possess global view, and link state and network flow are inquired by southbound interface；For inter-satellite link subset
Eh,There are corresponding time delay, remaining bandwidth, packet loss information, uses d respectivelyij、bij、sijIt indicates；Road can be used
The collection of diameter shares P expressions, for any subpath piUnder ∈ P,
Time delay function representation is：
Remaining bandwidth function representation is：
Packet loss function representation is：
Link state parameter is expressed as：
F=[f1(pi),f2(pi),f3(pi)] (6)
A4, available path collection is defined
With (vs,vd) indicate network model in source node and destination node node pair, thenGeneration
One, table is from source node vsTo destination node vdPath；It is assumed that sharing m path set after mission dispatching, it is expressed as：P=
{P1,P2,…,Pm, wherein PiIndicate i-th available path collection；In network model G (V, E, C), available path is expressed as P'
={ p1,p2,…,pq, wherein 1≤q≤m；
B, space tasks grade classification
The premise of multitask Route Selection is divided to space tasks grade；Fully consider that link state is joined in network model
Task is divided into 7 classes by number on demand, i.e.,：Time delay bandwidth packet loss is sensitive, only delay sensitive, only bandwidth sensitive, only packet loss is quick
Sense, time delay bandwidth sensitive, time delay packet loss is sensitive and bandwidth packet loss is sensitive；
At SDSIN, 7 kinds of tasks in above-mentioned network model are reduced to 4 kinds of typical space tasks, i.e.,：Contingency tasks lead to
Trust business, number/figure biography task and TT＆C task, for 4 kinds of different tasks, according to its task feature, task task is divided 4
A grade, i.e. Task={ Task1,Task2,Task3,Task4, above-mentioned 4 kinds of tasks are corresponded to respectively；
C, SDSIN multitask route selection method founding mathematical models are based on
Multitask routing problem based on SDSIN refers in available path collection P'={ p1,p2,…,pq}(1≤q≤m)
In, it fully considers the QoS demand of link state parameter f and different spaces task, how to obtain optimal；Time delay and packet loss all with
Service quality is in inverse ratio, that is, the smaller the better, and it is to be the bigger the better that bandwidth is proportional with service quality, so routing problem is
It is：
Wherein,Value indicate in link pqUplink parameter fiPriority size, wiIndicate i-th of link parameter
Preferred number, andR(Pq) indicate path bandwidth utilization rate；
Multitask route selection method uses a kind of method of machine learning, includes the following steps：
C1, communication source node v in SDSIN networks is determinedsWith destination node vd, for the available path P'={ p in network1,
p2,…,pq(1≤q≤m), according to supervised learning principle, obtain a large amount of training set D, be denoted as D={ (x1,y1),(x2,
y2),…,(xm,ym), wherein (xm,ym) what is indicated is the input and output result of m group training；
C2, the training set obtained for step C1, are trained training set, the number of fresh information gain function Gain (D, f)
Value, reselects the priority of link state parameter, obtains new W=(w1,w2,w3)T, as link state parameter
Priority criteria；
C3, the newer W=(w of foundation step C21,w2,w3)TNumerical value is to link state parameter f={ f1(pq),f2(pq),f3(pq)}
It is updated, obtains optimized link state parameter attribute collection f*=argmaxGain (D, f)；
C4, step tri- steps of C1-C3 are repeated, continues to update weights W=(w1,w2,w3)T, when loss function E (d) ＜ ε,
Iteration stopping, right value update terminate, and have selected optimal path；D ∈ D are the subsets of training sample；
D, link parameter decision
Route selection method based on multitask should carry out weight division to different link state parameters first, be increased with information
Beneficial function reflects Attribute transposition；
It is as follows to define information gain function：
Assuming that the ratio in training sample D shared by kth class sample is pk(k=1,2 ..., | y |), then the comentropy of D isIt is assumed that having u possible value { f to link state parameter f1,f2,…,fu, use link
State parameter f divides training sample D, will produce u branch, wherein u-th of branch contain it is all in link shape in D
Value is f on state parameter fuSample, be denoted as Du；
Weight is assigned to branchInformation gain is expressed as：
Information gain is bigger, it is meant that it is better to be divided obtained effect using link state parameter f, passes through information gain
Sizes values select optimized link state parameter f*=argmaxGain (D, f)；Input data is pre-processed, is needed to letter
Breath gain function is transformed, and is shown below：
Wherein, Gain (D, f, t) is the information gain after training sample D is divided based on division points t bis-；It is obtained most according to information gain
Suitable link state parameter f*=argmaxGain (D, f)；
E, right value update
After the completion of training, suitable link parameter requirement is had adjusted according to information gain, is needed below to W=(w1,w2,w3)TInto
Row right value update；
E1, definition activation primitive are as follows：
It enables
WhereinInitialization W=(w first1,w2,w3)T,wi∈ (0,1], input set is carried out
When training, one threshold θ of setting is needed to be set to 1 if F >=θ, otherwise is -1；Activation primitive characterization be under present weight whether
Correct categorization of perception can be exported；Weight update is expressed as：
Wherein,Y is the correct classification of input sample, and y' is calculated according to activation primitive
Classification, η is learning rate, and 0.3 is taken in machine learning field；
E2, definition loss function are as follows：
Wherein d ∈ D are the subset of training sample, tdFor target reality output, odIt is exported for algorithm, n indicates iterations, works as damage
When losing function E (d) ＜ ε, iteration stopping, right value update terminates, and has obtained best initial weights W=(w1,w2,w3)T,wi∈ (0,1], root
According to different weight distributions, it is an optimal communication path to look for most suitable end-to-end path and combine.
2. a kind of multitask route selection method based on SDSIN according to claim 1, it is characterised in that：Described
SDSIN frameworks include that plane, control plane and data Forwarding plane, the application plane is applied to be passed through by ground control centre
Internet ground networks are communicated；The control plane is made of rail satellite GEO three high, and control plane is flat with application
It is transmitted into row information by northbound interface agreement in face；The data plane is made of different local area network, data plane and
Control plane carries out data interaction by OpenFlow agreements.
3. a kind of multitask route selection method based on SDSIN according to claim 2, it is characterised in that：Described
The access net that the space-based backbone network and user's satellite that data plane is made of repeater satellite are constituted forms.
4. a kind of multitask route selection method based on SDSIN according to claim 2, it is characterised in that：Described
Local area network is loop network, and the loop network is connected into a closing by multiple middle rail satellite MEO and low orbit satellite LEO
Ring.
5. a kind of multitask route selection method based on SDSIN according to claim 2, it is characterised in that：Described
Local area network is star network, and the star network is made of multiple middle rail satellite MEO and low orbit satellite LEO；One of them
Satellite is in network center position, remaining satellite individually with the satellite communication at center.
6. a kind of multitask route selection method based on SDSIN according to claim 2, it is characterised in that：Described
Local area network is mesh network, and the mesh network is made of multiple middle rail satellite MEO and low orbit satellite LEO, mesh network
Topology be arbitrary, each satellite is communicated depending on Topology connection situation with other connected satellite nodes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810083451.4A CN108307435B (en) | 2018-01-29 | 2018-01-29 | Multitask routing method based on SDSIN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810083451.4A CN108307435B (en) | 2018-01-29 | 2018-01-29 | Multitask routing method based on SDSIN |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108307435A true CN108307435A (en) | 2018-07-20 |
| CN108307435B CN108307435B (en) | 2021-02-19 |
Family
ID=62866999
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810083451.4A Active CN108307435B (en) | 2018-01-29 | 2018-01-29 | Multitask routing method based on SDSIN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108307435B (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108900419A (en) * | 2018-08-17 | 2018-11-27 | 北京邮电大学 | Route decision method and device based on deeply study under SDN framework |
| CN109039885A (en) * | 2018-08-27 | 2018-12-18 | 广州爱拍网络科技有限公司 | A kind of data transfer path selection method and device |
| CN109067668A (en) * | 2018-09-19 | 2018-12-21 | 苏州瑞立思科技有限公司 | Global network based on intelligent equalization distribution accelerates link construction method |
| CN110519146A (en) * | 2019-08-23 | 2019-11-29 | 北京邮电大学 | Service protecting method and controller based on air-ground integrated annular link structure |
| CN110730131A (en) * | 2019-10-22 | 2020-01-24 | 电子科技大学 | SDN satellite network multi-QoS constraint routing method based on improved ant colony |
| CN111064667A (en) * | 2019-12-04 | 2020-04-24 | 北京邮电大学 | Satellite network route optimization method, controller and data system |
| CN112019381A (en) * | 2020-08-12 | 2020-12-01 | 苏州浪潮智能科技有限公司 | Cluster link detection method and system based on deep learning |
| CN113572686A (en) * | 2021-07-19 | 2021-10-29 | 大连大学 | Heaven and earth integrated self-adaptive dynamic QoS routing method based on SDN |
| CN113923125A (en) * | 2020-06-22 | 2022-01-11 | 北京交通大学 | Tolerance analysis method and device for multi-service flow converged communication in industrial heterogeneous network |
| WO2022009005A1 (en) * | 2020-07-08 | 2022-01-13 | International Business Machines Corporation | Continual learning using cross connections |
| CN113938176A (en) * | 2021-08-26 | 2022-01-14 | 西安空间无线电技术研究所 | Low-delay service space-based computing method |
| CN114363243A (en) * | 2021-06-07 | 2022-04-15 | 中宇联云计算服务(上海)有限公司 | Backbone link optimization method, system and equipment based on cloud network fusion technology |
| CN114710200A (en) * | 2022-04-07 | 2022-07-05 | 中国科学院计算机网络信息中心 | Satellite network resource arrangement method and system based on reinforcement learning |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104579454A (en) * | 2015-01-17 | 2015-04-29 | 浙江大学 | Multi-objective optimization satellite flow control method based on software defined network |
| CN105897329A (en) * | 2016-06-08 | 2016-08-24 | 大连大学 | Multi-service routing optimization method of LEO satellite network based on multi-objective decisions |
| CN105959232A (en) * | 2016-06-16 | 2016-09-21 | 清华大学 | Satellite network routing method based on control point optimization of software-defined network |
| CN106059650A (en) * | 2016-05-24 | 2016-10-26 | 北京交通大学 | Air-ground integrated network architecture and data transmission method based on SDN and NFV technology |
| CN106059960A (en) * | 2016-05-24 | 2016-10-26 | 北京交通大学 | Software defined network-based space network QoS guarantee method and management center |
| CN107294592A (en) * | 2017-06-16 | 2017-10-24 | 大连大学 | A kind of satellite network and its construction method based on distributed SDN |
| US20170311228A1 (en) * | 2016-04-21 | 2017-10-26 | At&T Intellectual Property I, Lp. | Vehicle-based mobile node fleet for network service deployment |
-
2018
- 2018-01-29 CN CN201810083451.4A patent/CN108307435B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104579454A (en) * | 2015-01-17 | 2015-04-29 | 浙江大学 | Multi-objective optimization satellite flow control method based on software defined network |
| US20170311228A1 (en) * | 2016-04-21 | 2017-10-26 | At&T Intellectual Property I, Lp. | Vehicle-based mobile node fleet for network service deployment |
| CN106059650A (en) * | 2016-05-24 | 2016-10-26 | 北京交通大学 | Air-ground integrated network architecture and data transmission method based on SDN and NFV technology |
| CN106059960A (en) * | 2016-05-24 | 2016-10-26 | 北京交通大学 | Software defined network-based space network QoS guarantee method and management center |
| CN105897329A (en) * | 2016-06-08 | 2016-08-24 | 大连大学 | Multi-service routing optimization method of LEO satellite network based on multi-objective decisions |
| CN105959232A (en) * | 2016-06-16 | 2016-09-21 | 清华大学 | Satellite network routing method based on control point optimization of software-defined network |
| CN107294592A (en) * | 2017-06-16 | 2017-10-24 | 大连大学 | A kind of satellite network and its construction method based on distributed SDN |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108900419B (en) * | 2018-08-17 | 2020-04-17 | 北京邮电大学 | Routing decision method and device based on deep reinforcement learning under SDN framework |
| CN108900419A (en) * | 2018-08-17 | 2018-11-27 | 北京邮电大学 | Route decision method and device based on deeply study under SDN framework |
| CN109039885B (en) * | 2018-08-27 | 2022-01-07 | 广州猎游信息科技有限公司 | Data transmission path selection method and device |
| CN109039885A (en) * | 2018-08-27 | 2018-12-18 | 广州爱拍网络科技有限公司 | A kind of data transfer path selection method and device |
| CN109067668A (en) * | 2018-09-19 | 2018-12-21 | 苏州瑞立思科技有限公司 | Global network based on intelligent equalization distribution accelerates link construction method |
| CN109067668B (en) * | 2018-09-19 | 2022-03-04 | 苏州瑞立思科技有限公司 | Global network acceleration link construction method based on intelligent balanced distribution |
| CN110519146A (en) * | 2019-08-23 | 2019-11-29 | 北京邮电大学 | Service protecting method and controller based on air-ground integrated annular link structure |
| CN110730131A (en) * | 2019-10-22 | 2020-01-24 | 电子科技大学 | SDN satellite network multi-QoS constraint routing method based on improved ant colony |
| CN111064667A (en) * | 2019-12-04 | 2020-04-24 | 北京邮电大学 | Satellite network route optimization method, controller and data system |
| CN113923125A (en) * | 2020-06-22 | 2022-01-11 | 北京交通大学 | Tolerance analysis method and device for multi-service flow converged communication in industrial heterogeneous network |
| CN113923125B (en) * | 2020-06-22 | 2022-12-06 | 北京交通大学 | Tolerance analysis method and device for multi-service flow converged communication in industrial heterogeneous network |
| WO2022009005A1 (en) * | 2020-07-08 | 2022-01-13 | International Business Machines Corporation | Continual learning using cross connections |
| GB2611731A (en) * | 2020-07-08 | 2023-04-12 | Ibm | Continual learning using cross connections |
| CN112019381A (en) * | 2020-08-12 | 2020-12-01 | 苏州浪潮智能科技有限公司 | Cluster link detection method and system based on deep learning |
| CN114363243A (en) * | 2021-06-07 | 2022-04-15 | 中宇联云计算服务(上海)有限公司 | Backbone link optimization method, system and equipment based on cloud network fusion technology |
| CN114363243B (en) * | 2021-06-07 | 2024-04-12 | 中宇联云计算服务(上海)有限公司 | Backbone link optimization method, system and equipment based on cloud network fusion technology |
| CN113572686A (en) * | 2021-07-19 | 2021-10-29 | 大连大学 | Heaven and earth integrated self-adaptive dynamic QoS routing method based on SDN |
| CN113572686B (en) * | 2021-07-19 | 2022-09-02 | 大连大学 | Heaven and earth integrated self-adaptive dynamic QoS routing method based on SDN |
| CN113938176A (en) * | 2021-08-26 | 2022-01-14 | 西安空间无线电技术研究所 | Low-delay service space-based computing method |
| CN114710200A (en) * | 2022-04-07 | 2022-07-05 | 中国科学院计算机网络信息中心 | Satellite network resource arrangement method and system based on reinforcement learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108307435B (en) | 2021-02-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108307435A (en) | A kind of multitask route selection method based on SDSIN | |
| CN109714219B (en) | Virtual network function rapid mapping method based on satellite network | |
| Mao et al. | A novel non-supervised deep-learning-based network traffic control method for software defined wireless networks | |
| CN109981438B (en) | Satellite network load balancing method oriented to SDN and NFV collaborative deployment framework | |
| CN113572686B (en) | Heaven and earth integrated self-adaptive dynamic QoS routing method based on SDN | |
| CN109257091B (en) | Global load balancing satellite-ground cooperative network networking device and method | |
| CN107294592A (en) | A kind of satellite network and its construction method based on distributed SDN | |
| CN107276662A (en) | A kind of software definition Information Network multi-controller dynamic deployment method | |
| Wang et al. | Fuzzy-CNN based multi-task routing for integrated satellite-terrestrial networks | |
| CN105282038A (en) | Distributed asterism networking optimization method based on stability analysis and used in mobile satellite network | |
| Tang et al. | Federated learning for intelligent transmission with space-air-ground integrated network toward 6G | |
| CN108111335A (en) | A kind of method and system dispatched and link virtual network function | |
| Han et al. | Time-varying topology model for dynamic routing in LEO satellite constellation networks | |
| Etengu et al. | AI-assisted framework for green-routing and load balancing in hybrid software-defined networking: Proposal, challenges and future perspective | |
| CN106452555A (en) | Multi-path optimization algorithm planning method based on medium and low earth orbit satellite network | |
| CN114221691A (en) | Software-defined air-space-ground integrated network route optimization method based on deep reinforcement learning | |
| Qi et al. | SDN-based dynamic multi-path routing strategy for satellite networks | |
| Chen et al. | Time-varying resource graph based resource model for space-terrestrial integrated networks | |
| CN114268575A (en) | Self-adaptive three-dimensional transmission method and system in heaven-earth integrated information network | |
| CN113114335B (en) | Software-defined space-based network networking architecture based on artificial intelligence | |
| Han et al. | Space edge cloud enabling service migration for on-orbit service | |
| CN114024894B (en) | Dynamic calculation method and system in software-defined heaven-earth integrated network | |
| CN108429577A (en) | A kind of satellite QoS routing algorithms based on PROMETHEE methods | |
| CN115225512B (en) | Multi-domain service chain active reconfiguration mechanism based on node load prediction | |
| Wu et al. | QoS provisioning in space information networks: Applications, challenges, architectures, and solutions |
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 |