CN109831386A - Optimal route selection algorithm based on machine learning under a kind of SDN - Google Patents
Optimal route selection algorithm based on machine learning under a kind of SDN Download PDFInfo
- Publication number
- CN109831386A CN109831386A CN201910174856.3A CN201910174856A CN109831386A CN 109831386 A CN109831386 A CN 109831386A CN 201910174856 A CN201910174856 A CN 201910174856A CN 109831386 A CN109831386 A CN 109831386A
- Authority
- CN
- China
- Prior art keywords
- network
- data
- packet loss
- node
- indicate
- 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
Abstract
Optimal route selection algorithm based on machine learning under a kind of SDN, build SDN platform, simulate true network environment, acquire discrete real-time network status data, classified according to the difference that different business in network transmission requires QoS index, and arrange experimental data, obtain sample data set, sample data set is screened into optimal path using heuritic approach according to consider standard of the different business to each index, and it labels for the corresponding optimal path of every group of data, finally use machine learning algorithm training dataset, obtain classifier, achieve the purpose that quick dynamic routing.The optimum results of result and heuritic approach of the invention are essentially identical, and the calculating time of model will be far smaller than heuritic approach, to meet the necessary condition of high-speed decision in real network operation.Compared with particle swarm algorithm, the cpu runing time needed for calculating is greatly shortened extreme learning machine algorithm of the invention, can satisfy live network deployment request completely.
Description
Technical field
The present invention relates to the general fashion method for routing under software defined network framework can generate higher time cost
Problem, the QoS route planning method of the multiple constraint consumed when proposing a kind of low, and in particular to machine learning is based under a kind of SDN
Optimal route selection algorithm.
Background technique
According to the 42nd time " China Internet network state of development statistical report " display of China Internet Network Information Center publication:
By in June, 2018, China's netizen's scale has reached 8.02 hundred million, moves end subscriber up to 7.88 hundred million.And with information technology
The emerging technologies such as high speed development and cloud computing, big data continue to bring out, and network data is no matter from scale or in type
Explosive growth is all presented.Ubiquitous network insertion, big bandwidth cause the dynamic management to network more important.Together
When, huge variation also has occurred in Internet service feature and demand for services at present, i.e., from original single point-to-point transmission side
Formula is gradually evolved into the communication pattern for supporting multiple services type and high quality to require, such as web browsing, e-commerce, video council
The multibusiness networks demands such as view.In order to achieve this goal, Internet provider or network operation manager have to basis
A variety of different discharge patterns provide different services, such as meet to realize high real-time in data center network
Voice service, videophone etc.；Congestion is reduced, as far as possible to keep the communication of the business such as E-mail, SMS, multi-media SMS just
Often.Software defined network (Software-DefinedNetwork, SDN) had become a kind of very attractive in recent years
Solution, pursued by people.There are two important characteristics for software defined network: one is that control plane and data are flat
It is separated from each other between face, the other is network has editability.Therefore, significantly more efficient configuration can be improved in SDN, preferably
Performance and higher flexibility more adapt to the demand for development of future network.At present about routing rule under SDN multi-constraint condition
The problem of drawing has been found to be a np problem, and in such way to solve the problem, accurate mathematical algorithm or cannot obtain
It can only work normally to reliable result or in small-sized networked environments or work as one relatively large network size of consideration
When, such algorithm seems helpless.Therefore up to the present, many research work all concentrate on the research of approximate algorithm
In, it obtains a kind of solution as excellent as possible and replaces to obtain optimal solution, and propose the heuritic approach of some classics, for effective
Managing network flow and balance network environment, wherein including the research of routing algorithm under a variety of SDN environment, although opening at present
The significant effect of hairdo algorithm has obtained approximate optimal solution, but there are critical defect, process of the heuritic approach in operation
It is middle to need to consume the more time, the selection of dynamic routing can not be completed in live network required time.
Summary of the invention
It is an object of the invention to significantly reduce for the different demands aspect for meeting multiple QoS indexes in different business
Algorithm when consume, solve the problems, such as that existing algorithm introduces higher time cost, propose under a kind of SDN based on machine learning
Optimal route selection algorithm can solve the quick dynamic routing of different business stream.
To achieve the above object, the present invention adopts the following technical scheme that:
Optimal route selection algorithm based on machine learning under a kind of SDN, comprising the following steps:
The first step builds software defined network platform, simulates true network environment, constructs network topology, and acquisition is real-time
Network status data forms network status data collection；
Second step, for business different in network, according to its different demands to time delay, shake and packet loss, by industry
Business is divided into four major class, respectively session service, stream class service, interactive class, background classes；
Third step, the data concentrated to network status data pre-process, define time delay rate using analytic hierarchy process (AHP), lose
Packet ratio indicator eliminates the dimension impact between former data, and sample data set is calculated using Euclidean distance formula；
Sample data sets X is used particle swarm algorithm according to consider standard of the different business to each index by the 4th step
Optimal path is screened, and is labelled for the corresponding optimal path of every group of data, path label data set is formed；
5th step, using extreme learning machine algorithm training raw data set and path label data set, in extreme learning machine
After algorithmic statement, classification data is treated using trained model and is classified, reach through street by dynamic purpose.
A further improvement of the present invention lies in that the middle controller of building of software defined network platform is selected in the first step
Floodlight, opendaylight, ryu controller or onos controller build fat tree-network topology, without hindrance complete are connected
Network topology simulates real network environment.
A further improvement of the present invention lies in that in the first step, for real-time network status data, according to the identical time
Interval acquires discrete data set, and every single sample that discrete data is concentrated is denoted as: sk, k ∈ 1,2,3...m, in each sample
Each of the links be denoted as xij, i ∈ (1,2,3...n), j ∈ (1,2,3...), n are terminal number, xijExpression source interchanger is to mesh
Interchanger end-to-end time delay, shake and packet loss data, be successively denoted as xi1,xi2,xi3..., can in sample data
It is expressed as the form of matrix.
A further improvement of the present invention lies in that the data that network status data is concentrated are pre-processed, are adopted in third step
With analytic hierarchy process (AHP) by each of the links x in each sampleijImportant indicator processing handle to obtain pair according to different business requirement
Sample data set X is calculated using range formula further according to weight in the weight answeredij, wherein (1,2,3...n) i ∈, j ∈ (1,
2,3...m), i.e. Xi=(xi1,xi2,...xim), n represents number of samples in experiment, and m represents of interchanger in network environment
Number, sample data sets X are indicated are as follows:
A further improvement of the present invention lies in that sample data sets X is obtained especially by following procedure:
Step 1: initial data pretreatment:
Standardization: initial data bi-directional scaling is allowed to fall into the specific region of one [0,1]；Remove initial data
Unit limitation, be translated into dimensionless pure values；
Step 2: obtaining experimental data
The dimensionless number that step 1 obtains is obtained into sample data sets X according to by Euclidean distance calculation formula.
A further improvement of the present invention lies in that detailed process is as follows for step 1:
(1) bandwidth availability ratio ηij, it is defined as current ink node v in networkiTo node vjThe link band used
Width accounts for the ratio of link maximum bandwidth in system, bandwidth availability ratio ηijIt is calculated by following formula:
Here, loadijIndicate current ink node viTo node vjThe link bandwidth used, loadklIndicate current
Link nodes vkTo node vlThe link bandwidth used,Indicate the maximum of chain road under current network state
Bandwidth capacity, i.e. peak transfer rate,Indicate the minimum bandwidth capacity of chain road under current network state, i.e., most
Small transmission rate；
(2) time delay rate trij, it is defined as current ink node v in networkiTo link nodes vjPropagation delay time and current net
Under network state in link maximum delay ratio, time delay rate trijIt is calculated by following formula:
Here,Indicate the chain road maximum delay in current network state lower network；Expression is worked as
Chain road minimal time delay in preceding network state lower network, tdijIndicate link nodes viTo link nodes vjTime delay, tdklTable
Show link nodes vkTo link nodes vlTime delay；
(3) packet loss ratio lrij, it is defined as current ink node v in networkiTo node vjWhen transmission, the chain road
The ratio of the difference of the difference and minimax packet loss of minimum packet loss in packet loss and current network environment；Packet loss ratio lrijIt is logical
Following formula is crossed to calculate:
Here,Indicate the chain road maximum delay in current network state lower network；It indicates
Chain road minimal time delay in current network state lower network, lossijIndicate viTo vjPacket loss；lossklIndicate vkTo vl's
Packet loss；
Use the bandwidth availability ratio η in above-mentioned formulaij, time delay rate trijAnd packet loss ratio lrijIt replaces in initial data
Bandwidth, time delay, packet loss, eliminate initial data between unit limitation, convert dimensionless number evidence for initial data.
A further improvement of the present invention lies in that the detailed process of the 4th step are as follows: open obtained sample data sets X use
Hairdo algorithm screens optimal path, runs heuritic approach repeatedly, obtains the optimal road for meeting QoS index under each business need
Diameter；Due to the diversity of network state, under identical network state, there can be a plurality of optimal path, due to being all optimal path, only
It is passed through link nodes sequence difference, so after same node point, the link conduct of minimum node in all optimal paths
Optimal path.
A further improvement of the present invention lies in that the detailed process of the 5th step are as follows: every group of sample in path label data set
Data correspond to an optimal path, then label for the corresponding optimal path of all sample datas, make to become single between them
Mapping relations obtain the data set of one group of tape label, there is specific label in the data set in each data, this data set is drawn
It is divided into training set and test set；Then supervised learning model is used, this training set of operating limit learning machine algorithm training passes through
Constantly prediction result is compared with actual result, the parameters for adjusting model reach scheduled accuracy rate；It reuses
The data of test set make further amendment to the parameter of supervised learning model, after model convergence, use trained mould
Type treats classification data and classifies, and reaches through street by dynamic purpose.
Compared with prior art, the device have the advantages that it is as follows:
Firstly, building SDN platform, true network environment is simulated, acquires discrete real-time network status data.Net herein
Under network environment, classified according to the difference that different business in network transmission requires QoS index.It is arranged according to classificating requirement real
Data are tested, weight is calculated and obtains sample data set.Secondly, sample data set is considered mark to each index according to different business
Standard screens optimal path using heuritic approach, and labels for the corresponding optimal path of every group of data.Finally use engineering
Algorithm training dataset is practised, classifier is obtained, to achieve the purpose that quick dynamic routing.
Due to building software defined network platform, true network environment is simulated, network topology is constructed, acquires real-time network
Status data has so data are true and reliable and convinces power.
The data set obtained according to type of service removes the unit limitation of data in standardisation process, is translated into nothing
Dimension pure values, the influence of unit between each index avoided.
The data set put in order is screened most according to consider standard of the different business to each index using heuritic approach
Shortest path, and label is accomplished fluently for the corresponding optimal path of every group of data, using machine learning algorithm training dataset, classified
Device, to achieve the purpose that quick dynamic routing.Heuritic approach carries out the significant effect of route planning at present, has obtained approximate
Optimal solution more convinces power in the searching process of path, then uses machine learning algorithm training dataset, obtains classifier,
To achieve the purpose that quick dynamic routing.The basic phase of optimum results of the result that method of the invention provides and heuritic approach
Together, but the calculating time of model will be far smaller than heuritic approach, thus meet real network operation in high-speed decision must
Want condition.Cpu and in same hardware environment limit inferior learning machine algorithm compared with particle swarm algorithm, needed for calculating
Runing time greatly shortens, and can satisfy live network deployment request completely.
Further, since the requirement difference according to different business to QoS index is classified, so that business has more needle
To property.
Detailed description of the invention
Fig. 1 is the detail flowchart of the invention that optimal path is solved using machine learning.
Specific embodiment
It elaborates with reference to the accompanying drawings and examples to the present invention, but protection scope of the present invention is not limited to institute
Embodiment.
Referring to Fig. 1, the invention proposes the optimal route selection algorithm under a kind of SDN based on machine learning, the algorithm packets
Include following steps:
The first step builds software defined network platform, simulates true network environment, constructs network topology, and acquisition is real-time
Network status data forms network status data collection；
The selection for building middle controller of software defined network platform, such as: floodlight, opendaylight,
Ryu, onos etc., the present invention in use floodlight controller, build fat tree-network topology, without hindrance full connected network is opened up
It the simulation real network environment such as flutters, acquires real-time network status data.
For real-time network status data, discrete data set, discrete data set are acquired according to identical time interval
In every single sample be denoted as: sk(k ∈ 1,2,3...m), each of the links in each sample are denoted as xij, i ∈ (1,2,
3...n), (1,2,3...) j ∈, n are terminal number, xijExpression source interchanger to purpose interchanger end-to-end time delay, shake
And the data such as packet loss, successively it is denoted as xi1,xi2,xi3..., it can be expressed as the form of matrix in sample data, for example, the
One sample s1It indicates are as follows:
Second step, for business different in network, according to its different demands to time delay, shake and packet loss, by industry
Business is divided into four major class, is respectively as follows: session service, stream class service, interactive class, background classes, specific classification method such as the following table 1 institute
Show:
The specific classification method of 1 business of table
Since transmittability of the different business to network has different requirements, i.e. the index of the QoS of business sensitivity is different,
By according to the real-time of business, business is divided into substantially by congestion situation in transmission process, the response time etc. when data interaction
It is divided into four major class.
Third step, the data concentrated to network status data pre-process, and consider that the influence degree of each index is different,
Reference analytic hierarchy process (AHP) defines the dimension impact between the former data of the indexs such as time delay rate, packet loss ratio elimination, utilizes Euclidean distance
Sample data set is calculated in formula.
Specifically, since each index is different for influence degree of the source mesh node in data transmission procedure, it will be original
Collected data are pre-processed, using analytic hierarchy process (AHP) by each of the links x in each sampleijImportant indicator handle root
It handles to obtain corresponding weight according to different business requirement, sample data set X is calculated using range formula further according to weightij,
Wherein (1,2,3...n) i ∈, j ∈ (1,2,3...m), i.e. Xi=(xi1,xi2,...xim), n represents number of samples in experiment, m
The number of interchanger in network environment is represented, sample data sets X is indicated are as follows:
Sample data sets X is obtained especially by following procedure:
Step 1: initial data pretreatment:
Standardization: initial data bi-directional scaling is allowed to fall into the specific region of one [0,1].Remove initial data
Unit limitation, be translated into dimensionless pure values, be able to carry out convenient for not commensurate or magnitude index and compare weighting.Specifically
Process is as follows:
(1) bandwidth availability ratio ηij, it is defined as current ink node v in networkiTo node vjThe link band used
Width accounts for the ratio of link maximum bandwidth in system, bandwidth availability ratio ηijIt is calculated by following formula:
Here, loadijIndicate current ink node viTo node vjThe link bandwidth used, loadklIndicate current
Link nodes vkTo node vlThe link bandwidth used,Indicate the maximum of chain road under current network state
Bandwidth capacity, i.e. peak transfer rate,Indicate the minimum bandwidth capacity of chain road under current network state, i.e., most
Small transmission rate.ηijAlthough expression formula it is simple, can intuitively react very much the load of each of the links.
(2) time delay rate trij, it is defined as current ink node v in networkiTo link nodes vjPropagation delay time and current net
Under network state in link maximum delay ratio, time delay rate trijIt is calculated by following formula:
Here,Indicate the chain road maximum delay in current network state lower network.Expression is worked as
Chain road minimal time delay in preceding network state lower network, tdijIndicate link nodes viTo link nodes vjTime delay, tdklTable
Show link nodes vkTo link nodes vlTime delay.
(3) packet loss ratio lrij, it is defined as current ink node v in networkiTo node vjWhen transmission, the chain road
The ratio of the difference of the difference and minimax packet loss of minimum packet loss in packet loss and current network environment.Packet loss ratio lrijIt is logical
Following formula is crossed to calculate:
Here,Indicate the chain road maximum delay in current network state lower network.It indicates
Chain road minimal time delay in current network state lower network, lossijIndicate viTo vjPacket loss.lossklIndicate vkTo vl's
Packet loss.
Use the bandwidth availability ratio η in formula defined aboveij, time delay rate trijAnd packet loss ratio lrijIt replaces original
Bandwidth, time delay, packet loss in data etc. eliminate the unit limitation between initial data, and by initial data, it is converted into immeasurable
Guiding principle data are compared weighting convenient for it.
Step 2: obtaining experimental data
By dimensionless number that step 1 obtains according to using Euclidean distance calculation formula to obtain sample data sets X.
Sample data sets X is used particle swarm algorithm according to consider standard of the different business to each index by the 4th step
Optimal path is screened, and is labelled for the corresponding optimal path of every group of data.
Obtained sample data sets X is screened into optimal path using heuritic approach, heuritic approach is run repeatedly, obtains
Meet the optimal path of QoS index under to each business need.Due to the diversity of network state, under identical network state, meeting
There are a plurality of optimal path, processing method: due to being all optimal path, only passed through link nodes sequence is different, advises herein
Determine after same node point, the link of minimum node is as optimal path in all optimal paths.
5th step is made after extreme learning machine algorithmic statement using extreme learning machine algorithm training path label data set
Classification data is treated with trained model to classify, and reaches through street by dynamic purpose.
Every group of sample data will correspond to an optimal path, be then the corresponding optimal path mark of all sample datas
Label make to become single mapping relations between them.The data set of one group of tape label will be obtained, is had in each data in the data set
This data set is divided into training set and test set by specific label.Then supervised learning model, this data of training are used
Collection, by being constantly compared prediction result with actual result, the parameters for adjusting model reach scheduled accuracy rate；
The data for reusing test set make further amendment to the parameter of model, after model convergence, can use trained mould
Type classifies to no label data (data i.e. to be sorted).It, can be more i.e. when controller receives new transmission request
Have a collected real-time network data, it is independent calculate with the approximate path optimizing of heuritic approach, then this is routed through
Journey is very fast, meets the needs of true network deployment.
In the present invention after extreme learning machine algorithmic statement, it can guarantee effectively to replace heuristic particle swarm algorithm, reach
Weaken the time-consuming problem solved during optimal path.
The present invention is to build SDN network platform based on implementing under SDN network architecture environment, simulate real network environment,
Network topology is constructed, discrete network status data is acquired, data is pre-processed, remove the influence between different dimensions,
Analytic hierarchy process (AHP) thought is used for reference, weight is determined, comprehensively considers influence of a variety of QoS indexes to optimal path, establish experiment sample
Data set.Route planning is carried out to entire sample data set using heuritic approach, obtains sample routing database, wherein sample
Data set and sample routing database are single mapping relations, to guarantee the uniqueness of label.Arrange sample data set and routing number
According to collection, k parts are divided into, as the data set of machine learning, is then operated in machine learning algorithm newly using k folding cross validation
Data set above, carry out routing decision, have what particle swarm algorithm was routed according to current network state calculation optimization to reach
Ability.The optimum results of result and heuritic approach that method of the invention provides are essentially identical, but the calculating time of model
It is far smaller than heuritic approach, to meet the necessary condition of high-speed decision in real network operation.To under SDN be based on machine
The optimal route selection model of study has done simple authentication, and effect is as shown in table 2 below.
2 Riming time of algorithm of table compares
Table 2 shows that under same hardware environment, particle swarm algorithm and extreme learning machine algorithm are for same current
Network state calculates the average cpu runing time that is spent of optimal path, as can be seen from the table, extreme learning machine algorithm with
Particle swarm algorithm is compared, and the cpu runing time needed for calculating greatly shortens, and can satisfy live network deployment request completely.
Claims (8)
1. the optimal route selection algorithm under a kind of SDN based on machine learning, which comprises the following steps:
The first step builds software defined network platform, simulates true network environment, constructs network topology, acquires real-time network
Status data forms network status data collection；
Second step, for business different in network, according to its different demands to time delay, shake and packet loss, by business point
For four major class, respectively session service, stream class service, interactive class, background classes；
Third step, the data concentrated to network status data pre-process, and define time delay rate, packet loss ratio using analytic hierarchy process (AHP)
Rate index eliminates the dimension impact between former data, and sample data set is calculated using Euclidean distance formula；
4th step screens sample data sets X according to standard of considering of the different business to each index using particle swarm algorithm
Optimal path, and label for the corresponding optimal path of every group of data, form path label data set；
5th step, using extreme learning machine algorithm training raw data set and path label data set, in extreme learning machine algorithm
After convergence, classification data is treated using trained model and is classified, reach through street by dynamic purpose.
2. the optimal route selection algorithm under a kind of SDN according to claim 1 based on machine learning, which is characterized in that
In the first step, the middle controller of building of software defined network platform selects floodlight, opendaylight, ryu controller
Or onos controller, build fat tree-network topology, without hindrance full connected network topology simulation real network environment.
3. the optimal route selection algorithm under a kind of SDN according to claim 2 based on machine learning, which is characterized in that
In the first step, for real-time network status data, discrete data set, discrete data set are acquired according to identical time interval
In every single sample be denoted as: sk, k ∈ 1,2,3...m, each of the links in each sample are denoted as xij, i ∈ (1,2,3...n),
J ∈ (1,2,3...), n are terminal number, xijExpression source interchanger is to the end-to-end time delay of purpose interchanger, shake and packet loss
Rate data, are successively denoted as xi1,xi2,xi3..., it can be expressed as the form of matrix in sample data.
4. the optimal route selection algorithm under a kind of SDN according to claim 1 based on machine learning, which is characterized in that
In third step, the data that network status data is concentrated are pre-processed, using analytic hierarchy process (AHP) by every in each sample
Link xijImportant indicator processing handled to obtain corresponding weight according to different business requirement, further according to weight utilize range formula
Sample data set X is calculatedij, wherein (1,2,3...n) i ∈, j ∈ (1,2,3...m), i.e. Xi=(xi1,xi2,...xim), n
Number of samples in experiment is represented, m represents the number of interchanger in network environment, and sample data sets X is indicated are as follows:
5. the optimal route selection algorithm under a kind of SDN according to claim 4 based on machine learning, which is characterized in that
Sample data sets X is obtained especially by following procedure:
Step 1: initial data pretreatment:
Standardization: initial data bi-directional scaling is allowed to fall into the specific region of one [0,1]；Remove the list of initial data
Position limitation, is translated into dimensionless pure values；
Step 2: obtaining experimental data
The dimensionless number that step 1 obtains is obtained into sample data sets X according to by Euclidean distance calculation formula.
6. the optimal route selection algorithm under a kind of SDN according to claim 5 based on machine learning, which is characterized in that
Detailed process is as follows for step 1:
(1) bandwidth availability ratio ηij, it is defined as current ink node v in networkiTo node vjThe link bandwidth used, which accounts for, is
The ratio of link maximum bandwidth in system, bandwidth availability ratio ηijIt is calculated by following formula:
Here, loadijIndicate current ink node viTo node vjThe link bandwidth used, loadklIndicate current ink
Node vkTo node vlThe link bandwidth used,Indicate the maximum bandwidth of chain road under current network state
Capacity, i.e. peak transfer rate,Indicate the minimum bandwidth capacity of chain road under current network state, i.e. most brief biography
Defeated rate；
(2) time delay rate trij, it is defined as current ink node v in networkiTo link nodes vjPropagation delay time and current network shape
Under state in link maximum delay ratio, time delay rate trijIt is calculated by following formula:
Here,Indicate the chain road maximum delay in current network state lower network；Indicate current net
Chain road minimal time delay in network state lower network, tdijIndicate link nodes viTo link nodes vjTime delay, tdklIndicate chain
Circuit node vkTo link nodes vlTime delay；
(3) packet loss ratio lrij, it is defined as current ink node v in networkiTo node vjWhen transmission, the packet loss of the chain road
With the ratio of the difference of the difference and minimax packet loss of packet loss minimum in current network environment；Packet loss ratio lrijBy as follows
Formula calculates:
Here,Indicate the chain road maximum delay in current network state lower network；Indicate current
Chain road minimal time delay in network state lower network, lossijIndicate viTo vjPacket loss；lossklIndicate vkTo vlPacket loss
Rate；
Use the bandwidth availability ratio η in above-mentioned formulaij, time delay rate trijAnd packet loss ratio lrijReplace the band in initial data
Width, time delay, packet loss eliminate the unit limitation between initial data, convert dimensionless number evidence for initial data.
7. the optimal route selection algorithm under a kind of SDN according to claim 4 based on machine learning, which is characterized in that
The detailed process of 4th step are as follows: obtained sample data sets X is screened into optimal path using heuritic approach, operation is opened repeatedly
Hairdo algorithm obtains the optimal path for meeting QoS index under each business need；Due to the diversity of network state, identical net
Under network state, there can be a plurality of optimal path, due to being all optimal path, only passed through link nodes sequence is different, so
After same node point, the link of minimum node is as optimal path in all optimal paths.
8. the optimal route selection algorithm under a kind of SDN according to claim 1 or claim 7 based on machine learning, feature exist
In the detailed process of the 5th step are as follows: the corresponding optimal path of every group of sample data in path label data set, is then institute
There is the corresponding optimal path of sample data to label, makes to become single mapping relations between them, obtain the data of one group of tape label
Collect, has specific label in each data in the data set, this data set is divided into training set and test set；Then using prison
Superintend and direct formula learning model, this training set of operating limit learning machine algorithm training, by constantly by prediction result and actual result into
Row compares, and the parameters for adjusting model reach scheduled accuracy rate；The data of test set are reused to supervised learning model
Parameter make further amendment, until model convergence after, treat classification data using trained model and classify, reach fast
Speed routes dynamic purpose.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910174856.3A CN109831386B (en) | 2019-03-08 | 2019-03-08 | Optimal path selection algorithm based on machine learning under SDN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910174856.3A CN109831386B (en) | 2019-03-08 | 2019-03-08 | Optimal path selection algorithm based on machine learning under SDN |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109831386A true CN109831386A (en) | 2019-05-31 |
| CN109831386B CN109831386B (en) | 2020-07-28 |
Family
ID=66865655
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910174856.3A Active CN109831386B (en) | 2019-03-08 | 2019-03-08 | Optimal path selection algorithm based on machine learning under SDN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109831386B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110287124A (en) * | 2019-07-03 | 2019-09-27 | 大连海事大学 | A kind of automatic marker software error reporting simultaneously carries out seriousness and knows method for distinguishing |
| CN110493068A (en) * | 2019-09-05 | 2019-11-22 | 烽火通信科技股份有限公司 | A kind of network route generating method and system |
| CN110888744A (en) * | 2019-11-29 | 2020-03-17 | 杭州电子科技大学 | Load balancing method based on automatic adjustment and optimization of workload |
| CN111083051A (en) * | 2019-12-20 | 2020-04-28 | 北京邮电大学 | Path planning method and device based on multiple intelligent agents and electronic equipment |
| CN111447114A (en) * | 2020-03-25 | 2020-07-24 | 深信服科技股份有限公司 | Data packet transmission method and device, electronic equipment and storage medium |
| CN112311670A (en) * | 2019-12-04 | 2021-02-02 | 重庆邮电大学 | Machine learning route optimization method for software defined network |
| CN112671648A (en) * | 2020-12-22 | 2021-04-16 | 北京浪潮数据技术有限公司 | SDN data transmission method, SDN, device and medium |
| US11368401B1 (en) | 2021-01-20 | 2022-06-21 | Cisco Technology, Inc. | Globally avoiding simultaneous reroutes in a network |
| CN114884879A (en) * | 2022-05-06 | 2022-08-09 | 鹏城实验室 | Software defined network hybrid switching method based on label technology |
| CN114938530A (en) * | 2022-06-10 | 2022-08-23 | 电子科技大学 | Wireless ad hoc network intelligent networking method based on deep reinforcement learning |
| CN116578924A (en) * | 2023-07-12 | 2023-08-11 | 太极计算机股份有限公司 | Network task optimization method and system for machine learning classification |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529667A (en) * | 2016-09-23 | 2017-03-22 | 中国石油大学(华东) | Logging facies identification and analysis method based on fuzzy depth learning in big data environment |
| US20170126314A1 (en) * | 2014-06-13 | 2017-05-04 | Ciena Corporation | Systems and methods for statistical multiplexing with otn and dwdm |
| CN106953802A (en) * | 2017-03-01 | 2017-07-14 | 浙江工商大学 | A kind of network optimal route selection method based on deep learning |
| CN108259367A (en) * | 2018-01-11 | 2018-07-06 | 重庆邮电大学 | A kind of Flow Policy method for customizing of the service-aware based on software defined network |
| CN108418756A (en) * | 2018-01-23 | 2018-08-17 | 南京邮电大学 | A kind of software definition backhaul network access selection method based on similarity measurement |
| CN108667734A (en) * | 2018-05-18 | 2018-10-16 | 南京邮电大学 | It is a kind of that the through street with LSTM neural networks is learnt by decision making algorithm based on Q |
| CN108923975A (en) * | 2018-07-05 | 2018-11-30 | 中山大学 | A kind of traffic behavior analysis method of Based on Distributed network |
-
2019
- 2019-03-08 CN CN201910174856.3A patent/CN109831386B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170126314A1 (en) * | 2014-06-13 | 2017-05-04 | Ciena Corporation | Systems and methods for statistical multiplexing with otn and dwdm |
| CN106529667A (en) * | 2016-09-23 | 2017-03-22 | 中国石油大学(华东) | Logging facies identification and analysis method based on fuzzy depth learning in big data environment |
| CN106953802A (en) * | 2017-03-01 | 2017-07-14 | 浙江工商大学 | A kind of network optimal route selection method based on deep learning |
| CN108259367A (en) * | 2018-01-11 | 2018-07-06 | 重庆邮电大学 | A kind of Flow Policy method for customizing of the service-aware based on software defined network |
| CN108418756A (en) * | 2018-01-23 | 2018-08-17 | 南京邮电大学 | A kind of software definition backhaul network access selection method based on similarity measurement |
| CN108667734A (en) * | 2018-05-18 | 2018-10-16 | 南京邮电大学 | It is a kind of that the through street with LSTM neural networks is learnt by decision making algorithm based on Q |
| CN108923975A (en) * | 2018-07-05 | 2018-11-30 | 中山大学 | A kind of traffic behavior analysis method of Based on Distributed network |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110287124A (en) * | 2019-07-03 | 2019-09-27 | 大连海事大学 | A kind of automatic marker software error reporting simultaneously carries out seriousness and knows method for distinguishing |
| CN110287124B (en) * | 2019-07-03 | 2023-04-25 | 大连海事大学 | Method for automatically marking software error report and carrying out severity identification |
| CN110493068B (en) * | 2019-09-05 | 2023-01-10 | 烽火通信科技股份有限公司 | Network route generation method and system |
| CN110493068A (en) * | 2019-09-05 | 2019-11-22 | 烽火通信科技股份有限公司 | A kind of network route generating method and system |
| CN110888744A (en) * | 2019-11-29 | 2020-03-17 | 杭州电子科技大学 | Load balancing method based on automatic adjustment and optimization of workload |
| CN112311670A (en) * | 2019-12-04 | 2021-02-02 | 重庆邮电大学 | Machine learning route optimization method for software defined network |
| CN112311670B (en) * | 2019-12-04 | 2024-01-26 | 重庆邮电大学 | Software defined network machine learning route optimization method |
| CN111083051A (en) * | 2019-12-20 | 2020-04-28 | 北京邮电大学 | Path planning method and device based on multiple intelligent agents and electronic equipment |
| CN111447114A (en) * | 2020-03-25 | 2020-07-24 | 深信服科技股份有限公司 | Data packet transmission method and device, electronic equipment and storage medium |
| CN112671648A (en) * | 2020-12-22 | 2021-04-16 | 北京浪潮数据技术有限公司 | SDN data transmission method, SDN, device and medium |
| US11368401B1 (en) | 2021-01-20 | 2022-06-21 | Cisco Technology, Inc. | Globally avoiding simultaneous reroutes in a network |
| US11777854B2 (en) | 2021-01-20 | 2023-10-03 | Cisco Technology, Inc. | Globally avoiding simultaneous reroutes in a network |
| CN114884879A (en) * | 2022-05-06 | 2022-08-09 | 鹏城实验室 | Software defined network hybrid switching method based on label technology |
| CN114884879B (en) * | 2022-05-06 | 2023-07-14 | 鹏城实验室 | Software defined network hybrid switching method based on label technology |
| CN114938530A (en) * | 2022-06-10 | 2022-08-23 | 电子科技大学 | Wireless ad hoc network intelligent networking method based on deep reinforcement learning |
| CN114938530B (en) * | 2022-06-10 | 2023-03-21 | 电子科技大学 | Wireless ad hoc network intelligent networking method based on deep reinforcement learning |
| CN116578924A (en) * | 2023-07-12 | 2023-08-11 | 太极计算机股份有限公司 | Network task optimization method and system for machine learning classification |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109831386B (en) | 2020-07-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109831386A (en) | Optimal route selection algorithm based on machine learning under a kind of SDN | |
| Mao et al. | A novel non-supervised deep-learning-based network traffic control method for software defined wireless networks | |
| CN108259367B (en) | Service-aware flow strategy customization method based on software defined network | |
| Chao et al. | Flow classification for software-defined data centers using stream mining | |
| CN107317758B (en) | High-reliability fine-grained SDN flow monitoring framework | |
| Guo et al. | Research on SDN/NFV network traffic management and optimization based on big data and artificial intelligence | |
| CN110290077B (en) | Industrial SDN resource allocation method based on real-time service configuration | |
| Tajiki et al. | CECT: computationally efficient congestion-avoidance and traffic engineering in software-defined cloud data centers | |
| CN109688056A (en) | Intelligent Network Control System and method | |
| CN114827021B (en) | Multimedia service flow acceleration system based on SDN and machine learning | |
| CN109617819B (en) | Software-defined backhaul network routing method based on traffic classification | |
| CN108880888A (en) | A kind of SDN network method for predicting based on deep learning | |
| Shirmarz et al. | An autonomic software defined network (SDN) architecture with performance improvement considering | |
| CN114143264A (en) | Traffic scheduling method based on reinforcement learning in SRv6 network | |
| CN104468403B (en) | A kind of SDN controllers for carrying out network flow classification to packet based on NACC | |
| CN116599904A (en) | Parallel transmission load balancing device and method | |
| CN112260953A (en) | Multi-channel data forwarding decision method based on reinforcement learning | |
| Barolli et al. | A genetic algorithm based routing method using two QoS parameters | |
| Hongvanthong | Novel four-layered software defined 5g architecture for ai-based load balancing and qos provisioning | |
| Tajiri et al. | Optimizing edge-cloud cooperation for machine learning accuracy considering transmission latency and bandwidth congestion | |
| CN113642736B (en) | Gradient polymerization method and system based on cold-hot separation | |
| Bolodurina et al. | Comprehensive approach for optimization traffic routing and using network resources in a virtual data center | |
| CN114513816A (en) | SDN network resource allocation method, system and storage medium | |
| Altamirano et al. | QoS-aware Network Self-management Architecture based on DRL and SDN for remote areas | |
| Song et al. | A survey of application of artificial intelligence methods in SDN |
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 |