CN108881028A - The SDN network resource regulating method of application perception is realized based on deep learning - Google Patents
The SDN network resource regulating method of application perception is realized based on deep learning Download PDFInfo
- Publication number
- CN108881028A CN108881028A CN201810573044.1A CN201810573044A CN108881028A CN 108881028 A CN108881028 A CN 108881028A CN 201810573044 A CN201810573044 A CN 201810573044A CN 108881028 A CN108881028 A CN 108881028A
- Authority
- CN
- China
- Prior art keywords
- network
- sdn
- data
- application
- node
- 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
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/50—Queue scheduling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- H04L45/74—Address processing for routing
- H04L45/745—Address table lookup; Address filtering
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/70—Admission control; Resource allocation
- H04L47/80—Actions related to the user profile or the type of traffic
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L49/00—Packet switching elements
- H04L49/25—Routing or path finding in a switch fabric
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The SDN network resource regulating method of application perception is realized based on deep learning, content is：Network characteristic based on SDN network, deep neural network DNN is disposed on the virtual network function VNF for being located at data plane, the application data stream that the DNN forwards interchanger is learnt and is classified, and classification results are reported to SDN controller, SDN controller carries out network resource scheduling according to classification results, generate the routing iinformation for meeting the application data stream network resource requirement, and the routing iinformation is handed down to interchanger, method of the invention substantially increases the flexibility and robustness of system, it realizes and rational management is carried out to Internet resources according to the resource requirement of application, to improve the service quality of network.
Description
Technical field
The present invention relates to a kind of SDN network resource regulating methods that application perception is realized based on deep learning, belong to information
Technical field especially belongs to SDN network technical field.
Background technique
There can be various access devices in Internet of Things, different access devices may carry different type
Network application, these network applications have different demands to Internet resources.Such as the audio for the Internet bearer
(VOIP), this kind of application has higher requirement to the time delay of network, we should distribute low time delay as far as possible for this kind of application
Path.And for video surveillance applications, it had not only needed the network path of low time delay, but also needing to occupy sufficiently large bandwidth could be real
The real-time Transmission of existing video data.Under scenes of internet of things, to realize that the scheduling on demand key of network is to obtain the tool of flow
Body type information.
Traditional flow scheduling is mostly based on datagram header packet information and carries out traffic classification, Internet engineering task force
(The Internet Engineering Task Force, abbreviation IETF) is that some standard agreements are assigned with fixed port
Number, network flow can be divided into different applicating categories by corresponding port numbers, but the presence of the classification based on port numbers is permitted
More problems：First is that increasing with network application, many application layer protocols do not distribute special port numbers, this certain applications
Agreement can not be distinguished by port numbers；Second is that certain application layer protocols may carry in a variety of different types of applications
Hold, different application contents is also completely different for the demand of network.Such as http agreement is answering of being most widely used at present
With layer protocol, when http agreement is used for web page browsing, its clock synchronization, which is postponed a meeting or conference, certain requirement, when http agreement is for carrying view
When frequency flow, it is bandwidth sensitive, and big bandwidth link can be substantially reduced the load time of video, promotes user experience, and
For time delay, there is no very high requirements.In conjunction with the above analysis, simple is far from by header packet information progress traffic classification
No more.Conventional means of the another kind for flow identification is deep packet inspection technical (DPI).DPI technology is a kind of based on application
The flow detection technology of layer, referred to as " deep-packet detection ".For DPI in addition to analyzing header packet information, it can also identify various answer
With specific load contents.It, should when IP data packet, TCP or UDP message stream are by bandwidth management system based on DPI technology
System recombinates the application layer message in seven layer protocol of OSI by the deep content for reading IP payload package, to obtain
Then the content of entire application program carries out shaping operation to flow according to the management strategy that system defines.The certain journey of DPI technology
Degree solve the problems, such as tradition based on header field carry out flow identification, but it there is also many problems：(1) scalability
Difference：Since this method has hysteresis quality to the flow identification that new P2P is applied, i.e., new answer can not be detected before non-upgrade feature library
With, it is necessary to after finding the load characteristic of new opplication, which could effectively be detected.This point becomes limitation this method
Bottleneck.(2) lack encryption data analytic function：Certain application loads use encrypted transmission, conceal the agreement sum number of application
According to feature, therefore depth data packet inspection technical is very limited to the detectability of encryption application.(3) cost is higher：Due to needing
The operations such as protocol analysis reduction and characteristic matching are completed, therefore calculating and storage overhead are big, flow detection algorithm performance is low.It carries
Lotus feature is more complicated, and detection cost is higher, and algorithm performance is also poorer.
To sum up, under scenes of internet of things, how quickly and effectively to identify that the concrete type of data traffic becomes technology of Internet of things
One urgent need in field solves technical problem.
Summary of the invention
Software defined network SDN is a kind of novel network architecture, realizes network data plane and controls point of plane
From providing the programmability of data plane equipment, can be realized the intelligent management of network.SDN controller is control plane
Hinge, all information of network can be got, these information include the topology of network, the bandwidth of link, time delay etc..Knot
These information are closed, controller can meet the transmission path of its network demand for different types of application traffic distribution.But SDN
Controller is to issue control strategy to SDN switch by specific control channel, once controller breaks down, network is just
Meeting is out of hand or even can collapse.
In view of this, inventing a kind of network flow kind identification method the purpose of the present invention is being based on SDN network, SDN is allowed
Controller is capable of the type of aware application, and then realizes the scheduling on demand to Internet resources.
In order to achieve the above object, the invention proposes the SDN network resource tune that application perception is realized based on deep learning
The content of degree method, the method is：Network characteristic based on SDN network, in the virtual network function VNF for being located at data plane
Upper deployment deep neural network DNN, the application data stream which forwards interchanger are learnt and are classified, and classification is tied
Fruit is reported to SDN controller, and SDN controller carries out network resource scheduling according to classification results, and generation meets the application data stream
The routing iinformation of network resource requirement, and the routing iinformation is handed down to interchanger.
The DNN will be trained in advance, and trained method is：Collect different type application in advance by interchanger
Data on flows is trained DNN using the method for supervised learning.
The method includes following operating procedures：
(1) client host is applied to server host and issues data packet, which enters SDN network；The visitor
The fringe node SA of family end main frame and SDN network is connected, and the fringe node SB of the server host and SDN network is connected；
(2) the fringe node SA of the SDN network described in inquires flow table after receiving the data packet, if there is matching stream accordingly
Table then forwards the data packet by flow table rule；If not matching flow table, which is passed through in packet_In message
It is transmitted to SDN controller；
(3) SDN controller receives the packet_In message, parses the data packet reported, obtains the number according to network topology
According to the source node and destination node of packet, respectively SA and SB；
(4) SDN controller calculates one from node SA to the transmission path of node SB using shortest path first, and will
The path integration is at OpenFlow flow table, then switching node all in the flow table issuance to the path；Described
All data packets that client host application issues can all be matched to above-mentioned flow table and are finally forwarded to according to the movement of the flow table
Server host；
(5) it is node SA that SDN controller, which calculates a source using shortest path first, and purpose is connected by the VNF
SDN network fringe node SC path, which can be converted into flow table and be issued to all nodes on the path；At this time
At SA, the data packet that the client host application is sent can be copied two parts, and a copy of it can be forwarded to node SB, this
Partial data packet is eventually forwarded to server host, another data packet can be forwarded to the node SC that VNF is connected；
(6) VNF is sampled according to the data packet that the sampling duration of setting applies the client host, and sampling is completed
Afterwards, VNF calculates the feature vector of data on flows and this feature vector is sent into the DNN and classifies；
(7) classification results are marked on the dscp field in the packet header IP of data packet, and the data packet can pass through later
Packet_In information reporting gives SDN controller；
(8) SDN controller receives the classification results that VNF is reported, and classification results are mapped to preset resource requirement, described
Resource requirement refer mainly to bandwidth requirement and delay requirement；
(9) SDN controller is SB's the purpose of source is SA using depth-priority-searching method DFS (Depth First Search)
One is searched in all paths and meets the path of resource requirement, and converts flow table for the path, which can be issued to institute
State all nodes on path；These flow tables priority with higher can cover the flow table issued before, at this time the client
The rear afterflow rate of end main frame application can be all forwarded along the path for meeting its resource requirement.
Feature vector described in step (6) refers to：According to the temporal aspect of the data flow of application feature calculated to
Amount；The stream data definition of the application is with identical { source IP, destination IP, source port number, destination slogan, agreement (TCP
Or UDP), i.e. { Source IP, Destination IP, Source Port, Destination Port and Protocol
(TCP or UDP) } a series of continuous data packets.
The feature vector is specifically made of following characteristics：The arrival time interval of forward data report, including maximum value,
Minimum value, average and standard deviation；The arrival time interval of backward datagram, including maximum value, minimum value, average value, standard
Difference；The arrival time interval of bi-directional data report, including maximum value, minimum value, average value, standard deviation；Forward data report hair per second
The message number and byte number sent；Afterwards to the message number and byte number of datagram transmission per second；Before being reached in time per second
The ratio between packet number to datagram and backward datagram and the ratio between byte number；The forward data report refers to that client is sent to service
The uplink traffic of device, the backward datagram refer to that server replies to the downlink traffic of client.
The beneficial effects of the present invention are：The classification task of flow is deployed in VNF by method of the invention, VNF work
In the data plane of SDN network, the sampling of data is to be switched through to distribute by switchboard direct, will not occupy control channel, and
The failure of VNF will not influence the basic function of controller, therefore method of the invention substantially increases flexibility and the Shandong of system
Stick realizes and carries out rational management to Internet resources according to the resource requirement of application, to improve the service quality of network.
Detailed description of the invention
Fig. 1 is the schematic diagram of the SDN network of one embodiment of the invention.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, the present invention is made with reference to the accompanying drawing further
Detailed description.
Network characteristic based on SDN network, in the virtual network function VNF (Virtualized for being located at data plane
Network Function) on dispose deep neural network DNN, the application data stream which forwards interchanger learns
And classification, and classification results are reported to SDN controller, SDN controller carries out network resource scheduling according to classification results, generates
Meet the routing iinformation of the application data stream network resource requirement, and the routing iinformation is handed down to interchanger.
The DNN will be trained in advance, and trained method is：Collect different type application in advance by interchanger
Data on flows is trained DNN using the method for supervised learning.
Referring to table 1, in order to verify the feasibility of the method for the present invention, inventor will acquire stream from 18 kinds of different applications
Amount data are classified as 8 seed types, respectively web page browsing, instant messaging, video flowing, audio stream, mail, IP phone, P2P and
File transmission.The present invention is concerned with the resource requirement applied to network, certain possible several types are to network in above-mentioned 8 seed type
Demand be it is similar, therefore inventor by 8 above-mentioned seed types cluster for shown in table 14 classification.
Table 1
| Type | Applicating category | Resource requirement |
| 1 | Instant messaging IP phone | Low time delay low bandwidth |
| 2 | Video flowing audio stream | High bandwidth moderate latency |
| 3 | Web page browsing | Time delay lower bandwidth is lower |
| 4 | P2P file transmission | High bandwidth |
Referring to Fig. 1, the method includes following operating procedures：
(1) client host (h1) application issues data packet to server host (h3), which enters SDN network；
The fringe node S1 of the client host and SDN network is connected, the fringe node of the server host and SDN network
S7 is connected；
(2) the fringe node S1 of the SDN network described in inquires flow table after receiving the data packet, if there is matching stream accordingly
Table then forwards the data packet by flow table rule；If not matching flow table, which is passed through in packet_In message
It is transmitted to SDN controller；
(3) SDN controller receives the packet_In message, parses the data packet reported, obtains the number according to network topology
According to the source node and destination node of packet, respectively S1 and S7；
(4) SDN controller calculates one from node S1 to the transmission path of node S7 using shortest path first
PathA, and by the path integration at OpenFlow flow table, then exchange section all in the flow table issuance to the path
Point；All data packets that the client host application issues can all be matched to above-mentioned flow table and finally according to the dynamic of the flow table
It is forwarded to server host；
(5) it is node S1 that SDN controller, which calculates a source using shortest path first, and purpose is connected by the VNF
Node S3 path P athB, which can be converted into flow table and be issued to all nodes on the path；At this time in node
At S1, the data packet that the client host application is sent can be copied two parts, and a copy of it can be forwarded to node S7, this portion
Divided data packet is eventually forwarded to server host, another data packet can be forwarded to the node S3 that VNF is connected；
(6) VNF is adopted according to the data packet that the sampling duration (such as 15 seconds) of setting applies the client host
Sample, after the completion of sampling, VNF, which calculates the feature vector of data on flows and this feature vector is sent into the DNN, to be divided
Class；(7) classification results are marked on DSCP (the Differentiated Services Code in the packet header IP of data packet
Point) field, such as：It by dscp field set is 00100000 if classification results are Class1, data later
Packet can give SDN controller by packet_In information reporting；
(8) SDN controller receives the classification results that VNF is reported, and classification results are mapped to preset resource requirement, described
Resource requirement refer mainly to bandwidth requirement and delay requirement；
(9) one is searched in all paths that SDN controller is node S7 the purpose of source is node S1 using DFS algorithm completely
The path P athC of sufficient resource requirement, and flow table is converted by the path, which can be issued to all sections on the path
Point；These flow tables priority with higher can cover the flow table issued before, at this time the rear afterflow rate of the client application
It will be forwarded along the path for meeting its resource requirement.
Feature vector described in step (6) refers to：According to the temporal aspect of the data flow of application feature calculated to
Amount；The stream data definition of the application is with identical { source IP, destination IP, source port number, destination slogan, agreement (TCP
Or UDP), i.e. { Source IP, Destination IP, Source Port, Destination Port and Protocol
(TCP or UDP) } a series of continuous data packets.
The feature vector is specifically made of following characteristics：The arrival time interval of forward data report, including maximum value,
Minimum value, average and standard deviation；The arrival time interval of backward datagram, including maximum value, minimum value, average value, standard
Difference；The arrival time interval of bi-directional data report, including maximum value, minimum value, average value, standard deviation；Forward data report hair per second
The message number and byte number sent；Afterwards to the message number and byte number of datagram transmission per second；Before being reached in time per second
The ratio between packet number to datagram and backward datagram and the ratio between byte number；The forward data report refers to that client is sent to service
The uplink traffic of device, the backward datagram refer to that server replies to the downlink traffic of client.
Inventor is based on Mininet and has built emulation experiment environment, and wherein interchanger uses the OVS soft switch of secondary development
Machine, controller are based on OpenDayLight secondary development.Inventor has carried out emulation experiment, simulation result to method of the invention
Show that method of the invention is fully effective.
Claims (5)
1. realizing the SDN network resource regulating method of application perception based on deep learning, it is characterised in that：The content of the method
It is：Network characteristic based on SDN network disposes deep neural network on the virtual network function VNF for being located at data plane
DNN, the application data stream which forwards interchanger are learnt and are classified, and classification results are reported to SDN controller,
SDN controller carries out network resource scheduling according to classification results, generates the routing for meeting the application data stream network resource requirement
Information, and the routing iinformation is handed down to interchanger.
2. the SDN network resource regulating method according to claim 1 for realizing application perception based on deep learning, feature
It is：The DNN will be trained in advance, and trained method is：Collect the stream of different type application in advance by interchanger
Data are measured, DNN is trained using the method for supervised learning.
3. the SDN network resource regulating method according to claim 1 or 2 that application perception is realized based on deep learning,
It is characterized in that：The method includes following operating procedures：
(1) client host is applied to server host and issues data packet, which enters SDN network；The client
Host and the fringe node SA of SDN network are connected, and the fringe node SB of the server host and SDN network is connected；
(2) the fringe node SA of the SDN network described in inquires flow table after receiving the data packet, if there is matching flow table accordingly,
Then the data packet is forwarded by flow table rule；If not matching flow table, which is uploaded to by packet_In message
SDN controller；
(3) SDN controller receives the packet_In message, parses the data packet reported, obtains the data packet according to network topology
Source node and destination node, respectively SA and SB；
(4) SDN controller calculates one from node SA to the transmission path of node SB, the road Bing Jianggai using shortest path first
Diameter is converted into OpenFlow flow table, then switching node all in the flow table issuance to the path；The client
All data packets that end main frame application issues can all be matched to above-mentioned flow table and finally be forwarded to service according to the movement of the flow table
Device host；
(5) it is node SA, the SDN that purpose is connected by the VNF that SDN controller, which calculates a source using shortest path first,
The path of network edge node SC, the path can be converted into flow table and be issued to all nodes on the path；At this time in SA
Place, the data packet that the client host application is sent can be copied two parts, and a copy of it can be forwarded to node SB, this part
Data packet is eventually forwarded to server host, another data packet can be forwarded to the node SC that VNF is connected；
(6) VNF is sampled according to the data packet that the sampling duration of setting applies the client host, after the completion of sampling,
VNF, which calculates the feature vector of data on flows and this feature vector is sent into the DNN, to classify；
(7) classification results are marked on the dscp field in the packet header IP of data packet, and the data packet can pass through later
Packet_In information reporting gives SDN controller；
(8) SDN controller receives the classification results that VNF is reported, and classification results are mapped to preset resource requirement, the money
Source demand refers mainly to bandwidth requirement and delay requirement；
(9) SDN controller using depth-priority-searching method DFS algorithm be the purpose of source is SA in all paths of SB search one it is full
The path of sufficient resource requirement, and flow table is converted by the path, which can be issued to all nodes on the path；This
The flow table that a little flow table priority with higher issue before covering, the rear afterflow rate that the client host is applied at this time is all
It can be forwarded along the path for meeting its resource requirement.
4. the SDN network resource regulating method according to claim 3 for realizing application perception based on deep learning, feature
It is：Feature vector described in step (6) refers to：According to the temporal aspect of the data flow of application feature vector calculated；
The stream data definition of the application be with it is identical source IP, destination IP, source port number, destination slogan, agreement (TCP or
UDP) }, i.e. { Source IP, Destination IP, Source Port, Destination Port and Protocol
(TCP or UDP) } a series of continuous data packets.
5. the SDN network resource regulating method according to claim 4 for realizing application perception based on deep learning, feature
It is：The feature vector is specifically made of following characteristics：The arrival time interval of forward data report, including maximum value, most
Small value, average and standard deviation；The arrival time interval of backward datagram, including maximum value, minimum value, average value, standard deviation；
The arrival time interval of bi-directional data report, including maximum value, minimum value, average value, standard deviation；The transmission per second of forward data report
Message number and byte number；Afterwards to the message number and byte number of datagram transmission per second；The forward direction number reached in time per second
According to the ratio between the ratio between report and the packet number of backward datagram and byte number；The forward data report refers to that client is sent to server
Uplink traffic, the backward datagram refer to that server replies to the downlink traffic of client.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810573044.1A CN108881028B (en) | 2018-06-06 | 2018-06-06 | SDN network resource scheduling method for realizing application awareness based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810573044.1A CN108881028B (en) | 2018-06-06 | 2018-06-06 | SDN network resource scheduling method for realizing application awareness based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108881028A true CN108881028A (en) | 2018-11-23 |
| CN108881028B CN108881028B (en) | 2020-06-26 |
Family
ID=64337123
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810573044.1A Active CN108881028B (en) | 2018-06-06 | 2018-06-06 | SDN network resource scheduling method for realizing application awareness based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108881028B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110298381A (en) * | 2019-05-24 | 2019-10-01 | 中山大学 | A kind of cloud security service functional tree Network Intrusion Detection System |
| CN110380940A (en) * | 2019-08-22 | 2019-10-25 | 北京大学深圳研究生院 | A kind of appraisal procedure of router and its data packet |
| CN110581802A (en) * | 2019-08-27 | 2019-12-17 | 北京邮电大学 | fully-autonomous intelligent routing method and device based on deep belief network |
| CN110601916A (en) * | 2019-08-14 | 2019-12-20 | 天津大学 | Flow sampling and application sensing system based on machine learning |
| CN110708260A (en) * | 2019-11-13 | 2020-01-17 | 鹏城实验室 | Data packet transmission method and related device |
| CN111355670A (en) * | 2018-12-24 | 2020-06-30 | 中移（杭州）信息技术有限公司 | Traffic identification method and device, electronic equipment and storage medium |
| CN112311578A (en) * | 2019-07-31 | 2021-02-02 | 中国移动通信集团浙江有限公司 | VNF scheduling method and device based on deep reinforcement learning |
| CN113452559A (en) * | 2021-06-24 | 2021-09-28 | 同济大学浙江学院 | Network resource allocation method, system and medium based on deep learning |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104518984A (en) * | 2014-12-22 | 2015-04-15 | 北京航空航天大学 | SDN controller for carrying out traffic classification on data package based on multiple classifiers |
| CN104994033A (en) * | 2015-05-13 | 2015-10-21 | 南京航空航天大学 | Method for guaranteeing QoS (quality of service) of SDN (software defined network) by means of dynamic resource management |
| US20160105364A1 (en) * | 2014-10-13 | 2016-04-14 | Nec Laboratories America, Inc. | Network traffic flow management using machine learning |
| CN107317758A (en) * | 2017-06-12 | 2017-11-03 | 西安交通大学 | A kind of fine granularity SDN traffic monitoring frameworks of high reliability |
| CN107547379A (en) * | 2016-06-23 | 2018-01-05 | 华为技术有限公司 | The method and relevant device of route test action are generated in software defined network |
-
2018
- 2018-06-06 CN CN201810573044.1A patent/CN108881028B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160105364A1 (en) * | 2014-10-13 | 2016-04-14 | Nec Laboratories America, Inc. | Network traffic flow management using machine learning |
| CN104518984A (en) * | 2014-12-22 | 2015-04-15 | 北京航空航天大学 | SDN controller for carrying out traffic classification on data package based on multiple classifiers |
| CN104994033A (en) * | 2015-05-13 | 2015-10-21 | 南京航空航天大学 | Method for guaranteeing QoS (quality of service) of SDN (software defined network) by means of dynamic resource management |
| CN107547379A (en) * | 2016-06-23 | 2018-01-05 | 华为技术有限公司 | The method and relevant device of route test action are generated in software defined network |
| CN107317758A (en) * | 2017-06-12 | 2017-11-03 | 西安交通大学 | A kind of fine granularity SDN traffic monitoring frameworks of high reliability |
Non-Patent Citations (1)
| Title |
|---|
| LU HE，等: ""vTC：Machine Learning Based Traffic Classification as a Virtual Network Function"", 《SDN-NFV SECURITY "16: PROCEEDINGS OF THE 2016 ACM INTERNATIONAL WORKSHOP ON SECURITY IN SOFTWARE DEFINED NETWORKS & NETWORK FUNCTION VIRTUALIZATION》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111355670A (en) * | 2018-12-24 | 2020-06-30 | 中移（杭州）信息技术有限公司 | Traffic identification method and device, electronic equipment and storage medium |
| CN111355670B (en) * | 2018-12-24 | 2023-05-02 | 中移（杭州）信息技术有限公司 | Traffic identification method and device, electronic equipment and storage medium |
| CN110298381A (en) * | 2019-05-24 | 2019-10-01 | 中山大学 | A kind of cloud security service functional tree Network Intrusion Detection System |
| CN110298381B (en) * | 2019-05-24 | 2022-09-20 | 中山大学 | Cloud security service function tree network intrusion detection system |
| CN112311578A (en) * | 2019-07-31 | 2021-02-02 | 中国移动通信集团浙江有限公司 | VNF scheduling method and device based on deep reinforcement learning |
| CN112311578B (en) * | 2019-07-31 | 2023-04-07 | 中国移动通信集团浙江有限公司 | VNF scheduling method and device based on deep reinforcement learning |
| CN110601916A (en) * | 2019-08-14 | 2019-12-20 | 天津大学 | Flow sampling and application sensing system based on machine learning |
| CN110380940A (en) * | 2019-08-22 | 2019-10-25 | 北京大学深圳研究生院 | A kind of appraisal procedure of router and its data packet |
| CN110380940B (en) * | 2019-08-22 | 2022-05-24 | 北京大学深圳研究生院 | Router and data packet evaluation method thereof |
| CN110581802A (en) * | 2019-08-27 | 2019-12-17 | 北京邮电大学 | fully-autonomous intelligent routing method and device based on deep belief network |
| CN110708260A (en) * | 2019-11-13 | 2020-01-17 | 鹏城实验室 | Data packet transmission method and related device |
| CN113452559A (en) * | 2021-06-24 | 2021-09-28 | 同济大学浙江学院 | Network resource allocation method, system and medium based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108881028B (en) | 2020-06-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108881028A (en) | The SDN network resource regulating method of application perception is realized based on deep learning | |
| CN104753704B (en) | State transfer method and interchanger in a kind of software defined network | |
| CN108353029A (en) | For managing the method and system for calculating the data service in network | |
| KR20060052982A (en) | Auto-ip traffic optimization in mobile telecommunications systems | |
| CN104158753A (en) | Dynamic flow dispatch method and system based on software definition network | |
| US20160241435A1 (en) | Apparatus for optimising a configuration of a communications network device | |
| WO2016161937A1 (en) | Method, device, and system for identifying traffic flow | |
| CN110324260A (en) | A kind of network function virtualization intelligent dispatching method based on flow identification | |
| Liu et al. | An adaptive approach for elephant flow detection with the rapidly changing traffic in data center network | |
| Adedayo et al. | QoS functionality in software defined network | |
| Xu et al. | IARA: An intelligent application-aware VNF for network resource allocation with deep learning | |
| Vladyko et al. | Fuzzy model of dynamic traffic management in software-defined mobile networks | |
| Nurhaida et al. | Performance Comparison based on Open Shortest Path First (OSPF) Routing Algorithm for IP Internet Networks | |
| Silva et al. | AVIRA: Enhanced multipath for content-aware adaptive virtual reality | |
| Mahboob et al. | A novel deep-learning-enabled qos management scheme for encrypted traffic in software-defined cellular networks | |
| Al-Haddad et al. | A Survey of Quality of Service (QoS) Protocols and Software-Defined Networks (SDN) From the Traditional to the Latest Network Architecture | |
| Bifulco et al. | CATENAE: A scalable service function chaining system for legacy mobile networks | |
| Chodorek et al. | A Linux kernel implementation of the traffic flow description option | |
| Sari et al. | Implementation of Bandwidth Management and Access Restrictions Using PCQ and Firewall Methods in SMP Tunas Bangsa Network | |
| Tosic et al. | Soft sensors in wireless networking as enablers for SDN based management of content delivery | |
| Calderón et al. | Predicting traffic through artificial neural networks | |
| Chahlaoui et al. | Towards QoS-enabled SDN networks | |
| Mohammed et al. | Comparison of Schecduling Schemes in IPV4 and IPV6 to Achieve High QoS | |
| Tansupasiri et al. | Using active networks technology for dynamic QoS | |
| CN115174393B (en) | Service function chain dynamic adjustment method based on in-band network telemetry |
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 |