CN107317758A - A kind of fine granularity SDN traffic monitoring frameworks of high reliability - Google Patents
A kind of fine granularity SDN traffic monitoring frameworks of high reliability Download PDFInfo
- Publication number
- CN107317758A CN107317758A CN201710439320.0A CN201710439320A CN107317758A CN 107317758 A CN107317758 A CN 107317758A CN 201710439320 A CN201710439320 A CN 201710439320A CN 107317758 A CN107317758 A CN 107317758A
- Authority
- CN
- China
- Prior art keywords
- stream
- monitoring
- controller
- flow
- feature
- 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
- H04L47/10—Flow control; Congestion control
-
- 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
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
- H04L47/24—Traffic characterised by specific attributes, e.g. priority or QoS
- H04L47/2441—Traffic characterised by specific attributes, e.g. priority or QoS relying on flow classification, e.g. using integrated services [IntServ]
Abstract
A kind of fine granularity SDN traffic monitoring frameworks of high reliability, by introducing the supervisory processor system independently of specific transactions and controller, the forwarding of controller is separated with monitoring, monitoring processor employs three-tier architecture, up and down controller and exchanger side are corresponded to respectively two layers, intermediate layer is handled the flow monitoring data that lower floor's exchanger side is collected, and produces stream characteristic set；Upper strata is made up of control order resolver and stream feature selector；The select command issued according to controller determines selected stream feature, the stream characteristic set that stream feature selector is provided intermediate layer according to selected stream feature carries out machine learning and data mining, the statistic of generation provides whole network data that is accurate and having predictability as the decision information of controller for forwarding；This framework has highly reliable, flexibility good and the advantage such as monitoring accuracy height, is communicated using application programming interfaces with controller and interchanger, with stronger scalability, facilitates and transplanted and safeguarded.
Description
Technical field
The present invention relates to the technical field that computer network monitors framework, more particularly to a kind of fine granularity of high reliability
SDN traffic monitoring frameworks, the traffic monitoring framework under software defined network is innovated and extended, with nowadays deep learning
Technology is combined.
Background technology
Modern data center network is operated on large-scale data exchange calculating, carries magnanimity business for example virtual
Cloud computing, big data application, data center services and multimedia transmission etc., legacy network also relies on network manager to locate
It is as network load dynamically changes, while legacy network is to manage such as network congestion not prompt enough of network transients problem very much
Hardly possible obtains the whole network information, and this is just difficult for dynamically carrying out, and network traffics scheduling offer is efficient and Reliability Assurance.
Therefore for many drawbacks of legacy network, next generation network (Next GenerationNetwork) technology
Arise at the historic moment, wherein software defined network SDN (Software Defined Network) is representational as a kind of comparison
Technology receives the extensive concern of industry.SDN has decoupled control plane and datum plane make it have it is efficient and flexible
Network management capabilities.This network architecture provides a kind of control of centralization allows router and interchanger by issuing flow table
With the rate processing network traffics of its circuit without introducing obvious overhead (according to the certain journey of specific framework of realization
Also overhead can be introduced on degree)；Controller can pass through the real-time management and control network of control passage simultaneously.SDN technologies are applied to
Instantly huge variety of network type, thus develop under the popularization of numerous research institutions and industrial group it is more and more faster.
SDN most prominent characteristic is exactly the whole network control, to realize that network control efficiently and accurately must be with efficient
Premised on accurate acquisition the whole network information, therefore SDN Network Monitoring Technology is determine SDN framework overall performance first
Step.Present SND network flow monitorings technology mainly has following a few point defects：
(1) because high-frequency stream statistics amount queue can introduce monitoring expense in controller.
(2) the problem of traffic monitoring generates non-flexible with forwarding using identical flow table, while for monitoring mesh
Bag feature with for forwarding the bag feature of purpose always not identical or overlapping.
(3) because the limited capacity of current hardware switch make it that stream entry number is very limited and then it is reliable to introduce
Sex chromosome mosaicism, router flow table congestion is easily caused when flow table quantity is excessive.
The content of the invention
In order to overcome the defect of above-mentioned prior art, it is an object of the invention to provide a kind of fine granularity of high reliability
SDN traffic monitoring frameworks, the characteristics of being separated by exclusive control with forwarding, improve the management and control ability of the whole network, are improving
The abilities such as fault recovery and prediction and load balancing are improved again while forward efficiency.
In order to achieve the above object, the technical scheme is that：
A kind of fine granularity SDN traffic monitoring frameworks of high reliability, the monitoring of SDN and forwarding capability are separated,
Traffic monitoring framework includes three layers：Controller side, stream information process layer and exchanger side；The superiors are controller sides, are contained
Control order resolver and stream feature selector, are responsible for entering the demand convection current characteristic set for monitoring stream information according to controller
Row study and excavation, extract the flow statistic needed for controller, data source are provided for its forwarding decision；Stream information processing
Layer is responsible for filtering the initial information of flow, and the statistic needed for extracting and then generation stream characteristic set, are upper strata control
The stream feature selecting of device side processed provides data source；Orlop is exchanger side, is made up of local control application and monitoring data storehouse,
It is responsible for filtering and storing the stream entry for monitoring stream according to the monitoring requirement of controller；Traffic monitoring framework comprising two kinds it is open should
Use routine interface：Data exchange and instruction between the controller side of monitoring processor and SDN controllers issue interface and prison
Control the data exchange and parameter setting interface between the exchanger side and interchanger of processor.
Described controller side is made up of control order resolver and stream feature selector, realizes and controller is issued
The parsing of order and the selection for flowing feature, controller mainly has two classes to monitoring processor transmitting order to lower levels in this framework：One class
It is the judgement order that controller is made whether sampling according to its monitoring demand to certain stream of exchanger side, via controller order solution
The local control application of exchanger side can be sent to after parser parsing, is filtered for controlling stream and flows sampling；Another kind of is controller
The select command that demand to the stream selector of controller side flow feature extraction is monitored according to it, controller side is directly controlled
Flow machine learning and data mining that feature selector carries out stream feature；The selected stream characteristic statisticses that stream feature selector is extracted
Amount is directly transmitted by the data exchange interface of monitoring processor and controller.
The principle of described stream feature selector is deep learning, is produced by the self study process of training set and test set
The current statistic of feature and premeasuring are flowed, and then convection current is classified：Feature selector is flowed in, the circulation that lower floor produces
Characteristic set according to the requirement of controller flow the screening of feature, and then by the algorithm of deep learning to according to selected stream
Feature convection current is classified, and sorted flow information is passed into controller for its forwarding provides decision-making.Flow feature selector
Including three modules：(1) characteristic format device is flowed：Its act on be by adfluxion close in statistical information according to feature selecting algorithm shape
Into training set and test set；(2) feature selector：The corresponding stream characteristic set of selection, middle stream letter are required according to controller
Stream characteristic set produced by breath process layer is the set of all stream statistics amounts, the feature that feature selector gives according to controller
The subset of individual features is selected in this set；(3) flow classifier：The training set that grader is produced by feature selector enters
The training of row sorting algorithm, is verified with test set and the data flow of monitoring is classified afterwards, finally by sorted knot
Fruit passes controller back, dispatches the stream that controller carries out next step according to classification results.
Described stream information process layer is by stream characteristic filter device, statistic maker and stream three module groups of characteristic set
Into the function of stream characteristic filter device is mainly filtered the stream entry in monitoring data storehouse, rejects invalid monitoring entry, control
The stream for some port that device regulation flows through some interchanger is all monitored, because disturbance caused by the complexity of network environment can
The invalid packet of energy formation is erroneously interpreted as effectively stream and is monitored, therefore is caused by a variety of validity checks after filtering
Monitoring data is all reliable, extracts stream statistics amount via statistic maker stream characteristic set is formed after classification afterwards, supply
Top level control device side carries out the flow point class of next step.
Described exchanger side by monitoring processor local control application (local control application) and
Monitoring data storehouse two parts composition, the effect that local control is applied is mainly according to the monitoring requirement of SDN controllers to router
Flow monitoring carry out parameter setting and filtering, wherein filtering out what need not be monitored by Bloom filter (Bloom filter)
Whether stream packets (carry out flow monitoring to be determined by the monitoring demand of controller), thus reduce the flow amount of monitoring and enhance stream
The flexibility of monitoring and accuracy；Monitoring data storehouse is responsible for storing stream information and bag statistical value that matching controller monitors demand,
Monitoring data storehouse tables of data is made up of three partial contents：Monitor matching domain, bag count value and cryptographic Hash (Hash Code), monitoring
The stream statistics amount of database can periodically send to stream information process layer the processing for carrying out stream information.
Described exchanger side, the work flow step of its monitoring process is as follows：
After the stream of interchanger reaches the local control application of monitoring processor exchanger side, corresponding stream can be checked first
Entry whether there is in Bloom filter, if there is, it was demonstrated that the stream need not be monitored, therefore arrive afterwards
Packet can all be controlled locally to apply and skip；If not present in Bloom filter, then proving that the stream is monitored
The stream either stream that newly reaches, it is necessary to be judged again；Now can again it be searched in the monitoring table in monitoring data storehouse,
If having found stream entry proves that this stream, just monitored, now updates the value of the flow counter, if in monitoring table
Corresponding stream entry is not found, local control application decides whether to adopt this stream according to the monitoring rules of controller
Sample；If without sampling, this stream entry is added in Bloom filter, the data after this stream are directly filtered, if
Sampled, the stream is added in monitoring table.
Data exchange and instruction between the controller side of described monitoring processor and SDN controllers issue interface, fixed
Justice is as follows：
(1)controllerCommondMessage(Controller message)：The interface is used for SDN controllers pair
Monitoring processor transmitting order to lower levels, a plurality of control order of the interface encapsulation, it will order by monitoring processor controller side
Resolver is parsed to control order, and the order of SDN controllers as claimed in claim 3 is divided into the stream of convection current selector
Feature sets and the flow monitoring of exchanger side is set, and resolver can be respectively sent to stream feature choosing after control order is parsed
Select the locally applied control of device and exchanger side.
(2)flowFeatureStatistic(flow ID,classification)：The interface is used for feature selector
Flow monitoring instruction according to controller is classified to monitored stream returns controller by information transmission after produced flow point class,
Mark and classification results comprising stream.
Data exchange and parameter setting interface between the exchanger side and interchanger of described monitoring processor, definition is such as
Under：
(1)setMonitoringParameter(Monitoring Match Fields,Counter Buffer
Size,Switch ID,Entry ID)：The interface is used to set monitoring parameter, is one of order of controller, by controller layer
The local control application of exchanger side is reached after the parsing of command analysis device, the interface is used to set monitoring matching domain, counter to delay
Deposit size, interchanger ID, the parameter value of port id；
(2) cutMonitoringEntry (Monitoring Match Fields, Switch ID, Entry ID)：Should
Interface is used to stop being monitored certain stream of certain Single port；
(3)setSamplingRadio(Sampling value,Switch ID,Entry ID)：The interface is used to set
Port sampling rate.
The fine granularity SDN traffic monitorings framework of high reliability proposed by the present invention mainly by introduce monitoring processor this
Independently of the mechanism of SDN controllers and interchanger, SDN controller monitorings and being separated from each other for forwarding are realized, makes controller only
Forwarding service is undertaken, and monitoring processor is responsible for the filtering, monitoring, the stream statistics amount need for handling and being issued according to controller of flow
Seek the feedback that the whole network information is provided for controller.The processing load of SDN controllers is greatly alleviated, monitoring processor passes through certainly
The API (application programming interfaces) of definition is communicated with SDN controllers and interchanger, makes it should independently of specific SDN
With realizing cross-platform characteristic, the monitoring mechanism of this lightweight causes storage and the processing module of monitoring data, simultaneously
Alleviate the flow table storage pressure of SDN switch.Due to similar to middleware independently of specific transactions and the prison of control plane
Control mechanism, adds the portability and autgmentability of the framework, and the exploitation by user's routine interface is also to reach at monitoring
Manage the programmable of device.
Compared to traditional traffic monitoring framework, this framework has following features：
(1) high reliability：The monitoring module of third party's form is employed, is carried out by application programming interfaces and SDN
Communication, independently of specific transactions and the network architecture.The extension of flexibility can be carried out according to the complexity of real network, to monitoring
The analysis of data calculates and employs distributed mode, greatly alleviates the calculating pressure of controller.
(2) fine granularity：Traditional traffic monitoring framework is only according to underlying protocol (southbound interface under such as SDN
OpenFlow the monitoring field) carried is monitored, the general value for only monitoring flow port counter under SDN, according to data
Bag number simply flowed feature judgement, the fine granularity monitoring of this framework be embodied in the excavation of convection current feature not only only according to
Rely in port counter, but count a variety of stream features, entered according to the environment of particular network, scale, service feature, performance indications
The extraction and excavation of row different characteristic so that monitored results are more accurate.
(3) scalability：The framework of this plug-in type can also be carried out can be real on function and Expansion, monitoring processor
Existing more business demands, it is only necessary to which developing new application programming interfaces just can be so that the function that monitoring processor is realized is to control
It is transparent for device, controller only needs transmitting order to lower levels and exchanges data.Monitoring processor can answering according to real network
Miscellaneous degree carries out quantity extension.
Brief description of the drawings
Fig. 1 is the fine granularity SDN traffic monitoring architectural configurations schematic diagrames of high reliability.
Fig. 2 is monitoring processor three-decker schematic diagram.
Fig. 3 is stream feature selector structural representation.
Fig. 4 is exchanger side structural representation.
Fig. 5 is flow monitoring schematic flow sheet.
Embodiment
The present invention is described in detail with example below in conjunction with the accompanying drawings.
Reference picture 2, the fine granularity SDN traffic monitoring frameworks of a kind of high reliability, by the monitoring of SDN and forwarding capability
It is separated, traffic monitoring framework includes three layers：Controller side, stream information process layer and exchanger side；The superiors are controller sides,
Control order resolver and stream feature selector are contained, is responsible for the demand convection current feature to monitoring stream information according to controller
Set is learnt and excavated, and extracts the flow statistic needed for controller, data source is provided for its forwarding decision；Stream letter
Breath process layer is responsible for filtering the initial information of flow, the statistic needed for extracting and then generation stream characteristic set, is
The stream feature selecting of top level control device side provides data source；Orlop is exchanger side, by local control application and monitoring data
Storehouse is constituted, and is responsible for filtering and storing the stream entry for monitoring stream according to the monitoring requirement of controller；Traffic monitoring framework includes two kinds
Open application programming interfaces：Data exchange and instruction between the controller side of monitoring processor and SDN controllers are issued and connect
Data exchange and parameter setting interface between mouth and the exchanger side and interchanger of monitoring processor.
Described controller side is made up of control order resolver and stream feature selector, realizes and controller is issued
The parsing of order and the selection for flowing feature, controller mainly has two classes to monitoring processor transmitting order to lower levels in this framework：One class
It is the judgement order that controller is made whether sampling according to its monitoring demand to certain stream of exchanger side, via controller order solution
The local control application of exchanger side can be sent to after parser parsing, is filtered for controlling stream and flows sampling；Another kind of is controller
The select command that demand to the stream selector of controller side flow feature extraction is monitored according to it, controller side is directly controlled
Flow machine learning and data mining that feature selector carries out stream feature；The selected stream characteristic statisticses that stream feature selector is extracted
Amount is directly transmitted by the data exchange interface of monitoring processor and controller.
Controller side is to mitigate controller with exchanging by separating calculating business as monitoring processor top layer main function
The mass data of machine is exchanged, and only the statistic of traffic monitoring is provided and pre- to controller by the way that the stream feature selector of top layer is final
Measurement, controller major function only undertakes control and forwarding strategy.All data related to monitoring are all handled via monitoring
Device is handled and calculated, and by monitoring the separation with forwarding data, drastically increases the overall reliability of network.
The principle of described stream feature selector is deep learning, is produced by the self study process of training set and test set
The current statistic of feature and premeasuring are flowed, and then convection current is classified：Feature selector is flowed in, the circulation that lower floor produces
Characteristic set according to the requirement of controller flow the screening of feature, and then by the algorithm of deep learning to according to selected stream
Feature convection current is classified, and sorted flow information is passed into controller for its forwarding provides decision-making.Flow feature selector
Including three modules, as shown in figure 3,：(1) characteristic format device is flowed：Its act on be by adfluxion close in statistical information according to spy
Levy selection algorithm formation training set and test set；(2) feature selector：The corresponding stream feature set of selection is required according to controller
Close, the stream characteristic set produced by middle stream information process layer is the set of all stream statistics amounts, and feature selector is according to control
The given feature of device processed selects the subset of individual features in this set；(3) flow classifier：Grader passes through feature selector
The training set of generation carries out the training of sorting algorithm, is verified afterwards with test set and the data flow of monitoring is classified, most
Pass sorted result back controller afterwards, dispatch the stream that controller carries out next step according to classification results.
Described stream information process layer is by stream characteristic filter device, statistic maker and stream three module groups of characteristic set
Into the function of stream characteristic filter device is mainly filtered the stream entry in monitoring data storehouse, rejects invalid monitoring entry, control
The stream for some port that device regulation flows through some interchanger is all monitored, because disturbance caused by the complexity of network environment can
The invalid packet of energy formation is erroneously interpreted as effectively stream and is monitored, therefore is caused by a variety of validity checks after filtering
Monitoring data is all reliable, extracts stream statistics amount via statistic maker stream characteristic set is formed after classification afterwards, supply
Top level control device side carries out the flow point class of next step.
Described exchanger side by monitoring processor local control application (local control application) and
Monitoring data storehouse two parts composition, such as Fig. 4, the effect of local control application is mainly according to the monitoring requirement pair of SDN controllers
The flow monitoring of router carries out parameter setting and filtering, need not wherein being filtered out by Bloom filter (Bloom filter)
The stream packets (whether carrying out flow monitoring to be determined by the monitoring demand of controller) of monitoring, thus reduce flow amount and the increasing of monitoring
The flexibility of strong flow monitoring and accuracy；It is responsible for storing the stream information and bag system of matching controller monitoring demand in monitoring data storehouse
Evaluation, monitoring data storehouse tables of data is made up of three partial contents：Matching domain, bag count value and cryptographic Hash (Hash Code) are monitored,
The stream statistics amount in monitoring data storehouse can periodically send to stream information process layer the processing for carrying out stream information.
Introducing Bloom filter in exchanger side has compared to other data structures, and its room and time aspect has
Big advantage.Bloom filter memory space and insertion/query time are all constants.In addition, Hash functions do not have each other
There is relation, it is convenient by hardware parallel realization.Bloom filter does not need storage element in itself；Bloom filter is it is desirable that one
Bit array (this is somewhat similar with bitmap) and k mapping function (similar with Hash tables), in original state, be for length
M bit array array, its all positions are all set to 0.For there is the set S={ s1, s2......sn } of n element, pass through
K mapping function f1, f2 ... and fk }, by each element sj (1 in set S<=j<=n) be mapped as k value g1,
G2......gk }, then again by array [g1] corresponding in bit array array, array [g2] ... array [gk] are put
For 1；If searching some element item whether in S, by mapping function { f1, f2.....fk } obtain k value g1,
G2.....gk }, array [g1] is then judged again, and whether array [g2] ... array [gk] are all 1, if being all 1,
Item is in S, and otherwise item is not in S.Therefore according to the flow monitoring order of controller, the stream entry that need not be monitored is mapped
Into Bloom filter, so the packet of the stream all can be monitored and filtered out by the grand device of cloth afterwards.Reduce the number of monitoring
According to amount and complexity.
Described exchanger side, the work flow step of its monitoring process is as follows, as shown in Figure 5：
After the stream of interchanger reaches the local control application of monitoring processor exchanger side, corresponding stream can be checked first
Entry whether there is in Bloom filter, if there is, it was demonstrated that the stream need not be monitored, therefore arrive afterwards
Packet can all be controlled locally to apply and skip；If not present in Bloom filter, then proving that the stream is monitored
The stream either stream that newly reaches, it is necessary to be judged again；Now can again it be searched in the monitoring table in monitoring data storehouse,
If having found stream entry proves that this stream, just monitored, now updates the value of the flow counter, if in monitoring table
Corresponding stream entry is not found, local control application decides whether to adopt this stream according to the monitoring rules of controller
Sample；If without sampling, this stream entry is added in Bloom filter, the data after this stream are directly filtered, if
Sampled, the stream is added in monitoring table.
Data exchange and instruction between the controller side of described monitoring processor and SDN controllers issue interface, fixed
Justice is as follows：
(1)controllerCommondMessage(Controller message)：The interface is used for SDN controllers pair
Monitoring processor transmitting order to lower levels, a plurality of control order of the interface encapsulation, it will order by monitoring processor controller side
Resolver is parsed to control order, and the order of SDN controllers as claimed in claim 3 is divided into the stream of convection current selector
Feature sets and the flow monitoring of exchanger side is set, and resolver can be respectively sent to stream feature choosing after control order is parsed
Select the locally applied control of device and exchanger side.
(2)flowFeatureStatistic(flow ID,classification)：The interface is used for feature selector
Flow monitoring instruction according to controller is classified to monitored stream returns controller by information transmission after produced flow point class,
Mark and classification results comprising stream.
Data exchange and parameter setting interface between the exchanger side and interchanger of described monitoring processor, definition is such as
Under：
(1)setMonitoringParameter(Monitoring Match Fields,Counter Buffer
Size,Switch ID,Entry ID)：The interface is used to set monitoring parameter, is one of order of controller, by controller layer
The local control application of exchanger side is reached after the parsing of command analysis device, the interface is used to set monitoring matching domain, counter to delay
Deposit size, interchanger ID, the parameter value of port id.
(2) cutMonitoringEntry (Monitoring Match Fields, Switch ID, Entry ID)：Should
Interface is used to stop being monitored certain stream of certain Single port.
(3)setSamplingRadio(Sampling value,Switch ID,Entry ID)：The interface is used to set
Port sampling rate.
Inventive network instance graph as shown in figure 1, the programmability of SDN be embodied in from data Layer to key-course again to
Each functional module of application layer, this programmability effectively improves legacy network very flexible, towards specific transactions
QOS (service quality) ensure the drawback such as bad adaptability.SDN the whole network management and control ability must be set up in efficient, low consumption and standard
On true the whole network flow information collection and analysis framework, for occur in current SDN traffic monitorings framework high expense, it is low can
Monitoring and analytical model by property and coarseness, this framework is by introducing the monitoring module of this lightweight of monitoring processor
Flow monitoring is carried out to SDN, monitoring processor is connected by controller with monitoring processor interface API with controller, led to
Cross interchanger to be connected with group of switches with control unit interface API, the interface made by this opening, controller can issue monitoring
Order, monitoring processor realizes the data exchange with controller and interchanger simultaneously also by such interface.
Claims (7)
1. the fine granularity SDN traffic monitoring frameworks of a kind of high reliability, it is characterised in that by the monitoring of SDN and forwarding work(
It can be separated, traffic monitoring framework includes three layers：Controller side, stream information process layer and exchanger side；The superiors are controllers
Side, contains control order resolver and stream feature selector, is responsible for the demand convection current to monitoring stream information according to controller
Characteristic set is learnt and excavated, and extracts the flow statistic needed for controller, and data source is provided for its forwarding decision；
Stream information process layer is responsible for filtering the initial information of flow, the statistic needed for extracting and then generation stream feature set
Close, the stream feature selecting for top level control device side provides data source；Orlop is exchanger side, is applied and is monitored by local control
Database is constituted, and is responsible for filtering and storing the stream entry for monitoring stream according to the monitoring requirement of controller；Traffic monitoring framework is included
Two kinds of open application programming interfaces：Under data exchange and instruction between the controller side of monitoring processor and SDN controllers
Send out the data exchange and parameter setting interface between interface and the exchanger side and interchanger of monitoring processor.
2. the fine granularity SDN traffic monitoring frameworks of a kind of high reliability according to claim 1, it is characterised in that described
Controller side by control order resolver and stream feature selector constitute, realize the parsing to controller transmitting order to lower levels and
The selection of feature is flowed, controller mainly there are two classes to monitoring processor transmitting order to lower levels in this framework：One class be controller according to
Its monitoring demand is made whether the judgement order of sampling, meeting after the parsing of via controller command analysis device to certain stream of exchanger side
The local control application of exchanger side is sent to, is filtered for controlling stream and flows sampling；Another kind of is that controller monitors need according to it
The select command that stream feature extraction is carried out to the stream selector of controller side is sought, the stream feature selector of controller side is directly controlled
Carry out the machine learning and data mining of stream feature；The selected stream characteristic statistic that stream feature selector is extracted is directly by prison
The data exchange interface of control processor and controller is transmitted.
3. the fine granularity SDN traffic monitoring frameworks of a kind of high reliability according to claim 2, it is characterised in that described
The principle of stream feature selector be deep learning, it is current to produce stream feature by the self study process of training set and test set
Statistic and premeasuring, and then convection current classified：Flow feature selector according in, lower floor generation circulation characteristic set according to
The requirement of controller carries out the screening of stream feature, and then by the algorithm of deep learning to being carried out according to selected stream feature convection current
Classification, passes to controller for its forwarding by sorted flow information and provides decision-making.Flowing feature selector includes three modules：
(1) characteristic format device is flowed：Its act on be by adfluxion close in statistical information according to feature selecting algorithm formation training set and survey
Examination collection；(2) feature selector：The corresponding stream characteristic set of selection is required according to controller, middle stream information process layer is produced
Raw stream characteristic set is the set of all stream statistics amounts, and the feature that feature selector gives according to controller is selected in this set
Select out the subset of individual features；(3) flow classifier：The training set that grader is produced by feature selector carries out sorting algorithm
Training, is verified with test set and the data flow of monitoring is classified, finally pass sorted result back controller afterwards,
The stream for making controller carry out next step according to classification results is dispatched.
4. the fine granularity SDN traffic monitoring frameworks of a kind of high reliability according to claim 1, it is characterised in that described
Stream information process layer by stream characteristic filter device, statistic maker and stream three modules of characteristic set constitute, flow feature mistake
The function of filter is mainly filtered the stream entry in monitoring data storehouse, rejects invalid monitoring entry, and controller regulation is flowed through
The stream of some port of some interchanger is all monitored, due to caused by the complexity of network environment disturbance be likely to form it is invalid
Packet be erroneously interpreted as that effectively stream is monitored, therefore the monitoring data after filtering is caused all by a variety of validity checks
It is reliable, stream statistics amount is extracted via statistic maker stream characteristic set is formed after classification afterwards, for top level control device
Side carries out the flow point class of next step.
5. the fine granularity SDN traffic monitoring frameworks of a kind of high reliability according to claim 1, it is characterised in that described
Exchanger side by monitoring processor local control apply (local control application) and monitoring data storehouse two
Part is constituted, and the effect of local control application is mainly carried out according to the monitoring requirement of SDN controllers to the flow monitoring of router
Parameter setting and filtering, wherein filtering out the stream packets that need not be monitored by Bloom filter (Bloom filter) (is
No progress flow monitoring is determined by the monitoring demand of controller), thus reduce the flow amount of monitoring and enhance the flexible of flow monitoring
Property and accuracy；It is responsible for storing stream information and bag statistical value that matching controller monitors demand, monitoring data storehouse in monitoring data storehouse
Tables of data is made up of three partial contents：Monitor matching domain, bag count value and cryptographic Hash (Hash Code), the stream in monitoring data storehouse
Statistic can periodically send to stream information process layer the processing for carrying out stream information.
6. the fine granularity SDN traffic monitoring frameworks of a kind of high reliability according to claim 5, it is characterised in that described
Exchanger side, the work flow step of its monitoring process is as follows：
After the stream of interchanger reaches the local control application of monitoring processor exchanger side, corresponding stream entry can be checked first
With the presence or absence of in Bloom filter, if there is, it was demonstrated that the stream need not be monitored, therefore the data arrived afterwards
Bao Douhui is controlled locally to apply and skipped；If not present in Bloom filter, then proving that the stream is monitored stream
The stream either newly reached, it is necessary to be judged again；Now can again it be searched in the monitoring table in monitoring data storehouse, if
Having found stream entry proves that this stream, just monitored, now updates the value of the flow counter, if do not had in monitoring table
Corresponding stream entry is found, local control application decides whether to sample to this stream according to the monitoring rules of controller；
If without sampling, this stream entry is added in Bloom filter, the data after this stream are directly filtered, if carried out
Sampling then adds the stream in monitoring table.
7. the fine granularity SDN traffic monitoring frameworks of a kind of high reliability according to claim 1, it is characterised in that described
Monitoring processor controller side and SDN controllers between data exchange and instruction issue interface, be defined as follows：
(1)controllerCommondMessage(Controller message)：The interface is used for SDN controllers to monitoring
Processor transmitting order to lower levels, a plurality of control order of the interface encapsulation, it will command analysis by monitoring processor controller side
Device is parsed to control order, and the order of SDN controllers as claimed in claim 3 is divided into the stream feature of convection current selector
Set and the flow monitoring of exchanger side is set, resolver can be respectively sent to flow feature selector after control order is parsed
With the locally applied control of exchanger side.
(2)flowFeatureStatistic(flow ID,classification)：The interface be used for feature selector according to
The flow monitoring instruction of controller is classified to monitored stream returns controller by information transmission after produced flow point class, comprising
The mark and classification results of stream.
Data exchange and parameter setting interface between the exchanger side and interchanger of described monitoring processor, are defined as follows：
(1)setMonitoringParameter(Monitoring Match Fields,Counter Buffer Size,
Switch ID,Entry ID)：The interface is used to set monitoring parameter, is one of order of controller, by controller layer order
The local control application of exchanger side is reached after resolver parsing, the interface is used to set monitoring matching domain, counter caching big
Small, interchanger ID, the parameter value of port id；
(2) cutMonitoringEntry (Monitoring Match Fields, Switch ID, Entry ID)：The interface
Certain stream for stopping to certain Single port is monitored；
(3)setSamplingRadio(Sampling value,Switch ID,Entry ID)：The interface is used to set port
Sampling rate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710439320.0A CN107317758B (en) | 2017-06-12 | 2017-06-12 | High-reliability fine-grained SDN flow monitoring framework |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710439320.0A CN107317758B (en) | 2017-06-12 | 2017-06-12 | High-reliability fine-grained SDN flow monitoring framework |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107317758A true CN107317758A (en) | 2017-11-03 |
| CN107317758B CN107317758B (en) | 2020-06-26 |
Family
ID=60184033
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710439320.0A Active CN107317758B (en) | 2017-06-12 | 2017-06-12 | High-reliability fine-grained SDN flow monitoring framework |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107317758B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107770098A (en) * | 2017-09-05 | 2018-03-06 | 全球能源互联网研究院有限公司 | A kind of transformer station's station communication drainage method and system based on SDN |
| CN108282481A (en) * | 2018-01-30 | 2018-07-13 | 电子科技大学 | A kind of data center network flow-measuring method based on SDN |
| CN108460423A (en) * | 2018-03-26 | 2018-08-28 | 大连大学 | A kind of business recognition method based on SDN frameworks |
| CN108566388A (en) * | 2018-03-27 | 2018-09-21 | 西安电子科技大学 | SDN stream rule conflict detection methods based on Bloom Filter and system |
| CN108881028A (en) * | 2018-06-06 | 2018-11-23 | 北京邮电大学 | The SDN network resource regulating method of application perception is realized based on deep learning |
| CN111126606A (en) * | 2020-03-30 | 2020-05-08 | 同盾控股有限公司 | Platform-based deep learning model autonomous training method, device, equipment and medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104468403A (en) * | 2014-11-14 | 2015-03-25 | 北京航空航天大学 | SDN controller for performing network flow classification on data packets based on NACC |
| CN106101163A (en) * | 2016-08-29 | 2016-11-09 | 北京工业大学 | Network architecture safety monitoring system based on OpenFlow |
-
2017
- 2017-06-12 CN CN201710439320.0A patent/CN107317758B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104468403A (en) * | 2014-11-14 | 2015-03-25 | 北京航空航天大学 | SDN controller for performing network flow classification on data packets based on NACC |
| CN106101163A (en) * | 2016-08-29 | 2016-11-09 | 北京工业大学 | Network architecture safety monitoring system based on OpenFlow |
Non-Patent Citations (3)
| Title |
|---|
| A.S.D.SILVA,ET AL.: "Identification and Selection of Flow Feature for Accurate Traffic Classification in SDN", 《2015 IEEE 14TH INTERNATIONAL SYMPOSIUM ON NETWORK COMPUTING AND APPLICATIONS》 * |
| X.T.PHAN,K.FUKUDA: "SDN-Mon:Fine-Grained Traffic Monitoring Framework in Software-Defined Networks", 《JOURNAL OF INFORMATION PROCESSING》 * |
| 王海 等: "基于SDN的分布式网络流量监控平台", 《电力信息与通信技术》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107770098A (en) * | 2017-09-05 | 2018-03-06 | 全球能源互联网研究院有限公司 | A kind of transformer station's station communication drainage method and system based on SDN |
| CN108282481A (en) * | 2018-01-30 | 2018-07-13 | 电子科技大学 | A kind of data center network flow-measuring method based on SDN |
| CN108282481B (en) * | 2018-01-30 | 2020-07-17 | 电子科技大学 | SDN-based data center network flow measurement method |
| CN108460423A (en) * | 2018-03-26 | 2018-08-28 | 大连大学 | A kind of business recognition method based on SDN frameworks |
| CN108460423B (en) * | 2018-03-26 | 2020-06-02 | 大连大学 | Service identification method based on SDN architecture |
| CN108566388A (en) * | 2018-03-27 | 2018-09-21 | 西安电子科技大学 | SDN stream rule conflict detection methods based on Bloom Filter and system |
| CN108566388B (en) * | 2018-03-27 | 2020-10-16 | 西安电子科技大学 | SDN flow rule conflict detection method and system based on bloom filter |
| CN108881028A (en) * | 2018-06-06 | 2018-11-23 | 北京邮电大学 | The SDN network resource regulating method of application perception is realized based on deep learning |
| CN108881028B (en) * | 2018-06-06 | 2020-06-26 | 北京邮电大学 | SDN network resource scheduling method for realizing application awareness based on deep learning |
| CN111126606A (en) * | 2020-03-30 | 2020-05-08 | 同盾控股有限公司 | Platform-based deep learning model autonomous training method, device, equipment and medium |
| CN111126606B (en) * | 2020-03-30 | 2020-07-28 | 同盾控股有限公司 | Platform-based deep learning model autonomous training method, device, equipment and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107317758B (en) | 2020-06-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107317758A (en) | A kind of fine granularity SDN traffic monitoring frameworks of high reliability | |
| CN109981691B (en) | SDN controller-oriented real-time DDoS attack detection system and method | |
| CN109831386B (en) | Optimal path selection algorithm based on machine learning under SDN | |
| CN107231384A (en) | A kind of ddos attack detection defence method cut into slices towards 5g networks and system | |
| CN108282497A (en) | For the ddos attack detection method of SDN control planes | |
| CN112491619A (en) | Self-adaptive distribution technology for service customized network resources based on SDN | |
| CN110298381A (en) | A kind of cloud security service functional tree Network Intrusion Detection System | |
| CN105099916B (en) | Open flows route exchange device and its processing method to data message | |
| CN105308920B (en) | Multicast Flows reorder scheme | |
| CN107959625A (en) | Virtual router with dynamic stream relieving capacity | |
| CN106559407A (en) | A kind of Network traffic anomaly monitor system based on SDN | |
| CN106972985A (en) | Accelerate the method and DPI equipment of the processing of DPI device datas and forwarding | |
| CN111478854B (en) | Real-time network node importance ordering method based on flow data | |
| CN108183917A (en) | DDoS attack cross-layer cooperative detection method based on software defined network | |
| CN110324327A (en) | User and server ip address caliberating device and method based on specific enterprise domain name data | |
| CN110213175A (en) | A kind of intelligent managing and control system and management-control method towards knowledge definition network | |
| CN106789242A (en) | A kind of identification application intellectual analysis engine based on mobile phone client software behavioral characteristics storehouse | |
| CN108880888A (en) | A kind of SDN network method for predicting based on deep learning | |
| CN105471670A (en) | Flow data classification method and device | |
| CN108833376A (en) | Software-oriented defines the DoS attack detection method of network | |
| CN102387082A (en) | Flow-classification-based grouping flow control system and control method | |
| CN108696453A (en) | The power telecom network SDN Business Streams cognitive method and system of lightweight | |
| CN109587000A (en) | High latency method for detecting abnormality and system based on collective intelligence network measurement data | |
| CN116668380A (en) | Message processing method and device of convergence diverter equipment | |
| CN108809752A (en) | A kind of adaptive process monitoring method, apparatus of network flow, NPB equipment and medium |
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 |