WO2017032254A1 - 网络控制策略的生成方法、装置及网络控制器 - Google Patents

网络控制策略的生成方法、装置及网络控制器 Download PDF

Info

Publication number
WO2017032254A1
WO2017032254A1 PCT/CN2016/095696 CN2016095696W WO2017032254A1 WO 2017032254 A1 WO2017032254 A1 WO 2017032254A1 CN 2016095696 W CN2016095696 W CN 2016095696W WO 2017032254 A1 WO2017032254 A1 WO 2017032254A1
Authority
WO
WIPO (PCT)
Prior art keywords
network
state information
data stream
data
current
Prior art date
Application number
PCT/CN2016/095696
Other languages
English (en)
French (fr)
Inventor
耿彦辉
陈志堂
张宝峰
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Priority to EP16838513.6A priority Critical patent/EP3297211B1/en
Publication of WO2017032254A1 publication Critical patent/WO2017032254A1/zh
Priority to US15/903,051 priority patent/US10749757B2/en

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0803—Configuration setting
    • H04L41/0813—Configuration setting characterised by the conditions triggering a change of settings
    • H04L41/0816—Configuration setting characterised by the conditions triggering a change of settings the condition being an adaptation, e.g. in response to network events
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0893—Assignment of logical groups to network elements
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0894—Policy-based network configuration management
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0895—Configuration of virtualised networks or elements, e.g. virtualised network function or OpenFlow elements
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/147—Network analysis or design for predicting network behaviour
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/40—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using virtualisation of network functions or resources, e.g. SDN or NFV entities
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00—Arrangements for monitoring or testing data switching networks
    • H04L43/06—Generation of reports
    • H04L43/065—Generation of reports related to network devices
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00—Arrangements for monitoring or testing data switching networks
    • H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
    • H04L43/0876—Network utilisation, e.g. volume of load or congestion level
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0803—Configuration setting
    • H04L41/0823—Configuration setting characterised by the purposes of a change of settings, e.g. optimising configuration for enhancing reliability

Definitions

  • Embodiments of the present invention relate to the field of network data routing, and more specifically, to a method, an apparatus, and a network controller for generating a network control policy, and a network data routing method and apparatus thereof.
  • the traditional network control strategy generation process generally transforms the problem into a series of optimization problems.
  • the input of the optimization problem is the topology of the network, the bandwidth of the link in the network, the traffic matrix of the network, etc.
  • the solution of the optimization problem is generally the optimized end-to-end path of the network, or the transmission rate of the transmitting end.
  • the traditional network control strategy generation method has the following defects: 1.
  • the optimization problem is generally obtained by linear programming or integer programming. Limited by the complexity of linear programming or integer programming, the scalability of this method is relatively weak, especially as the number of network elements, service types and traffic continue to increase, the optimization solution will become too complicated to solve or solve the cost ( If the calculation time is too large. At the same time, it is difficult to implement dynamic and immediate policy adjustment through the offline optimization scheme.
  • the embodiment of the invention provides a method, a device and a network controller for generating a network control strategy, which have a self-learning function and can adaptively adjust the control strategy dynamically.
  • an embodiment of the present invention provides a method for generating a network control policy, which is applied to an SDN network, where the method includes:
  • the network state data includes network current state data, where the network state information includes network current state information, and the data according to the network state Before estimating the network status information, the method further includes:
  • the estimating the network state information according to the network state data specifically includes:
  • the network state information further includes network history state information, and the predicting the network state information according to the Before the network next time status information, the method further includes:
  • the predicting the network next time state information according to the network state information comprises: predicting the network next time state information according to the network current state information and the network historical state information.
  • the first possible implementation of the first aspect or any one of the possible implementations of the second possible implementation of the first aspect, in a third possible implementation of the first aspect ,
  • the predicting the next time state information of the network according to the network state information is:
  • the first possible implementation of the first aspect, or any one of the possible implementations of the third possible implementation of the first aspect, in a fourth possible implementation of the first aspect The preprocessing the data stream to obtain a data stream preprocessing result, including:
  • Extracting characteristics of the data stream including a packet header quintuple, a packet length sequence, or a packet arrival time interval sequence;
  • the application current network control policy is based on the network state information, the network next time state information, and
  • the data stream pre-processing result generation control action is:
  • S t represents the network state information
  • i i represents a tag of the data flow
  • i represents an i th data stream
  • ⁇ t represents the current network control policy
  • Express S t , l i is the probability distribution of a under the current network control strategy of the condition variable, where a is a control variable
  • the control action a i is randomly sampled in the probability distribution of the control variable a.
  • an embodiment of the present invention provides a method for generating a network control policy, which is applied to an SDN network, where the method includes:
  • Network status information of the network is used to describe a network link status of the network, a network element queue status in the network, and the The network data flow distribution of the network, the next time state information of the network is predicted according to the network state information, and the data stream pre-processing result is obtained by pre-processing the data stream currently entering the network;
  • the network next time state information is predicted according to the network state information, where the prediction method is:
  • the data stream pre-processing result is: pre-processing the data stream currently entering the network
  • the method of preprocessing includes:
  • Extracting characteristics of the data stream including a packet header quintuple, a packet length sequence, or a packet arrival time interval sequence;
  • the embodiment of the present invention provides a method for generating a network control policy, which is applied to an SDN network, where the method includes: receiving feedback information from the SDN network, and forwarding the feedback information to a control module, where And causing the control module to perform real-time adjustment on the current network control policy according to the feedback information to obtain a next-time network control policy, where the feedback information is based on a control action issued by the control module in the network operation at a last moment.
  • the network is controlled.
  • an embodiment of the present invention provides a device for generating a network control policy, which is applied to an SDN network, where the device includes:
  • An obtaining module configured to acquire network state data of the network
  • a prediction module configured to estimate network state information according to the network state data, and predict, according to the network state information, status information of the network next time, where the network state information is used to describe a network link of the network, The network element queue situation in the network, and the network data flow distribution of the network;
  • a stream data pre-processing module configured to receive a data stream currently entering the network, and pre-process the data stream to obtain a data stream pre-processing result
  • control module configured to apply a current network control policy to generate a control action according to the network state information, the network next time state information, and the data flow pre-processing result
  • An execution module configured to control the network according to the control action, and obtain feedback information of the network controlled by the control action
  • the control module is further configured to receive feedback information of the network, and obtain a next-time network control policy according to real-time adjustment of the current network control policy according to the feedback information.
  • the network state data includes network current state data
  • the network state information includes network current state information
  • the prediction module is specifically configured to:
  • the network state information further includes network history state information, where the prediction module is further configured to: obtain from a database And the network history state information; and predicting the next time state information of the network according to the current state information of the network and the historical state information of the network.
  • the prediction module is specifically configured to:
  • the stream data preprocessing module includes:
  • An extracting unit configured to extract a feature of the data stream, where the feature includes a packet header quintuple, a data packet length sequence, or a data packet arrival time interval sequence;
  • a mining analysis unit configured to perform time-space data mining processing and/or causal relationship analysis processing on the data stream by using the feature, to obtain a processing result
  • an identifier unit configured to label the data stream according to the processing result to obtain a label of the data stream, where the label of the data stream is the data stream pre-processing result.
  • control module is specifically configured to:
  • S t represents the network state information
  • l i represents a tag of the data stream
  • i represents an i th data stream
  • ⁇ t represents the current network control policy
  • Express S t , l i is the probability distribution of a under the current network control strategy of the condition variable, where a is a control variable
  • the control action a i is randomly sampled in the probability distribution of the control variable a.
  • an embodiment of the present invention provides a device for generating a network control policy, which is applied to an SDN network, where the device includes:
  • a receiving module configured to receive network status information of the network, network next time status information, and data stream pre-processing result; wherein the network status information is used to describe a network link status of the network, and a network element in the network a queue situation, a network data flow distribution of the network, the network next time state information is predicted according to the network state information, and the data stream pre-processing result is a pre-predetermined data stream currently entering the network Processed
  • control module configured to apply a current network control policy to generate a control action according to the network state information, the network next time state information, and the data flow pre-processing result, where the control action is used to control the network and obtain a The feedback information of the network controlled by the control action;
  • the control module is further configured to receive feedback information of the network, and perform real-time adjustment on the current network control policy according to the feedback information to obtain a next-time network control policy.
  • the information that the network next time state information is predicted according to the network state information is:
  • the data stream pre-processing result is: pre-processing the data stream currently entering the network
  • the specifics include:
  • Extracting characteristics of the data stream including a packet header quintuple, a packet length sequence, or a packet arrival time interval sequence;
  • an embodiment of the present invention provides a network controller, which is applied to an SDN network, and is configured to receive feedback information from the SDN network, and forward the feedback information to a control module, where the control module is configured to The feedback information is adjusted in real time to the current network control policy to obtain a next-time network control policy, where the feedback information is obtained by controlling the network according to a control action sent by the control module in the network operation at a last moment.
  • the network state information is predicted according to the network state information, and the data stream currently entering the network is preprocessed to obtain a data stream pre-processing result, thereby applying the current network control policy according to the network state information,
  • the network next time state information and the data stream preprocessing result generating control action, controlling the network according to the control action to obtain feedback information, wherein the feedback information is obtained by controlling the network according to the control action,
  • the current network control policy is adjusted in real time according to the feedback information to obtain a next-time network control policy.
  • the network control strategy is always dynamically optimized and adjusted according to the current data and network conditions, which is called adaptive in this paper.
  • the network is controlled according to the network control strategy updated in real time to improve the network control efficiency.
  • FIG. 1 is a structural diagram of a data transmission network in which a method for generating a network control policy according to an embodiment of the present invention is deployed;
  • FIG. 2 is a schematic flowchart of a method 200 for generating a network control policy according to an embodiment of the present invention
  • FIG. 3 is a structural diagram showing an example of a method 200 for generating a network control policy according to an embodiment of the present invention
  • FIG. 4 is another structural diagram of an example 200 of generating a network control policy according to an embodiment of the present invention.
  • FIG. 5 is still another structural diagram of a method for generating a network control policy according to an embodiment of the present invention.
  • FIG. 6 is a schematic flowchart of a method 600 for generating a network control policy according to an embodiment of the present invention
  • FIG. 7 is a structural diagram showing an example of a method 600 for generating a network control policy according to an embodiment of the present invention.
  • FIG. 8 is a schematic flowchart of a method 800 for generating a network control policy according to an embodiment of the present invention.
  • FIG. 9 is a structural block diagram of a network control policy generating apparatus 900 according to an embodiment of the present invention.
  • FIG. 10 is a structural block diagram of a network control policy generating apparatus 1000 according to an embodiment of the present invention.
  • the method for generating a network control policy provided by the present invention may be deployed and implemented in the data transmission network 100 as described in FIG.
  • the data transmission network 100 includes a central control server 110, a source server 120, a network switch 130, a transmission network 140, a client device 150, a prediction server 160, and a training server 170.
  • the central control server 110 is a core processing device of the transmission network 100, and is connected to the network switch 130, the transmission network 140, the client device 150, the prediction server 160, and the training server 170, respectively, for connecting with the network switch 130.
  • the transmission network 140, the training server 170, and the prediction server 160 perform information exchange, command delivery, and network data transmission.
  • the source server 120, the network switch 130, the transmission network 140, the client device 150, the training server 170, and the prediction server 160 can be connected to each other according to different service requirements and functional requirements, and information and instructions are exchanged. The function and composition of each device will be described in detail below.
  • the central control server 110 is mainly composed of main components such as a processor, a memory, and a data interface.
  • the processor mainly performs a corresponding processing function by calling a processing program stored in the storage device, and the data interface is mainly responsible for each device inside the central server 110. Data is transmitted and received between the central processing server 110 and external components.
  • the processor may be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a Field Programmable Gate Array (FPGA), or Other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.
  • DSP digital signal processor
  • ASIC application specific integrated circuit
  • FPGA Field Programmable Gate Array
  • the general purpose processor may be a microprocessor or the processor or any conventional processor or the like.
  • the steps in the method disclosed in the embodiment of the present invention may be directly implemented as a hardware decoding processor, or may be performed by a combination of hardware and software modules in the decoding processor.
  • the software module can be located in a random access memory (RAM), a flash memory, a read-only memory (ROM), a programmable read only memory or an electrically erasable programmable memory, a register, etc. In the storage medium.
  • the central control server 110 receives the routing request reported by the network interaction machine 130 through the data interface, where the routing request includes current data flow information to be routed, and sends the current data flow information in the routing request to the data interface through the data interface. Prediction server 160. The prediction server 160 will next the network according to the information of the current data stream. The time state information is sent to the central control server 110 as a prediction result, and the central processing unit of the central control server 110 calls the pre-stored processing program in the memory, according to the prediction result returned by the prediction server 160 and the current network control policy stored in the called memory.
  • the central control server 110 will configure a route for the current data stream according to the optimal transmission path and will be configured
  • the route is sent to the network switch 130.
  • the central control server 110 is further configured to collect historical data flow information of the completed route reported by the network switch 130 through its data interface, and obtain feedback information according to the pre-routing data and the post-route data, and provide feedback information.
  • the training server 170 sends a real-time adjustment to the current network control policy according to the feedback information to obtain a next-time network control policy.
  • the source server 120 is mainly composed of main components such as a processor, a memory, and a data interface.
  • the processor is mainly used to send data stored in the memory to the client device 150 through the network switch 130 and the transmission network 140 through the data interface.
  • the network switch 130 whose main task is to extract the header of the first data packet of the data stream for the data stream newly entering the network, that is, the current data stream, and report it to the central control server 110, wait for and calculate and configure according to the central control server 110.
  • the route forwards the packets of the current data stream.
  • the network switch 130 also collects the collected historical data stream information of the completed route and reports it to the central control server 110.
  • the transmission network 140 is composed of a transmission cable and a plurality of network nodes, and is mainly used for data transmission according to an instruction of the network switch 130.
  • the client device 150 is composed of a main component such as a processor, a memory, and a data interface.
  • the processor is mainly configured to send a data request of the client to the source server 120 through the data interface and the network switch 130, and pass the source server 120 through the network switch 130.
  • the data content transmitted by the transport network 140 is stored in the memory.
  • the prediction server 160 may be a separate processing device, for example, composed of a separate processor, a memory, and a data interface.
  • the processor is mainly used to call a processing program stored in the memory to receive current data received through the data interface.
  • the stream information is processed to predict the next time state information of the network, and the predicted network next time state information is sent to the central control server 110 through the data interface.
  • the information of the current data stream is from the central control server 110.
  • the prediction server 160 may also be a working unit attached to the central control server 110, that is, share the same memory and data interface with the central control server 110, but adopt different processors. At this time, the prediction server 160 and the central unit Data transfer between the control servers 110 is primarily accomplished through a data bus internal to the central control server 110.
  • the prediction server 160 may also share the same processor with the central control server 110, and the shared manner may be by using the processor.
  • the same processing resource performs synchronous (parallel processing) or asynchronous (time-sharing processing) data processing, or performs asynchronous (time-sharing processing) processing using the same processing resources in the processor.
  • the prediction server 160 can also be implemented by a virtual machine, that is, the central control server 110 uses its own processor, memory, and data interface to simulate a function with a complete hardware system by calling a program stored in the memory. A complete computer system running in a completely isolated environment that can independently perform all of the work required to predict server 160 without affecting the operation of central control server 110 itself. It can be understood that the prediction server 160 can be further simplified as a software program stored in the memory of the central control server 110, which is invoked by the central control server 110 at a suitable time to implement the network next time state information described above. Forecast function.
  • the training server 170 may be an independent processing device, and is composed of main components such as a processor, a memory, and a data interface.
  • the processor is mainly used to control the current network stored in the memory according to the received feedback information.
  • the policy performs real-time adjustment to obtain the next-time network control strategy, and stores the next-time network control policy in the memory for the next round of call.
  • the training server 170 may also be a working unit attached to the central control server 110, that is, share the same memory and data interface with the central control server 110, but use different processors. At this time, the training server Data transfer between 170 and central control server 110 is primarily accomplished via a data bus internal to central control server 110.
  • the training server 170 may also share the same processor with the central control server 110, which may be synchronous (parallel processing) or asynchronous (time-sharing processing) data processing by using different processing resources of the processor. Or asynchronous (time-sharing) data processing using the same processing resources in the processor.
  • the training server 170 can also be implemented by a virtual machine, that is, the central control server 110 uses its own processor, memory, and data interface to simulate a function with a complete hardware system by calling a program stored in the memory. A complete computer system running in a fully isolated environment that can perform all of the work required by the training server 170 independently. It will be appreciated that the training server 170 can be further simplified as a software program stored in the memory of the central control server 110 that is invoked by the central control server 110 at an appropriate time to implement the aforementioned network control strategy. Real-time adjustments and real-time updates.
  • the above is a hardware environment for implementing the network data stream type detection of the present invention, but it should be noted that the above hardware environment is not the only way to implement the present invention, and may be dynamically adjusted according to hardware resources and service requirements, that is, data transmission requirements.
  • the central control server 110, the network switch 130, the prediction server 160, and the training server 170 are centralized in one data control center to improve processing efficiency.
  • the method of the present invention can be deployed and implemented based on the application environment described above.
  • the network data stream type detecting method of the present invention will be described below with reference to the accompanying drawings.
  • Embodiment 1 is a diagrammatic representation of Embodiment 1:
  • FIG. 2 is a schematic flowchart of a method 200 for generating a network control policy according to an embodiment of the present invention.
  • the method 200 for generating a network control policy is applied to an SDN network.
  • the network status data is raw data used to indicate the status of the network.
  • S203 Estimate network state information according to the network state data, and predict, according to the network state information, status information of the network next time, where the network state information is used to describe a network link situation of the network, and the network The status of the network element queue and the distribution of the network data stream of the network.
  • the network status information under the description describes the network link status at time t, the network element queue status in the network, the network data flow distribution, etc.; the network status information at time t-1 (which we can call historical time) Describe the network link situation at time t-1, the network element queue status in the network, the distribution of network data flow, etc.; the network status information at time t+1 (which we can call the next time or future time) Describes the network link status at time t+1, the network element queue status in the network, and the distribution of network data flows, which are not described here.
  • S205 Receive a data stream currently entering the network, and perform pre-processing on the data stream to obtain a data stream pre-processing result.
  • the pre-processing may include, but is not limited to, processing by means of clustering, classification, regression, or causal analysis.
  • S211 Perform real-time adjustment on the current network control policy according to the feedback information to obtain a next-time network control policy.
  • the steps are sequentially performed by numbers such as S201, S203, S205, and the like, the numbers are not specific to the sequence of steps in the solution.
  • Restrictions for example, in this scenario, S203 and S205 do not have such a sequence limit, they can be the same When it occurs, S203 may be earlier than S205, or S203 may occur later than S205. That is, the sequence of steps in the embodiment of the present invention is only related to the direction of the data flow, and is not related to the sequence of the description.
  • the embodiment of the present invention predicts the next state information of the network according to the network state information, and preprocesses the data stream currently entering the network to obtain a data stream pre-processing result, thereby applying the current network control policy according to the network state information and the Determining the network next time state information and the data stream preprocessing result generating control action, controlling the network according to the control action to obtain feedback information, wherein the feedback information is obtained by controlling the network according to the control action
  • the current network control policy is adjusted in real time according to the feedback information to obtain a next-time network control policy.
  • the process is cycled.
  • the next-time network control strategy obtained by this real-time adjustment is adjusted in real time accordingly, and the loop is continued.
  • the network control strategy is always dynamically optimized and adjusted according to the current data and network conditions, which is called adaptive in this paper.
  • the network is controlled according to the network control strategy updated in real time to improve the network control efficiency.
  • the above embodiment can be implemented on the structure shown in FIG.
  • the SDN controller by periodically or randomly triggering a command to collect network status data (represented by line segment 1 in the figure), the SDN controller collects and forwards these network status data commands based on the command (in the figure) Indicated by the line segment 2 to the corresponding processing device, the corresponding processing device estimates the network state information according to the network state data, and the network state information predicts the state information of the next moment of the network, it is worth noting that, corresponding to the network Status information at a moment, where the network status information is used to describe network status information different from the next moment, such as network historical time status information or network current status information.
  • the SDN controller forwards the current data stream entering the network in real time (indicated by line 3 in the figure) to another corresponding The processing device, the corresponding processing device preprocesses the data stream currently entering the network to obtain the data stream preprocessing result, and emphasizes the action of the data stream preprocessing and the network state information estimation and the network current state information prediction action
  • the order of the present invention may be simultaneous, or earlier than or later, because in the operation of the real solution, the original operation is not limited thereto, so that the above actions are performed in any order in the protection scope of the embodiment of the present invention. It will not be repeated here.
  • the network status information, the status information of the next moment of the network, and the data stream pre-processing result are used as input information of the control module, and the control module will invoke the current network control policy in the cache device or in the memory, using the network status information, and the network next. Status information of the time and data stream preprocessing result.
  • the information generates a control action according to the current network control policy, and outputs the control action to the SDN.
  • the controller (indicated by line 4 in the figure), the SDN controller performs specific control commands on the network according to the control action.
  • the network will feedback the feedback effect generated by the control action as feedback information (indicated by line 5 in the figure) to the SDN controller, and the SDN controller forwards the feedback information (indicated by line segment 5 in the figure) to the control module, and the control module according to
  • the feedback information adjusts the current network control policy in real time to obtain the next-time network control policy, and caches the real-time network control policy by the cache device of the control module or stores it in the memory for the next cycle of standby.
  • the control action generated in the first cycle is an optimal control action required for the network in the time or period of the cycle.
  • the feedback information is generated, and the next time network obtained by the feedback information is obtained.
  • this next moment network control strategy will enter the next cycle as the current network control strategy in the next cycle, thereby expanding the next second cycle, the third cycle, the fourth cycle...
  • Each current moment has optimal control of the network state and data flow information at the current moment to improve the control efficiency of the entire control system, and will not be described again.
  • FIG. 3 is an example of the method, and is not limited to the unique structure.
  • the method 200 for generating a network control policy according to the embodiment of the present invention generates a network control policy or Network control and the like should be included in the scope of protection of the present invention and will not be described again.
  • the network status information is time-sensitive.
  • the network status data should also be time-sensitive, that is, the network status data and network status information at the current time may be present, and Network status data and network status information at historical times.
  • the network status data includes network current status data
  • the network status information includes network current status information
  • the method further includes:
  • the estimating the network state information according to the network state data specifically includes:
  • the network state information may further include network history state information (samely, the network state data may also include network history state data), and the network time information is used to predict the next moment of the network.
  • the method may further include:
  • the network history state information is obtained from the database; the database here may be pre-existing data information in the memory.
  • the predicting the next time state information of the network according to the network state information includes: The network current state information and the network history state information predict the network next time state information.
  • estimating the network state information according to the network state data, and predicting the next time state information of the network according to the network state information may be implemented as follows:
  • the method for estimating the network state information according to the network state data is more common in reality, and we illustrate by way of example, for example, we can estimate the link utilization rate of the switch according to the collected switch port counter data;
  • the switch queue counter data can be collected to estimate the switch delay; for example, the collected switch flow table data is used to estimate the number of active data streams in the network.
  • the details of the specific estimation method need not be described here, and those skilled in the art should understand.
  • the SDN controller sends a network state data collection command to the network in real time, or periodically, or by an event-triggered type (line segment 1 in the figure). Representation), the SDN controller collects the collected network state data D t and forwards D t (indicated by line 2 in the figure) to the corresponding processing device.
  • the SDN controller sends a network state data collection command to the network in real time, or periodically, or by an event-triggered type (line segment 1 in the figure). Representation), the SDN controller collects the collected network state data D t and forwards D t (indicated by line 2 in the figure) to the corresponding processing device.
  • D t indicated by line 2 in the figure
  • the state The estimator estimates network state information S t based on D t , where S t is used to represent the current state information of the network, and S t is used as an input to the state predictor (in another embodiment, S t and the network from the database)
  • the historical state information S t-1 , S t-2 , . . . together as the input of the state predictor) the state predictor performs prediction based on the input information to obtain the state information of the next time of the network.
  • the specific prediction method may be the above method, and the function form specifically represented by f here is not limited herein.
  • the network state information that is not necessarily the next time, that is, the next time state information of the network described above, It may be the next two moments, three moments..., we can control this state information as the state information of the future moment, and its prediction idea is consistent with the idea of predicting the next moment.
  • the S t is stored in the database, and may be stored in the database cache device or in the memory to form the network history state information at the next moment for standby.
  • FIG. 4 is an example of the method, and is not limited to the unique structure.
  • the method for generating a network control policy according to the embodiment of the present invention generates a network control policy or Network control and the like should be included in the scope of protection of the present invention and will not be described again.
  • S205 preprocessing the data stream to obtain a data stream preprocessing result may be implemented as follows:
  • Extracting features of the data stream including but not limited to a packet header quintuple, a packet length sequence, or a packet arrival time interval sequence;
  • the label of the data stream is the data stream pre-processing result
  • the pre-processing result may be the label itself, the label described herein It can be an identity, or other related tag that is used to distinguish between stream types and/or flow relationships.
  • step S207 applying the current network control policy according to the network state information, the network next time state information, and the data flow pre-processing result generating control action may be specifically implemented in the following manner:
  • S t represents the network state information
  • l i represents a tag of the data stream
  • i represents an i th data stream
  • ⁇ t represents the current network control policy
  • Express S t , l i is the probability distribution of a under the current network control strategy of the condition variable, where a is a control variable
  • the control action a i is randomly sampled in the probability distribution of the control variable a.
  • the foregoing solution may be implemented on the structure shown in FIG. 5.
  • the SDN controller forwards the data stream currently entering the network to the corresponding processing device.
  • the feature extraction module we call it the feature extraction module and the feature extraction module.
  • the features of the data stream are extracted, and the extracted features are used for spatiotemporal data mining processing and/or causal relationship analysis processing.
  • the following operations can be performed:
  • the extracted features are denoted by x.
  • the spatiotemporal information contained therein may include, but is not limited to:
  • Source IP address 32-bit binary string
  • Destination IP address 32-bit binary string
  • Server port 16-bit binary string
  • Client port 16-bit binary string
  • Transmission protocol category type
  • Starting time real type
  • Packet length integer type
  • Packet inter-arrival time Real type.
  • the specific method of pre-processing includes: spatio-temporal data mining and/or causality mining with respect to data streams.
  • the space-time data mining specifically includes but is not limited to:
  • Clustering A feature set extracted by a data stream or a partial feature set as a feature vector, representing the data stream, and clustering the feature vectors to obtain spatio-temporal information of the data stream. For example, by clustering, Co-flow information can be obtained, that is, it can be analyzed which data streams may belong to the same task.
  • Data streams can often be classified according to their nature. For example, depending on the flow size or duration, the data stream can be divided into an elephant flow and a mice flow. For another example, according to an application generated by a data stream, the data stream can be divided into a video stream, a data backup, and the like.
  • By offline manual tagging we can obtain training data and train the classifier based on the training data. The classifier is used to classify data streams in real time online.
  • Regression and classification have similarities, that is, learning function g is obtained based on training data, function g takes feature as input, and function g outputs a feature of data stream. For example, we can establish a regression model and utilize data flow. The characteristics of the data stream are estimated.
  • causal relationship mining includes but is not limited to:
  • Causal analysis There may be a causal relationship between the data streams. For example, some clients may send some requests to the server, and the server responds accordingly. In this case, there is a causal relationship between the data stream sent by the client to the server and sent by the server to the client.
  • the frequently-used IP address pairs Source IP, Destination IP
  • ⁇ 0 has a causal relationship:
  • the processing result is obtained, and the data stream is tagged according to the processing result to obtain the label l i shown in the figure, that is, the data stream preprocessing result.
  • FIG. 5 is an example of the method, and is not limited to the unique structure.
  • the method 200 for generating a network control policy according to the embodiment of the present invention generates a network control policy or Network control and the like should be included in the scope of protection of the present invention and will not be described again.
  • Embodiment 2 is a diagrammatic representation of Embodiment 1:
  • FIG. 6 is a schematic flowchart of a method 600 for generating a network control policy according to an embodiment of the present invention.
  • the method 600 for generating a network control policy is applied to an SDN network.
  • S601 Receive network state information of the network, state information of the next time of the network, and a data stream preprocessing result.
  • the network status information is used to describe a network link status of the network, a network element queue status in the network, and a network data flow distribution of the network, where the next time status information of the network is according to the The network status information is predicted, and the data stream pre-processing result is pre-processing the data stream currently entering the network. owned.
  • the control action is for controlling the network and obtaining feedback information of the network controlled by the control action.
  • S605. Receive feedback information of the network, and perform real-time adjustment on the current network control policy according to the feedback information to obtain a next-time network control policy.
  • the network next time state information is predicted according to the network state information, where the method for predicting may be specifically:
  • the data stream pre-processing result is obtained by pre-processing the data stream that is currently entering the network, where the pre-processing method may be specifically:
  • Extracting characteristics of the data stream including a packet header quintuple, a packet length sequence, or a packet arrival time interval sequence;
  • the embodiment of the present invention receives network state information, network next-time state information, and data stream pre-processing result, thereby generating control based on network state information, network next-time state information, and data stream pre-processing results according to current network control policies.
  • the control action is used to control the network to obtain feedback information obtained by controlling the network according to the control action, and finally, receiving the feedback information, and controlling the current network according to the feedback information
  • the policy is adjusted in real time to get the next-time network control strategy. When the next network data comes over, the process is cycled. Finally, based on the next feedback information, the real-time adjusted network control strategy is adjusted in real time to obtain the network control strategy at the next moment, and the loop is continued.
  • the network control strategy is always dynamically optimized and adjusted according to the current data and network conditions, which is called adaptive in this paper.
  • the network is controlled according to the network control strategy updated in real time to improve the network control efficiency.
  • the complete solution implementation process in the control system 300 should be: by periodically or randomly triggering a command to collect network status data to the network, the SDN controller collects these network status data commands based on the commands and forwards them to the corresponding processing.
  • the device estimates the network state information according to the network state data by the corresponding processing device, and predicts the state information of the next moment of the network by the network state information, and it is worth noting that the state information corresponding to the next moment of the network, where
  • the network status information is used to describe network status information different from the next moment, such as network historical time status information or network current status information.
  • the SDN controller forwards the current data stream entering the network to another corresponding processing device in real time, and the corresponding processing device The data stream currently entering the network is preprocessed to obtain the data stream preprocessing result.
  • the reason why the action of the data stream preprocessing and the network state information estimation and the network state information prediction action may be emphasized at the same time, or The reason for the above-mentioned actions in the order of the present invention is not limited thereto, and the above-mentioned actions are performed within the scope of protection of the embodiments of the present invention, and will not be described again.
  • the network status information, the status information of the next moment of the network, and the data stream pre-processing result are used as input information of the control module, and the control module will invoke the current network control policy in the cache device or in the memory, using the network status information, and the network next.
  • the status information of the time and the data stream pre-processing result The information generates a control action according to the current network control policy, and outputs the control action to the SDN controller, and the SDN controller performs a specific control command on the network according to the control action.
  • the network feedbacks the control effect generated by the control action to the SDN controller as feedback information, and the SDN controller forwards the feedback information to the control module, and the control module adjusts the current network control policy according to the feedback information to obtain the next time network control.
  • the policy, and the next time network control policy is cached by the control module's cache device or stored in the memory for the next cycle of standby.
  • the control action generated in the first cycle is the optimal control action required for the network in the cycle time or period.
  • the feedback information is generated, and the feedback information is used to obtain the next time network control.
  • This next moment network control strategy will enter the next cycle as the current network control strategy in the next cycle, thereby expanding the next second cycle, the third cycle, the fourth cycle... to guarantee each A current moment has optimal control of the network state and data flow information at the current moment to improve the control efficiency of the entire control system, and will not be described again.
  • the method execution body provided by the embodiment of the present invention may be a control module in the structure 300.
  • the control module receives network status information, network next time status information, and data stream pre-processing results, according to the invoked policy ⁇ t and received information.
  • the control action a i is generated, which is specifically performed as follows: among them Representing the network next time state information, where S t represents the network state information, where l i represents a tag of the data stream, where i represents an i th data stream, where ⁇ t represents the current network control policy ,among them Express S t , l i is the probability distribution of a under the current network control strategy of the condition variable, where a is a control variable, and the control action a i is randomly sampled in the probability distribution of the control variable a, the control action a I obey Probability distributions.
  • the Q learning method or MDP (Markov decision process), or POMDP (partial observable Markov decision process) can be used as shown in the figure.
  • the control action a i is sent to the SDN controller for causing the SDN controller to perform the control action.
  • S is a set of finite network state information
  • A is a set of finite control actions
  • J( ⁇ ) represents an objective function that measures the superiority of the strategy
  • s t , a t ) is a system in action a t
  • the probability of transition from state s t to state s t+1 , R(s t , a t , s t+1 ) is the return of the system from state s t to s t+1
  • s) is a strategy function, indicating the probability that the network state s described by the network state information adopts the control action a
  • ⁇ ⁇ (s) is the distribution of the state s under the policy ⁇ , then the best strategy can be as follows Limit optimization problems to get:
  • Control strategy ⁇ 1 so continuous looping (this loop process can be done offline or online) until we think that a good enough network control strategy can be used as the current network control strategy in the above method 200 or 600, then enter The first cycle in the embodiment provided above.
  • the feedback information may be stored in a cache device of the control module or in a memory, and details are not described herein.
  • FIG. 3 and FIG. 7 is an example of the method, and is not limited to the unique structure.
  • the method for generating a network control policy according to the embodiment of the present invention is used to generate a network control policy. Or network control, etc., should be included in the scope of protection of the present invention, and will not be described again.
  • Embodiment 3 is a diagrammatic representation of Embodiment 3
  • FIG. 8 is a schematic flowchart of a method 800 for generating a network control policy according to an embodiment of the present invention.
  • the network control policy generating device 800 is applied to an SDN network.
  • the embodiment of the present invention receives the feedback information, and forwards the feedback information to the control module, so that the control module adjusts the current network control policy in real time according to the feedback information to obtain a next-time network control policy.
  • the control module adjusts the current network control policy in real time according to the feedback information to obtain a next-time network control policy.
  • the process is cycled.
  • the real-time adjusted network control strategy is adjusted in real time to obtain the network control strategy at the next moment, and the loop is continued.
  • the network control strategy is always dynamically optimized and adjusted according to the current data and network conditions, which is called adaptive in this paper.
  • the network is controlled according to the network control strategy updated in real time to improve the network control efficiency.
  • the complete solution implementation process in the control system 300 should be: by periodically or randomly triggering a command to collect network status data to the network, the SDN controller collects these network status data commands based on the commands and forwards them to the corresponding processing.
  • the device estimates the network state information according to the network state data by the corresponding processing device, and predicts the state information of the next moment of the network by the network state information, and it is worth noting that the state information corresponding to the next moment of the network, where network status
  • the information is used to describe network status information different from the next moment, such as network historical time status information or network current status information.
  • the SDN controller forwards the current data stream entering the network to another corresponding processing device in real time, and the corresponding processing device The data stream currently entering the network is preprocessed to obtain the data stream preprocessing result.
  • the reason why the action of the data stream preprocessing and the network state information estimation and the network state information prediction action may be emphasized at the same time, or The reason for the above-mentioned actions in the order of the present invention is not limited thereto, and the above-mentioned actions are performed within the scope of protection of the embodiments of the present invention, and will not be described again.
  • the network status information, the status information of the next moment of the network, and the data stream pre-processing result are used as input information of the control module, and the control module will invoke the current network control policy in the cache device or in the memory, using the network status information, and the network next.
  • the status information of the time and the data stream pre-processing result The information generates a control action according to the current network control policy, and outputs the control action to the SDN controller, and the SDN controller performs a specific control command on the network according to the control action.
  • the network feedbacks the control effect generated by the control action to the SDN controller as feedback information, and the SDN controller forwards the feedback information to the control module, and the control module adjusts the current network control policy according to the feedback information to obtain the next time network control.
  • the policy, and the next time network control policy is cached by the control module's cache device or stored in the memory for the next cycle of standby.
  • the control action generated in the first cycle is the optimal control action required for the network in the cycle time or period.
  • the feedback information is generated, and the feedback information is used to obtain the next time network control.
  • This next moment network control strategy will enter the next cycle as the current network control strategy in the next cycle, thereby expanding the next second cycle, the third cycle, the fourth cycle... to guarantee each A current moment has optimal control of the network state and data flow information at the current moment to improve the control efficiency of the entire control system, and will not be described again.
  • the method execution body provided by the embodiment of the present invention may be an SDN controller in the structure 300.
  • Embodiment 4 is a diagrammatic representation of Embodiment 4:
  • FIG. 9 is a structural block diagram of a network control policy generating apparatus 900 according to an embodiment of the present invention.
  • the network control policy generating device 900 is applied to an SDN network.
  • the obtaining module 901 is configured to acquire network state data of the network.
  • the network status data is raw data used to indicate the status of the network.
  • a prediction module 903 configured to estimate network state information according to the network state data, and according to the network state The information predicts the next time state information of the network, where the network state information is used to describe a network link condition of the network, a network element queue condition in the network, and a network data flow distribution of the network.
  • the network status information under the description describes the network link status at time t, the network element queue status in the network, the network data flow distribution, etc.; the network status information at time t-1 (which we can call historical time) Describe the network link situation at time t-1, the network element queue status in the network, the distribution of network data flow, etc.; the network status information at time t+1 (which we can call the next time or future time) Describes the network link status at time t+1, the network element queue status in the network, and the distribution of network data flows, which are not described here.
  • the stream data pre-processing module 905 is configured to receive a data stream currently entering the network, and pre-process the data stream to obtain a data stream pre-processing result.
  • the control module 907 is configured to apply a current network control policy to generate a control action according to the network state information, the network next time state information, and the data flow pre-processing result.
  • the executing module 909 is configured to control the network according to the control action, and obtain feedback information of the network controlled by the control action.
  • the control module 907 is further configured to receive feedback information of the network, and perform real-time adjustment on the current network control policy according to the feedback information to obtain a next-time network control policy.
  • execution module 909 may be implemented by the same hardware or by different hardware implementations.
  • the acquisition module 901 can be implemented by independent software, and the execution module 909 can be a functional unit disposed within the SDN controller.
  • the embodiment of the present invention predicts the next state information of the network according to the network state information, and preprocesses the data stream currently entering the network to obtain a data stream pre-processing result, thereby based on the network state information, the network next-time state information, and the data.
  • a flow pre-processing result generating a control action according to the current network control policy, controlling the network according to the control action to obtain feedback information, wherein the feedback information is obtained by controlling the network according to the control action, and finally according to the The feedback information adjusts the current network control policy in real time to obtain a next-time network control policy.
  • the process is cycled.
  • the real-time adjusted network control strategy is adjusted in real time to obtain the network control strategy at the next moment, and the loop is continued.
  • the network control strategy is always dynamically optimized and adjusted according to the current data and network conditions, which is called adaptive in this paper.
  • the network is controlled according to the network control strategy updated in real time to improve the network control efficiency.
  • the network status data may include current status data of the network, and correspondingly, the network status information
  • the information may include the current state information of the network, and the foregoing prediction module 903 may be specifically configured to: send a state collection command; receive current network state data collected according to the state collection command; and estimate the network according to the current state data of the network.
  • Current state information, the current state information of the network includes the current link status of the network, the current network element queue status in the network, and the current data flow distribution of the network.
  • the network status information may further include network history status information (correspondingly, the network status data may further include network history status data), and the prediction module 903 is further configured to: obtain network history status information from a database. And predicting the next time state information of the network according to the current state information of the network and the historical state information of the network.
  • the prediction module may be specifically configured to execute Operation
  • the stream data pre-processing module 905 may specifically include:
  • An extracting unit configured to extract a feature of the data stream, where the feature includes a packet header quintuple, a data packet length sequence, or a data packet arrival time interval sequence;
  • a mining analysis unit configured to perform time-space data mining processing and/or causal relationship analysis processing on the data stream by using the feature, to obtain a processing result
  • an identifier unit configured to label the data stream according to the processing result to obtain a label of the data stream, where the label of the data stream is the data stream pre-processing result.
  • control module may be specifically configured to perform the following operations:
  • S t represents the network state information
  • l i represents a tag of the data stream
  • i represents an i th data stream
  • ⁇ t represents the current network control policy
  • Express S t , l i is the probability distribution of a under the current network control strategy of the condition variable, where a is a control variable
  • the control action a i is randomly sampled in the probability distribution of the control variable a.
  • Embodiment 5 is a diagrammatic representation of Embodiment 5:
  • FIG. 10 is a structural block diagram of a device 100 for generating a network control policy according to an embodiment of the present invention.
  • the receiving module 1001 is configured to receive network status information of the network, network next time status information, and data stream pre-processing result, where the network status information is used to describe a network link status of the network, and the network in the network a meta-queue situation, a network data flow distribution of the network to which the network belongs, the state information of the next time of the network is predicted according to the network state information, and the pre-processing result of the data stream is to pre-process the data flow currently entering the network. Processed.
  • the control module 1003 is configured to apply a current network control policy to generate a control action according to the network state information, the network next time state information, and the data flow pre-processing result, where the control action is used to control the network and obtain Feedback information of the network controlled by the control action.
  • the control module 1003 is further configured to receive feedback information of the network, and perform real-time adjustment on the current network control policy according to the feedback information to obtain a next-time network control policy.
  • the embodiment of the present invention receives network state information, network next-time state information, and data stream pre-processing result, thereby generating control based on network state information, network next-time state information, and data stream pre-processing results according to current network control policies.
  • the control action is used to control the network to obtain feedback information obtained by controlling the network according to the control action, and finally, receiving the feedback information, and controlling the current network according to the feedback information
  • the policy is adjusted in real time to get the next-time network control strategy.
  • the real-time adjusted network control strategy is adjusted in real time according to the current network state information and data flow, and thus looped. .
  • the network control strategy is always dynamically optimized and adjusted according to the current data and network conditions, which is called adaptive in this paper.
  • the network is controlled according to the network control strategy updated in real time to improve the network control efficiency.
  • the specific information of the network next time state is predicted according to the network state information:
  • the data stream pre-processing result is obtained by pre-processing the data stream currently entering the network, and specifically includes:
  • Extracting characteristics of the data stream including a packet header quintuple, a packet length sequence, or a packet arrival time interval sequence;
  • Embodiment 5 is a diagrammatic representation of Embodiment 5:
  • An embodiment of the present invention provides a network controller, which is applied to an SDN network, and is configured to receive feedback information from the SDN network, and forward the feedback information to a control module, where the control module is configured to The feedback information is adjusted in real time to the current network control policy to obtain a next-time network control policy; wherein the feedback information is controlled according to a control action sent by the control module in the network operation at a last moment. Arrived.
  • the disclosed systems, devices, and methods may be implemented in other manners.
  • the device embodiments described above are merely illustrative.
  • the division of the unit is only a logical function division.
  • there may be another division manner for example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored or not executed.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be in an electrical, mechanical or other form.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.
  • each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
  • the functions may be stored in a computer readable storage medium if implemented in the form of a software functional unit and sold or used as a standalone product.
  • the technical solution of the present invention which is essential or contributes to the prior art, or a part of the technical solution, may be embodied in the form of a software product, which is stored in a storage medium, including
  • the instructions are used to cause a computer device (which may be a personal computer, server, or network device, etc.) or a processor to perform all or part of the steps of the methods described in various embodiments of the present invention.
  • the foregoing storage medium includes: a U disk, a mobile hard disk, and a read only memory. (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Environmental & Geological Engineering (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

本发明提供一种网络控制策略的生成方法包括：获取网络状态数据；根据所述网络状态数据估计网络状态信息，并根据所述网络状态信息预测所述网络下一时刻状态信息；接收当前进入所述网络的数据流，对所述数据流进行预处理得到数据流预处理结果；应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作；根据所述控制动作控制所述网络，并得到经所述控制动作控制的所述网络的反馈信息；根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。通过自适应实时调整网络控制策略，从而提高网络控制效率。

Description

网络控制策略的生成方法、装置及网络控制器 技术领域
本发明实施例涉及网络数据路由领域，并且更具体地，涉及网络控制策略的生成方法、装置及网络控制器，以及其应用的网络数据路由方法及装置。
背景技术
传统网络控制策略生成过程一般是将问题转化为一系列的优化问题。优化问题的输入为网络的拓扑结构，网络中链路的带宽，网络的流量矩阵等；而优化问题的解则一般为优化得到的网络端到端的路径，或者是发送端的发送速率等。传统网络控制策略生成方法具有如下的缺陷：1、优化问题一般通过线性规划或者整数规划得到。受限于线性规划或者整数规划的复杂度，该方法可扩展性比较弱，尤其是随着网元数量，业务类型和流量的不断增加，优化求解将变得过于复杂而无法求解或求解代价(如计算时间)过大。同时这种通过离线优化的方案难以实现动态即时的策略调整。2、当网络状态信息，比如拓扑发生变化(节点增加或减少)，需要重新进行优化问题求解。重新求解优化问题，一方面具有明显滞后性，另一方面需要大量的人力配置以使优化模型适应新的场景。这些缺陷使得由传统网络控制策略生成方法生成的网络控制策略导致网络控制效率低下。
发明内容
本发明实施例提供一种网络控制策略的生成方法、装置及网络控制器，具有自我学习功能，能够自适应的动态调整控制策略。
第一方面，本发明实施例提供，一种网络控制策略的生成方法，应用于SDN网络，所述方法包括：
获取网络的网络状态数据；
根据所述网络状态数据估计网络状态信息，并根据所述网络状态信息预测所述网络下一时刻状态信息，所述网络状态信息用于描述所述网络的网络链路情况、所述网络中网元队列情况、所述网络的网络数据流分布情况；
接收当前进入所述网络的数据流，对所述数据流进行预处理得到数据流预处理结果；
应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作；
根据所述控制动作控制所述网络，并得到经所述控制动作控制的所述网络的反馈信息；
根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。
结合第一方面，在第一方面的第一种可能的实现方式中，所述网络状态数据包括网络当前状态数据，所述网络状态信息包括网络当前状态信息，在所述根据所述网络状态数据估计所述网络状态信息之前，所述方法还包括：
发送状态采集命令；
接收根据所述状态采集命令采集得到的网络当前状态数据；
对应地，所述根据所述网络状态数据估计所述网络状态信息具体包括：
根据所述网络当前状态数据估计所述网络当前状态信息，所述网络当前状态信息包括网络当前链路情况、网络中当前网元队列情况、网络当前数据流分布情况。
结合第一方面的第一种可能的实现方式，在第一方面的第二种可能的实现方式中，所述网络状态信息还包括网络历史状态信息，所述根据所述网络状态信息预测所述网络下一时刻状态信息之前，所述方法还包括：
从数据库获取网络历史状态信息；
对应地，所述根据所述网络状态信息预测所述网络下一时刻状态信息包括：根据所述网络当前状态信息和所述网络历史状态信息预测所述网络下一时刻状态信息。
结合第一方面、第一方面的第一种可能的实现方式或第一方面的第二种可能的实现方式中任意一种可能的实现方式，在第一方面的第三种可能的实现方式中，
所述根据所述网络状态信息预测所述网络下一时刻状态信息为：
Figure PCTCN2016095696-appb-000001
其中
Figure PCTCN2016095696-appb-000002
表示所述网络下一时刻状态信息，其中St-τ表示所述网络状态信息，且0≤τ≤L；其中L为所述网络状态信息的时间窗口长度，其中f为线性函数或非线性函数。
结合第一方面、第一方面的第一种可能的实现方式至第一方面的第三种可能的实现方式中任意一种可能的实现方式，在第一方面的第四种可能的实现方式中，所述对所述数据流进行预处理得到数据流预处理结果，包括：
提取所述数据流的特征，所述特征包括数据包报头五元组、数据包长度序列或数据包到达时间间隔序列；
通过所述特征对所述数据流进行时空数据挖掘处理和/或因果关系分析处理，得到处理结果；
根据所述处理结果对所述数据流打标签得到所述数据流的标签，所述数据流的标签为所述数据流预处理结果。
结合第一方面的第四种可能的实现方式，在第二方面的第五种可能的实现方式中，所述应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作为：
Figure PCTCN2016095696-appb-000003
其中
Figure PCTCN2016095696-appb-000004
表示所述网络下一时刻状态信息，其中St表示所述网络状态信息，其中li表 示所述数据流的标签，其中i表示第i个数据流，其中πt表示所述当前网络控制策略，其中
Figure PCTCN2016095696-appb-000005
表示以
Figure PCTCN2016095696-appb-000006
St，li作为条件变量的当前网络控制策略下a的概率分布，其中a为控制变量，在所述控制变量a的概率分布中随机采样得到所述控制动作ai。
第二方面，本发明实施例提供一种网络控制策略的生成方法，应用于SDN网络，所述方法包括：
接收网络的网络状态信息，网络下一时刻状态信息，数据流预处理结果；其中，所述网络状态信息用于描述所述网络的网络链路情况、所述网络中网元队列情况、所述网络的网络数据流分布情况，所述网络下一时刻状态信息是根据所述网络状态信息进行预测得到的，所述数据流预处理结果是对当前进入网络的数据流进行预处理得到的；
应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作；所述控制动作用于控制所述网络并得到经所述控制动作控制的所述网络的反馈信息；
接收所述网络的反馈信息，并根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。
结合第二方面，在第二方面的第一种可能的实现方式中，所述网络下一时刻状态信息是根据所述网络状态信息进行预测得到的，其中所述预测的方法为：
Figure PCTCN2016095696-appb-000007
其中
Figure PCTCN2016095696-appb-000008
表示所述网络下一时刻状态信息，其中St-τ表示所述网络状态信息，且0≤τ≤L；其中L为所述网络状态信息的时间窗口长度，其中f为线性函数或非线性函数。
结合第二方面或第二方面的第一种可能的实现方式，在第二方面的第二种可能的实现方式中，所述数据流预处理结果是对当前进入网络的数据流进行预处理得到的，其中所述预处理的方法包括：
提取所述数据流的特征，所述特征包括数据包报头五元组、数据包长度序列或数据包到达时间间隔序列；
通过所述特征对所述数据流进行时空数据挖掘处理和/或因果关系分析处理，得到处理结果；
根据所述处理结果对所述数据流打标签得到所述数据流的标签，所述数据流的标签为所述数据流预处理结果。
第三方面，本发明实施例提供一种网络控制策略的生成方法，应用于SDN网络，所述方法包括：从所述SDN网络接收反馈信息，并将所述反馈信息转发给控制模块，用于使所述控制模块根据所述反馈信息对当前网络控制策略进行实时调整得到下一时刻网络控制策略；其中，所述反馈信息是根据上一时刻网络运行中所述控制模块发出的控制动作对所述网络进行控制得到的。
第四方面，本发明实施例提供一种网络控制策略的生成装置，应用于SDN网络，所述装置包括：
获取模块，用于获取网络的网络状态数据；
预测模块，用于根据所述网络状态数据估计网络状态信息，并根据所述网络状态信息预测所述网络下一时刻状态信息，所述网络状态信息用于描述所述网络的网络链路情况、所述网络中网元队列情况、所述网络的网络数据流分布情况；
流数据预处理模块，用于接收当前进入所述网络的数据流，对所述数据流进行预处理得到数据流预处理结果；
控制模块，用于应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作；
执行模块，用于根据所述控制动作控制所述网络，并得到经所述控制动作控制的所述网络的反馈信息；
所述控制模块还用于接收所述网络的反馈信息，并根据根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。
结合第四方面，在第四方面的第一种可能的实现方式中，所述网络状态数据包括网络当前状态数据，所述网络状态信息包括网络当前状态信息，所述预测模块具体用于：
发送状态采集命令；
接收根据所述状态采集命令采集得到的网络当前状态数据；
根据所述网络当前状态数据估计所述网络当前状态信息，所述网络当前状态信息包括网络当前链路情况、网络中当前网元队列情况、网络当前数据流分布情况。
结合第四方面的第一种可能的实现方式，在第四方面的第二种可能的实现方式中，所述网络状态信息还包括网络历史状态信息，所述预测模块还用于：从数据库获取网络历史状态信息；并根据所述网络当前状态信息和所述网络历史状态信息预测所述网络下一时刻状态信息。
结合第四方面、第四方面的第一种可能的实现方式或第四方面的第二种可能的实现方式中任意一种可能的实现方式，在第四方面的第三种可能的实现方式中，所述预测模块具体用于：
Figure PCTCN2016095696-appb-000009
其中
Figure PCTCN2016095696-appb-000010
表示所述网络下一时刻状态信息，其中St-τ表示所述网络状态信息，且0≤τ≤L；其中L为所述网络状态信息的时间窗口长度，其中f为线性函数或非线性函数。
结合第四方面、第四方面的第一种可能的实现方式至第四方面的第三种可能的实现方式中任意一种可能的实现方式，在第四方面的第四种可能的实现方式中，所述流数据预处理模块包括：
提取单元，用于提取所述数据流的特征，所述特征包括数据包报头五元组、数据包长度序列或数据包到达时间间隔序列；
挖掘分析单元，用于通过所述特征对所述数据流进行时空数据挖掘处理和/或因果关系分析处理，得到处理结果；
标识单元，用于根据所述处理结果对所述数据流打标签得到所述数据流的标签，所述数据流的标签为所述数据流预处理结果。
结合第四方面的第四种可能的实现方式，在第四方面的第五种可能的实现方式中，所述控制模块具体用于：
Figure PCTCN2016095696-appb-000011
其中
Figure PCTCN2016095696-appb-000012
表示所述网络下一时刻状态信息，其中St表示所述网络状态信息，其中li表示所述数据流的标签，其中i表示第i个数据流，其中πt表示所述当前网络控制策略，其中
Figure PCTCN2016095696-appb-000013
表示以
Figure PCTCN2016095696-appb-000014
St，li作为条件变量的当前网络控制策略下a的概率分布，其中a为控制变量，在所述控制变量a的概率分布中随机采样得到所述控制动作ai。
第五方面，本发明实施例提供一种网络控制策略的生成装置，应用于SDN网络，所述装置包括：
接收模块，用于接收网络的网络状态信息，网络下一时刻状态信息，数据流预处理结果；其中，所述网络状态信息用于描述所述网络的网络链路情况、所述网络中网元队列情况、所述网络的网络数据流分布情况，所述网络下一时刻状态信息是根据所述网络状态信息进行预测得到的，所述数据流预处理结果是对当前进入网络的数据流进行预处理得到的；
控制模块，用于应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作，所述控制动作用于控制所述网络并得到经所述控制动作控制的所述网络的反馈信息；
所述控制模块还用于接收所述网络的反馈信息，并根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。
结合第五方面，在第五方面的第一种可能的实现方式中，所述网络下一时刻状态信息是根据所述网络状态信息进行预测得到的具体为：
Figure PCTCN2016095696-appb-000015
其中
Figure PCTCN2016095696-appb-000016
表示所述网络下一时刻状态信息，其中St-τ表示所述网络状态信息，且0≤τ≤L；其中L为所述网络状态信息的时间窗口长度，其中f为线性函数或非线性函数。
结合第五方面或第五方面的第一种可能的实现方式，在第五方面的第二种可能的实现方式中，所述数据流预处理结果是对当前进入网络的数据流进行预处理得到的具体包括：
提取所述数据流的特征，所述特征包括数据包报头五元组、数据包长度序列或数据包到达时间间隔序列；
通过所述特征对所述数据流进行时空数据挖掘处理和/或因果关系分析处理，得到处理结果；
根据所述处理结果对所述数据流打标签得到所述数据流的标签，所述数据流的标签为所述数据流预处理结果。
第六方面，本发明实施例提供一种网络控制器，应用于SDN网络，用于从所述SDN网络接收反馈信息，并将所述反馈信息转发给控制模块，用于使所述控制模块根据所述反馈信息对当前网络控制策略进行实时调整得到下一时刻网络控制策略；其中，所述反馈信息是根据上一时刻网络运行中所述控制模块发出的控制动作对所述网络进行控制得到的。
本发明中根据网络状态信息预测网络下一时刻状态信息，并对当前进入所述网络的数据流进行预处理得到数据流预处理结果，从而应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作，根据该控制动作控制所述网络以得到反馈信息，所述反馈信息是根据所述控制动作对所述网络进行控制得到的，最后根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。这样，网络控制策略总是动态的根据当前的数据及网络情况进行实时的优化和调整，即本文所称自适应。每一个时刻都有针对本时刻网络及数据所需要的最优的网络控制策略，根据实时更新的网络控制策略对网络进行控制以此提高网络控制效率。
附图说明
为了更清楚地说明本发明实施例的技术方案，下面将对本发明实施例中所需要使用的附图作简单地介绍，显而易见地，下面所描述的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1是部署本发明实施例提供的网络控制策略的生成方法的数据传输网络架构图；
图2是本发明实施例提供的网络控制策略的生成方法200的示意性流程图；
图3是本发明实施例提供的执行网络控制策略的生成方法200的结构图示例；
图4是本发明实施例提供的执行网络控制策略的生成方法200的另一结构图示例；
图5是本发明实施例提供的执行网络控制策略的生成方法200的又一结构图示例；
图6是本发明实施例提供的网络控制策略的生成方法600的示意性流程图；
图7是本发明实施例提供的执行网络控制策略的生成方法600的结构图示例；
图8是本发明实施例提供的网络控制策略的生成方法800的示意性流程图；
图9是本发明实施例提供的网络控制策略的生成装置900的结构框图；
图10是本发明实施例提供的网络控制策略的生成装置1000的结构框图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所述的实施例是本发明的一部分实施例，而不是全部实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例，都应属于本发明保护的范围。
本发明所提供的网络控制策略的生成方法可以在如图1所述的数据传输网络100中部署和实施。所述数据传输网络100包括有中央控制服务器110，源服务器120，网络交换机130，传输网络140，客户端设备150，预测服务器160，及训练服务器170。
所述中央控制服务器110为所述传输网络100的核心处理设备，其分别与网络交换机130，传输网络140，客户端设备150，预测服务器160及训练服务器170相互连接，用于与网络交换机130，传输网络140，训练服务器170及预测服务器160进行信息交流，指令传递及网络数据传输。同时，源服务器120，网络交换机130，传输网络140，客户端设备150，训练服务器170及预测服务器160之间可以根据不同的业务需求及功能需求相互连接，进行信息及指令的交互。以下，将详细介绍各个设备的功能及组成。
中央控制服务器110主要由处理器、存储器及数据接口等主要部件组成，处理器主要通过调用存储在存储设备中的处理程序以完成相应处理功能，而数据接口则主要负责中央服务器110内部的各个器件之间以及中央处理服务器110与外部组件之间的数据收发。所述的处理器，可以是通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现成可编程门阵列(Field Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本发明以下实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本发明实施例所公开的方法中的步骤可以直接体现为硬件译码处理器执行完成，或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存取存储器(Random Access Memory，RAM)、闪存、只读存储器(Read-Only Memory，ROM)、可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。
所述中央控制服务器110通过数据接口接受网络交互机130上报的路由请求，该路由请求中包括了将要路由的当前数据流信息，并将所述路由请求中的当前数据流信息通过数据接口发送至预测服务器160。预测服务器160根据当前数据流的信息，将网络下一 时刻状态信息作为预测结果发送给中央控制服务器110，中央控制服务器110的中央处理器将调用存储器中预存的处理程序，根据预测服务器160返回的预测结果及调用的存储器中存储的当前网络控制策略，生成控制动作该控制动作中包含与当前数据流的类型匹配的最优的传输路径，随后，中央控制服务器110将根据所述最优的传输路径配置针对当前数据流的路由，并将配置好的路由发送给网络交换机130。此外，所述中央控制服务器110还用于通过其数据接口收集由所述网络交换机130上报的已经完成路由的历史数据流信息，并根据路由前数据和路由后数据得到反馈信息，并将反馈信息发送给训练服务器170，训练服务器170根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。
源服务器120主要由处理器、存储器及数据接口等主要部件组成，处理器主要用于通过数据接口将存储在存储器内的数据经过网络交换机130及传输网络140发送给客户端设备150。
网络交换机130，其主要任务是对新进入网络的数据流，即当前数据流，提取数据流第一个数据包的报头，并上报给中央控制服务器110，等待并根据中央控制服务器110计算和配置的路由对当前数据流的数据包进行转发。同时，网络交换机130还将收集已完成路由的历史数据流信息的收集，并上报给中央控制服务器110。
传输网络140，其由传输线缆和多个网络节点构成，主要用于根据网络交换机130的指令进行数据传输。
客户端设备150，由处理器、存储器及数据接口等主要部件组成，处理器主要用于将客户的数据请求通过数据接口及网络交换机130发送至源服务器120，并将源服务器120通过网络交换机130及传输网络140发送的数据内容存储在存储器中。
预测服务器160，可以是独立的处理设备，例如由独立的处理器，存储器，数据接口组成，所述处理器主要用于调用存储在所述存储器中的处理程序对通过数据接口接收到的当前数据流的信息进行处理，以对网络下一时刻状态信息进行预测，并将预测得到的网络下一时刻状态信息通过数据接口发送至中央控制服务器110。其中，所述当前数据流的信息来自于中央控制服务器110。此外，预测服务器160，也可以是依附于中央控制服务器110的一个工作单元，即与中央控制服务器110共用相同存储器，数据接口，但采用不同的处理器，此时，预测服务器160与所述中央控制服务器110之间的数据传输主要通过中央控制服务器110内部的数据总线完成。此外，所述预测服务器160也可以和中央控制服务器110共用相同的处理器，所述共用的方式可以是通过利用处理器中不 同的处理资源进行同步(并行处理)或者异步(分时处理)数据处理，或者是利用处理器中相同的处理资源进行异步(分时处理)处理。此外，所述预测服务器160也可以通过虚拟机来实现，即由中央控制服务器110利用自身的处理器，存储器，数据接口，通过调用存储在存储器中的程序模拟出具有完整的硬件系统功能的、运行在一个完全隔离环境中的完整计算机系统，该计算机系统可以独立完成所有预测服务器160所需完成的工作而不影响中央控制服务器110自身的运行。可以理解，所述预测服务器160可以进一步简化为存储在中央控制服务器110的存储器中的一个软件程序，该软件程序由中央控制服务器110在适合的时机调用，以实现上述的网络下一时刻状态信息预测功能。
训练服务器170，可以是独立的处理设备，由处理器、存储器及数据接口等主要部件组成，所述处理器主要用于根据接收到的反馈信息对调用的存储于所述存储器中的当前网络控制策略进行实时调整得到下一时刻网络控制策略，，并将该下一时刻网络控制策略存储在存储器中供下一次循环的调用。与预测服务器160相同，所述训练服务器170也可以是依附于中央控制服务器110的一个工作单元，即与中央控制服务器110共用相同存储器，数据接口，但采用不同的处理器，此时，训练服务器170与中央控制服务器110之间的数据传输主要通过中央控制服务器110内部的数据总线完成。此外，训练服务器170也可以和中央控制服务器110共用相同的处理器，所述共用的方式可以是通过利用处理器的不同的处理资源进行同步(并行处理)或者异步(分时处理)的数据处理，或者是利用处理器中相同的处理资源进行异步(分时处理)数据处理。此外，训练服务器170也可以通过虚拟机来实现，即由中央控制服务器110利用自身的处理器，存储器，数据接口，通过调用存储在所述存储器中的程序模拟出具有完整的硬件系统功能的、运行在一个完全隔离环境中的完整计算机系统，该计算机系统可以独立完成所有训练服务器170所需完成的工作。可以理解，训练服务器170可以进一步简化为存储在中央控制服务器110的存储器中的一个软件程序，该软件程序由所述中央控制服务器110在适合的时机调用，以实现上述的对当前网络控制策略的实时调整及实时更新。
以上为实现本发明网络数据流类型检测的硬件环境，但需要注意的是，上述硬件环境并不是执行本发明的唯一方式，其可以根据硬件资源，以及业务需求，即数据传输需求动态的调整，例如将央控制服务器110，网络交换机130，预测服务器160及训练服务器170集中在一个数据控制中心，以提高处理效率。
即，本发明的方法可以基于上述的应用环境来进行部署及实施。以下将结合附图来介绍本发明网络数据流类型检测方法。
实施例一：
图2是根据本发明实施例提供的网络控制策略的生成方法200的示意性流程图。该网络控制策略的生成方法200应用于SDN网络。
S201，获取网络的网络状态数据。
所述网络状态数据是用于表示网络状态的原始数据。
S203，根据所述网络状态数据估计网络状态信息，并根据所述网络状态信息预测所述网络下一时刻状态信息，所述网络状态信息用于描述所述网络的网络链路情况、所述网络中网元队列情况、所述网络的网络数据流分布情况。
需要说明书的，我们从具有低层次物理含义的网络状态数据中估计出具有高层次物理含义的网络状态信息，这个网络状态信息是带有时效性的，即t时刻(我们可以称之为当前时刻)下的网络状态信息描述的是t时刻的网络链路情况、网络中网元队列情况、网络数据流分布等情况；在t-1时刻(我们可以称之为历史时刻)下的网络状态信息描述的是t-1时刻的网络链路情况、网络中网元队列情况、网络数据流分布等情况；在t+1时刻(我们可以称之为下一时刻或将来时刻)下的网络状态信息描述的是t+1时刻的网络链路情况、网络中网元队列情况、网络数据流分布等情况，此处不再赘述。
S205，接收当前进入所述网络的数据流，对所述数据流进行预处理得到数据流预处理结果。
具体的，所述预处理可以包括但不限于通过聚类、分类、回归或因果分析等方式进行处理。
S207，应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作。
S209，根据所述控制动作控制所述网络，并得到经所述控制动作控制的所述网络的反馈信息。
S211，根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。
需要说明的是，本发明实施例提供的网络控制策略的生成方法200中虽然以S201、S203、S205……等依次增大的数字进行步骤描述，但这些数字并不对本方案中的具体步骤顺序进行限制，例如，在本方案中S203与S205就没有这样的顺序限制，它们可以同 时发生，也可以S203早于S205，还可以S203晚于S205发生。即本发明实施例中的步骤顺序只与数据流的走向相关，不与描述的先后相关。
本发明实施例根据网络状态信息预测网络下一时刻状态信息，并对当前进入所述网络的数据流进行预处理得到数据流预处理结果，从而应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作，根据该控制动作控制所述网络以得到反馈信息，所述反馈信息是根据所述控制动作对所述网络进行控制得到的，最后根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。在下一次网络数据过来时，循环此过程，最后基于下一次的反馈信息对本次实时调整后得到的下一时刻网络控制策略再进行相应地实时调整，并由此循环下去。这样，网络控制策略总是动态的根据当前的数据及网络情况进行实时的优化和调整，即本文所称自适应。每一个时刻都有针对本时刻网络及数据所需要的最优的网络控制策略，根据实时更新的网络控制策略对网络进行控制以此提高网络控制效率。
具体的，上述实施例可以在图3所示的结构上实现。例如，在控制系统300中，通过周期性的或者随机触发的向网络中发出采集网络状态数据命令(图中用线段①表示)，SDN控制器基于命令收集这些网络状态数据命令并转发(图中用线段②表示)给相应的处理装置，由相应的处理装置根据网络状态数据估计出网络状态信息，并由网络状态信息预测出网络下一时刻的状态信息，值得说明的是，对应于网络下一时刻的状态信息，此处的网络状态信息用于描述与下一时刻不同的网络状态信息，比如网络历史时刻状态信息或者网络当前状态信息。同时的，或者早于，或者晚于网络状态数据收集及网络状态信息估计等动作的时序，SDN控制器将进入网络的当前数据流实时的转发(图中用线段③表示)给另一个相应的处理装置，这个相应的处理装置对当前进入网络的数据流进行预处理从而得到数据流预处理结果，之所以要强调数据流预处理的动作与网络状态信息估计及网络下一时刻状态信息预测动作的先后顺序可以是同时，或早于或晚于，是因为在现实方案的运行中，原本对此处并不做额外限定，因此无论以何顺序执行上述动作均在本发明实施例的保护范围之内，不再赘述。网络状态信息、网络下一时刻的状态信息及数据流预处理结果作为控制模块的输入信息，并且，控制模块将调用缓存设备中或存储器中的当前网络控制策略，利用网络状态信息、网络下一时刻的状态信息及数据流预处理结果这些信息根据当前网络控制策略生成控制动作，并将该控制动作输出给SDN 控制器(图中用线段④表示)，SDN控制器根据该控制动作对网络执行具体的控制命令。网络将根据该控制动作产生的控制效果作为反馈信息反馈(图中用线段⑤表示)给SDN控制器，SDN控制器将该反馈信息转发(图中用线段⑤表示)给控制模块，控制模块根据反馈信息对当前网络控制策略进行实时调整得到下一时刻网络控制策略，并由控制模块的缓存设备对该实时的网络控制策略进行缓存或将其存储在存储器中留待下一次循环的备用。这样，第一次循环生成的控制动作是针对该次循环时刻或时期下网络所需要的最优的控制动作，这个控制动作被执行之后产生本次反馈信息，由反馈信息得到的下一时刻网络控制策略，这个下一时刻网络控制策略将作为下一次循环中的当前网络控制策略进入下一次循环，由此展开接下来第二次循环，第三次循环，第四次循环……以保证在每一个当前时刻都有针对该当前时刻下的网络状态及数据流信息进行最优的控制从而提高整个控制系统的控制效率，不再赘述。
值得说明的是，上述图3所示的结构是本方法的一种示例，并不作为唯一的结构限制，凡是依本发明实施例提供的网络控制策略的生成方法200进行网络控制策略的生成或网络控制等，就应该被纳入本发明保护范围内，不再赘述。
具体的，如S203步骤所解释，网络状态信息是带有时效性的，对应地，网络状态数据也应该是带有时效性的，即可以有当前时刻的网络状态数据及网络状态信息，还可以有历史时刻的网络状态数据和网络状态信息。
具体来看，所述网络状态数据包括网络当前状态数据，所述网络状态信息包括网络当前状态信息，在上述实施例提供的方法的基础上，本发明实施例提供另一种可选的实施方式：
在S203根据所述网络状态数据估计所述网络状态信息之前，所述方法还包括：
发送状态采集命令；
接收根据所述状态采集命令采集得到的网络当前状态数据；
对应地，S203、所述根据所述网络状态数据估计所述网络状态信息具体包括：
根据所述网络当前状态数据估计所述网络当前状态信息，所述网络当前状态信息包括网络当前链路情况、网络中当前网元队列情况、网络当前数据流分布情况。
如上所述，所述网络状态信息还可以包括网络历史状态信息(同理，所述网络状态数据也还可以包括网络历史状态数据)，所述根据所述网络状态信息预测所述网络下一时刻状态信息之前，所述方法还可以包括：
从数据库获取网络历史状态信息；这里的数据库可以是预存在存储器中数据信息。
对应地，所述根据所述网络状态信息预测所述网络下一时刻状态信息包括：根据所 述网络当前状态信息和所述网络历史状态信息预测所述网络下一时刻状态信息。
具体的，根据网络状态数据估计网络状态信息，并根据网络状态信息预测网络下一时刻状态信息可以通过如下方式实现：
Figure PCTCN2016095696-appb-000017
其中
Figure PCTCN2016095696-appb-000018
表示所述网络下一时刻状态信息，其中St-τ表示所述网络状态信息，且0≤τ≤L；其中L为所述网络状态信息的时间窗口长度，其中f为线性函数或非线性函数。
其中，需要说明的是，根据网络状态数据估计网络状态信息的方法，现实中比较常见，我们通过举例来说明，例如：我们可以根据采集到的交换机端口计数器数据，估计交换机的链路利用率；又比如可以通过采集到的交换机队列计数器数据，估计交换机延迟；再比如通过采集到的交换机流表数据，估计网络中的活跃数据流数量等等。具体的估计方法细节此处无需赘述，本领域相关技术人员应该理解。
具体的，上述方案可以在图4所示的结构上实现，例如，SDN控制器实时的、或周期性的、亦或由事件触发型的向网络发出网络状态数据采集命令(图中用线段①表示)，SDN控制器收集该采集得到的网络状态数据Dt并将Dt转发(图中用线段②表示)给相应的处理装置，在图4的示例中我们叫它状态估计器，该状态估计器根据Dt估计出网络状态信息St，此处的St用来表示网络当前状态信息，St作为状态预测器的输入(在另一种实施例中St和来自数据库中的网络历史状态信息St-1,St-2,...一起作为状态预测器的输入)，状态预测器根据上述输入信息进行预测得到网络下一时刻状态信息
Figure PCTCN2016095696-appb-000019
具体的预测方法可以为上述方法，此处的f具体表示的函数形式此处不做限定。值得说明的是，我们根据当前时刻或根据当前时刻及历史时刻的网络状态信息预测得到不一定只能是下一时刻的网络状态信息，即上文所述的网络下一时刻状态信息，还有可能是下两个时刻，三个时刻……，我们可以管这个状态信息叫做未来时刻的状态信息，其预测思路与预测下一时刻的思路一致，另外，如图4所示，我们将估计得到的St存储在数据库中，具体可以存储在数据库缓存设备中或存储器中，以形成下一时刻的网络历史状态信息留待备用。
值得说明的是，上述图4所示的结构是本方法的一种示例，并不作为唯一的结构限制，凡是依本发明实施例提供的网络控制策略的生成方法200进行网络控制策略的生成或网络控制等，就应该被纳入本发明保护范围内，不再赘述。
具体的，S205、对所述数据流进行预处理得到数据流预处理结果可以按如下方法实施：
提取所述数据流的特征，所述特征包括但不限于数据包报头五元组、数据包长度序列或数据包到达时间间隔序列；
通过所述特征对所述数据流进行时空数据挖掘处理和/或因果关系分析处理，得到处 理结果；
根据所述处理结果对所述数据流打标签得到所述数据流的标签，所述数据流的标签为所述数据流预处理结果，该预处理结果可以是标签本身，此处所述的标签可以是标识，或者别的相关的用于区分流类型和/或流间关系的标记。
具体的，所述步骤S207，应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作具体可以以下述方式实现：
Figure PCTCN2016095696-appb-000020
其中
Figure PCTCN2016095696-appb-000021
表示所述网络下一时刻状态信息，其中St表示所述网络状态信息，其中li表示所述数据流的标签，其中i表示第i个数据流，其中πt表示所述当前网络控制策略，其中
Figure PCTCN2016095696-appb-000022
表示以
Figure PCTCN2016095696-appb-000023
St，li作为条件变量的当前网络控制策略下a的概率分布，其中a为控制变量，在所述控制变量a的概率分布中随机采样得到所述控制动作ai。
具体的，上述方案可以在图5所示的结构上实现，例如，SDN控制器将当前进入网络的数据流转发给相应的处理装置，在图5示例中我们叫它特征提取模块，特征提取模块提取所述数据流的特征，并利用提取的特征进行时空数据挖掘处理和/或因果关系分析处理。具体可以如下操作：
对数据流进行特征提取，具体可以通过Xi＝F(fi)进行特征提取，其中fi表示第i个数据流，F表示特征提取函数，Xi表示从数据流i中提取到的特征向量。将提取到的特征用x表示。x是一个高维的向量，如下所示：X＝{x1,x2...xn}T。
例如，一个从数据流中提取到的特征中，其所包含的时空信息可以包括但不限于：
源端地址(Source IP)，32位二进制串；目的端地址(Destination IP)，32位二进制串；服务器端口(Server Port)，16位二进制串；客户端端口(Client Port)，16位二进制串；传输协议(Protocol)，类别类型；发送时间(Starting time)，实数类型；数据包长度序列(Packet length)，整数类型；数据包到达时间间隔序列(Packet inter-arrival time)：实数类型等。
所述预处理的具体方法包括：关于数据流的时空数据挖掘和/或因果关系挖掘。
其中，所述时空数据挖掘具体包括但不限于：
聚类(Clustering)：以数据流提取的特征全集或者是部分特征集合，作为特征向量，代表该数据流，对特征向量进行聚类，可以获得数据流的时空信息。比如，通过聚类，可以获得协同流信息(Co-flow information)，即是可以分析哪一些数据流可能属于同一个 任务。
分类(Classification)：数据流往往可以根据其本身的性质，进行分类。比如，根据数据流长度(flow size)或者持续时间(duration)，可以将数据流分为大象流(elephant flow)和老鼠流(mice flow)。又比如，根据数据流产生的应用，可以将数据流分为视频流(video stream)，数据备份(data backup)等等。通过离线人工标签的方式，我们可以获得训练数据，根据训练数据训练分类器(Classifier)。利用分类器，在线实时地对数据流进行分类。
回归(Regression)：回归与分类有类似的地方，即是根据训练数据学习得到函数g，函数g将特征作为输入，函数g输出数据流某个特征，比如我们可以建立一个回归模型，利用数据流的特征，估计数据流的长度。
其中，所述因果关系挖掘包括但不限于：
因果分析(Causal analysis)：数据流之间可能还存在因果关系。比如，某些客户端可能通过向服务器端发送一些请求，而服务器端则相应地做出响应。在这种情形下，客户端发送到服务器端以及服务器端发送到客户端的数据流存在因果关系。假设我们研究网络中频繁交流的IP地址对(Source IP,Destination IP)之间发送的数据流之间是否有因果关系，我们可以将研究的时间划分为多个时间段，定义
Figure PCTCN2016095696-appb-000024
为t时间段内，从SourceIP发送到DestinationIP的数据流数量。我们研究
Figure PCTCN2016095696-appb-000025
与
Figure PCTCN2016095696-appb-000026
τ≥0是否有因果关系：
Figure PCTCN2016095696-appb-000027
通过上述处理，得到处理结果，并根据所述处理结果对所述数据流打标签得到即图中所示标签li，即所述数据流预处理结果。
值得说明的是，上述图5所示的结构是本方法的一种示例，并不作为唯一的结构限制，凡是依本发明实施例提供的网络控制策略的生成方法200进行网络控制策略的生成或网络控制等，就应该被纳入本发明保护范围内，不再赘述。
实施例二：
图6是根据本发明实施例提供的网络控制策略的生成方法600的示意性流程图。该网络控制策略的生成方法600应用于SDN网络。
S601，接收网络的网络状态信息，网络下一时刻状态信息，数据流预处理结果。
其中，所述网络状态信息用于描述所述网络的网络链路情况、所述网络中网元队列情况、所述网络的网络数据流分布情况，所述网络下一时刻状态信息是根据所述网络状态信息进行预测得到的，所述数据流预处理结果是对当前进入网络的数据流进行预处理 得到的。
S603，应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作。
所述控制动作用于控制所述网络并得到经所述控制动作控制的所述网络的反馈信息。
S605，接收所述网络的反馈信息，并根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。
具体的，所述网络下一时刻状态信息是根据所述网络状态信息进行预测得到的，其中所述预测的方法具体可以为：
Figure PCTCN2016095696-appb-000028
其中
Figure PCTCN2016095696-appb-000029
表示所述网络下一时刻状态信息，其中St-τ表示所述网络状态信息，且0≤τ≤L；其中L为所述网络状态信息的时间窗口长度，其中f为线性函数或非线性函数。
具体的，所述数据流预处理结果是对当前进入网络的数据流进行预处理得到的，其中所述预处理的方法具体可以为：
提取所述数据流的特征，所述特征包括数据包报头五元组、数据包长度序列或数据包到达时间间隔序列；
通过所述特征对所述数据流进行时空数据挖掘处理和/或因果关系分析处理，得到处理结果；
根据所述处理结果对所述数据流打标签，得到所述数据流预处理结果。
本发明实施例通过接收网络状态信息，网络下一时刻状态信息，数据流预处理结果，从而基于网络状态信息、网络下一时刻状态信息和数据流预处理结果，并根据当前网络控制策略生成控制动作，所述控制动作用于控制所述网络以得到根据所述控制动作对所述网络进行控制得到的反馈信息，最后，接收所述反馈信息，并根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。在下一次网络数据过来时，循环此过程，最后基于下一次的反馈信息对本次实时调整后的网络控制策略再进行实时调整得到再下一个时刻网络控制策略，并由此循环下去。这样，网络控制策略总是动态的根据当前的数据及网络情况进行实时的优化和调整，即本文所称自适应。每一个时刻都有针对本时刻网络及数据所需要的最优的网络控制策略，根据实时更新的网络控制策略对网络进行控制以此提高网络控制效率。
具体的，上述实施例可以在图3所示的结构上实现。例如，在控制系统300中完整的方案实现过程应该是：通过周期性的或者随机触发的向网络中发出采集网络状态数据命令，SDN控制器基于命令收集这些网络状态数据命令并转发给相应的处理装置，由相应的处理装置根据网络状态数据估计出网络状态信息，并由网络状态信息预测出网络下一时刻的状态信息，值得说明的是，对应于网络下一时刻的状态信息，此处的网络状态信息用于描述与下一时刻不同的网络状态信息，比如网络历史时刻状态信息或者网络当前状态信息。同时的，或者早于，或者晚于网络状态数据收集及网络状态信息估计等动作的时序，SDN控制器将进入网络的当前数据流实时的转发给另一个相应的处理装置，这个相应的处理装置对当前进入网络的数据流进行预处理从而得到数据流预处理结果，之所以要强调数据流预处理的动作与网络状态信息估计及网络下一时刻状态信息预测动作的先后顺序可以是同时，或早于或晚于，是因为在现实方案的运行中，原本对此处并不做额外限定，因此无论以何顺序执行上述动作均在本发明实施例的保护范围之内，不再赘述。网络状态信息、网络下一时刻的状态信息及数据流预处理结果作为控制模块的输入信息，并且，控制模块将调用缓存设备中或存储器中的当前网络控制策略，利用网络状态信息、网络下一时刻的状态信息及数据流预处理结果这些信息根据当前网络控制策略生成控制动作，并将该控制动作输出给SDN控制器，SDN控制器根据该控制动作对网络执行具体的控制命令。网络将根据该控制动作产生的控制效果作为反馈信息反馈给SDN控制器，SDN控制器将该反馈信息转发给控制模块，控制模块根据反馈信息对当前网络控制策略进行实时调整得到下一时刻网络控制策略，并由控制模块的缓存设备对该下一时刻网络控制策略进行缓存或将其存储在存储器中留待下一次循环的备用。这样，第一次循环生成的控制动作是针对该次循环时刻或时期下网络所需要的最优的控制动作，这个控制动作被执行之后产生本次反馈信息，由反馈信息得到下一时刻网络控制策略，这个下一时刻网络控制策略将作为下一次循环中的当前网络控制策略进入下一次循环，由此展开接下来第二次循环，第三次循环，第四次循环……以保证在每一个当前时刻都有针对该当前时刻下的网络状态及数据流信息进行最优的控制从而提高整个控制系统的控制效率，不再赘述。本发明实施例所提供的方法执行主体可以是结构300中的控制模块。
更具体的，上述方案还可以在图7所示的结构上实现，例如，控制模块接收网络状态信息，网络下一时刻状态信息以及数据流预处理结果，根据调用的策略πt及接收的信息生成控制动作ai，具体的通过如下方式进行：
Figure PCTCN2016095696-appb-000030
其中
Figure PCTCN2016095696-appb-000031
表示所述网络下一时刻状态信息，其中St表示所述网络状态信息，其中li表示所述数据流的标签，其中i表示第i个数据流，其中πt表示所述当前网络控制策略，其中
Figure PCTCN2016095696-appb-000032
表示以
Figure PCTCN2016095696-appb-000033
St，li作为条件变量的当前网络控制策略下a的概率分布，其中a为控制变量，在所述控制变量a的概率分布中随机采样得到所述控制动作ai，所述控制动作ai服从的
Figure PCTCN2016095696-appb-000034
概率分布。具体的，可以如图中所示采用Q学习的方法，或MDP(马尔科夫决策过程)，或POMDP(部分可观察马尔科夫决策过程)进行。控制动作ai被发送给SDN控制器，用于使SDN控制器去执行该控制动作。网络在接受控制动作ai，从状态st转移到实际的状态st+1时，反馈给控制模块该控制动作所产生的回报Rt＝R(st,ai,st+1)，我们把该回报叫做反馈信息。进一步的，控制模块根据该反馈信息相应地对策略πt进行实时调整得到下一时刻网络控制策略：πt+1＝L(πt,Rt)，其中L表示利用反馈对策略进行调整的学习算法，其中πt+1为下一时刻网络控制策略。具体的，S为有限网络状态信息集合，A为有限控制动作集合，J(π)表示衡量策略优越性的目标函数，P(st+1|st,at)为系统在动作at的作用下，从状态st转移到状态st+1的概率，R(st,at,st+1)为系统从状态st转移到st+1获得的回报，π(a|s)为策略函数，表示在网络状态信息所描述的网络状态s采用控制动作a的概率，μπ(s)为在策略π下状态s的分布，则最佳的策略可以由如下的有限制优化问题获得：
Figure PCTCN2016095696-appb-000035
Figure PCTCN2016095696-appb-000036
Figure PCTCN2016095696-appb-000037
假设在初始化的时候，我们通过离线收集到的数据：
(s0,a0,s′0,R0),(s1,a1,s′1,R1),…,(sn,an,s′n,Rn)
并学习得到初始化的网络控制策略π0，系统在初始化运转的时候通过网络控制策略π0控制网络运行，等收集足够多新的数据集时，我们用新的数据集，重新学习得到新的网络控制策略π1，如此不断循环(这个循环的过程可以离线进行也可以在线进行)，直到我们认为一个足够好的网络控制策略可以被当作上述方法200或600中的当前网络控制策略，则进入上述提供的实施例中的第一个循环。
可选的，如图7所示，在本方案实施例中，我们可以将该反馈信息存储在控制模块的缓存设备中或存储器中，此处不做赘述。
值得说明的是，上述图3和图7所示的结构是本方法的示例，并不作为唯一的结构限制，凡是依本发明实施例提供的网络控制策略的生成方法600进行网络控制策略的生成或网络控制等，就应该被纳入本发明保护范围内，不再赘述。
实施例三：
图8是根据本发明实施例提供的网络控制策略的生成方法800的示意性流程图。该网络控制策略的生成装置800应用于SDN网络。
S801，从所述网络接收反馈信息。
S803，将所述反馈信息转发给控制模块，用于使所述控制模块根据所述反馈信息对当前网络控制策略进行实时调整得到下一时刻网络控制策略。
本发明实施例接收反馈信息，并将所述反馈信息转发给控制模块，用于使所述控制模块根据所述反馈信息对当前网络控制策略进行实时调整得到下一时刻网络控制策略。在下一次网络数据过来时，循环此过程，最后基于下一次的反馈信息对本次实时调整后网络控制策略再进行实时调整得到再下一个时刻网络控制策略，并由此循环下去。这样，网络控制策略总是动态的根据当前的数据及网络情况进行实时的优化和调整，即本文所称自适应。每一个时刻都有针对本时刻网络及数据所需要的最优的网络控制策略，根据实时更新的网络控制策略对网络进行控制以此提高网络控制效率。
具体的，上述实施例可以在图3所示的结构上实现。例如，在控制系统300中完整的方案实现过程应该是：通过周期性的或者随机触发的向网络中发出采集网络状态数据命令，SDN控制器基于命令收集这些网络状态数据命令并转发给相应的处理装置，由相应的处理装置根据网络状态数据估计出网络状态信息，并由网络状态信息预测出网络下一时刻的状态信息，值得说明的是，对应于网络下一时刻的状态信息，此处的网络状态 信息用于描述与下一时刻不同的网络状态信息，比如网络历史时刻状态信息或者网络当前状态信息。同时的，或者早于，或者晚于网络状态数据收集及网络状态信息估计等动作的时序，SDN控制器将进入网络的当前数据流实时的转发给另一个相应的处理装置，这个相应的处理装置对当前进入网络的数据流进行预处理从而得到数据流预处理结果，之所以要强调数据流预处理的动作与网络状态信息估计及网络下一时刻状态信息预测动作的先后顺序可以是同时，或早于或晚于，是因为在现实方案的运行中，原本对此处并不做额外限定，因此无论以何顺序执行上述动作均在本发明实施例的保护范围之内，不再赘述。网络状态信息、网络下一时刻的状态信息及数据流预处理结果作为控制模块的输入信息，并且，控制模块将调用缓存设备中或存储器中的当前网络控制策略，利用网络状态信息、网络下一时刻的状态信息及数据流预处理结果这些信息根据当前网络控制策略生成控制动作，并将该控制动作输出给SDN控制器，SDN控制器根据该控制动作对网络执行具体的控制命令。网络将根据该控制动作产生的控制效果作为反馈信息反馈给SDN控制器，SDN控制器将该反馈信息转发给控制模块，控制模块根据反馈信息对当前网络控制策略进行实时调整得到下一时刻网络控制策略，并由控制模块的缓存设备对该下一时刻网络控制策略进行缓存或将其存储在存储器中留待下一次循环的备用。这样，第一次循环生成的控制动作是针对该次循环时刻或时期下网络所需要的最优的控制动作，这个控制动作被执行之后产生本次反馈信息，由反馈信息得到下一时刻网络控制策略，这个下一时刻网络控制策略将作为下一次循环中的当前网络控制策略进入下一次循环，由此展开接下来第二次循环，第三次循环，第四次循环……以保证在每一个当前时刻都有针对该当前时刻下的网络状态及数据流信息进行最优的控制从而提高整个控制系统的控制效率，不再赘述。本发明实施例所提供的方法执行主体可以是结构300中的SDN控制器。
实施例四：
图9是根据本发明实施例提供的网络控制策略的生成装置900的结构框图。该网络控制策略的生成装置900应用于SDN网络。
获取模块901，用于获取网络的网络状态数据。
所述网络状态数据是用于表示网络状态的原始数据。
预测模块903，用于根据所述网络状态数据估计网络状态信息，并根据所述网络状态 信息预测所述网络下一时刻状态信息，所述网络状态信息用于描述所述网络的网络链路情况、所述网络中网元队列情况、所述网络的网络数据流分布情况。
需要说明书的，我们从具有低层次物理含义的网络状态数据中估计出具有高层次物理含义的网络状态信息，这个网络状态信息是带有时效性的，即t时刻(我们可以称之为当前时刻)下的网络状态信息描述的是t时刻的网络链路情况、网络中网元队列情况、网络数据流分布等情况；在t-1时刻(我们可以称之为历史时刻)下的网络状态信息描述的是t-1时刻的网络链路情况、网络中网元队列情况、网络数据流分布等情况；在t+1时刻(我们可以称之为下一时刻或将来时刻)下的网络状态信息描述的是t+1时刻的网络链路情况、网络中网元队列情况、网络数据流分布等情况，此处不再赘述。
流数据预处理模块905，用于接收当前进入所述网络的数据流，对所述数据流进行预处理得到数据流预处理结果。
控制模块907，用于应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作。
执行模块909，用于根据所述控制动作控制所述网络，并得到经所述控制动作控制的所述网络的反馈信息。
所述控制模块907还用于接收所述网络的反馈信息，并根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。
值得说明的是，执行模块909与获取模块901的关系，可以是同一个硬件来实现，也可以是不同硬件实现。获取模块901可以由独立的软件实现，执行模块909可以为设置在SDN控制器内的功能单元。
本发明实施例根据网络状态信息预测网络下一时刻状态信息，并对当前进入所述网络的数据流进行预处理得到数据流预处理结果，从而基于网络状态信息、网络下一时刻状态信息和数据流预处理结果，根据当前网络控制策略生成控制动作，根据该控制动作控制所述网络以得到反馈信息，所述反馈信息是根据所述控制动作对所述网络进行控制得到的，最后根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。在下一次网络数据过来时，循环此过程，最后基于下一次的反馈信息对本次实时调整后的网络控制策略再进行实时调整得到再下一个时刻网络控制策略，并由此循环下去。这样，网络控制策略总是动态的根据当前的数据及网络情况进行实时的优化和调整，即本文所称自适应。每一个时刻都有针对本时刻网络及数据所需要的最优的网络控制策略，根据实时更新的网络控制策略对网络进行控制以此提高网络控制效率。
具体的，所述网络状态数据可以包括网络当前状态数据，相应的，所述网络状态信 息可以包括网络当前状态信息，则上述预测模块903具体可以用于：发送状态采集命令；接收根据所述状态采集命令采集得到的网络当前状态数据；并根据所述网络当前状态数据估计所述网络当前状态信息，所述网络当前状态信息包括网络当前链路情况、网络中当前网元队列情况、网络当前数据流分布情况。
具体的，所述网络状态信息还可以包括网络历史状态信息(相应的，所述网络状态数据还可以包括网络历史状态数据)，则所述预测模块903还用于：从数据库获取网络历史状态信息；并根据所述网络当前状态信息和所述网络历史状态信息预测所述网络下一时刻状态信息。
具体的，所述预测模块具体可以用于执行
Figure PCTCN2016095696-appb-000038
运算；
其中
Figure PCTCN2016095696-appb-000039
表示预测的所述网络下一时刻状态信息，其中St-τ表示所述网络状态信息，且0≤τ≤L；其中L为所述网络状态信息的时间窗口长度，其中f为线性函数或非线性函数。
作为一种可选的实施方案，所述流数据预处理模块905具体可以包括：
提取单元，用于提取所述数据流的特征，所述特征包括数据包报头五元组、数据包长度序列或数据包到达时间间隔序列；
挖掘分析单元，用于通过所述特征对所述数据流进行时空数据挖掘处理和/或因果关系分析处理，得到处理结果；
标识单元，用于根据所述处理结果对所述数据流打标签得到所述数据流的标签，所述数据流的标签为所述数据流预处理结果。
作为一种可选的实施方案，所述控制模块具体可以用于进行如下操作：
Figure PCTCN2016095696-appb-000040
其中
Figure PCTCN2016095696-appb-000041
表示所述网络下一时刻状态信息，其中St表示所述网络状态信息，其中li表示所述数据流的标签，其中i表示第i个数据流，其中πt表示所述当前网络控制策略，其中
Figure PCTCN2016095696-appb-000042
表示以
Figure PCTCN2016095696-appb-000043
St，li作为条件变量的当前网络控制策略下a的概率分布，其中a为控制变量，在所述控制变量a的概率分布中随机采样得到所述控制动作ai。
实施例五：
图10是根据本发明实施例提供的网络控制策略的生成装置1000的结构框图。
接收模块1001，用于接收网络的网络状态信息，网络下一时刻状态信息，数据流预处理结果；其中，所述网络状态信息用于描述所述网络的网络链路情况、所述网络中网元队列情况、所属网络的网络数据流分布情况，所述网络下一时刻状态信息是根据所述网络状态信息进行预测得到的，所述数据流预处理结果是对当前进入网络的数据流进行预处理得到的。
控制模块1003，用于应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作，所述控制动作用于控制所述网络并得到经所述控制动作控制的所述网络的反馈信息。
所述控制模块1003还用于接收所述网络的反馈信息，并根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。
本发明实施例通过接收网络状态信息，网络下一时刻状态信息，数据流预处理结果，从而基于网络状态信息、网络下一时刻状态信息和数据流预处理结果，并根据当前网络控制策略生成控制动作，所述控制动作用于控制所述网络以得到根据所述控制动作对所述网络进行控制得到的反馈信息，最后，接收所述反馈信息，并根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。在下一次网络数据过来时，循环此过程，最后基于下一次的反馈信息对本次实时调整后的网络控制策略再进行与当时的网络状态信息及数据流情况相应地实时调整，并由此循环下去。这样，网络控制策略总是动态的根据当前的数据及网络情况进行实时的优化和调整，即本文所称自适应。每一个时刻都有针对本时刻网络及数据所需要的最优的网络控制策略，根据实时更新的网络控制策略对网络进行控制以此提高网络控制效率。
其中，所述网络下一时刻状态信息是根据所述网络状态信息进行预测得到的具体为：
Figure PCTCN2016095696-appb-000044
其中
Figure PCTCN2016095696-appb-000045
表示预测的所述网络下一时刻状态信息，其中St-τ表示所述网络状态信息，且0≤τ≤L；其中L为所述网络状态信息的时间窗口长度，其中f为线性函数或非线性函数。
所述数据流预处理结果是对当前进入网络的数据流进行预处理得到的具体包括：
提取所述数据流的特征，所述特征包括数据包报头五元组、数据包长度序列或数据包到达时间间隔序列；
通过所述特征对所述数据流进行时空数据挖掘处理和/或因果关系分析处理，得到处理结果；
根据所述处理结果对所述数据流打标签得到所述数据流的标签，所述数据流的标签为所述数据流预处理结果。
实施例五：
本发明实施例提供一种网络控制器，应用于SDN网络，其特征在于，用于从所述SDN网络接收反馈信息，并将所述反馈信息转发给控制模块，用于使所述控制模块根据所述反馈信息对当前网络控制策略进行实时调整得到下一时刻网络控制策略；其中，所述反馈信息是根据上一时刻网络运行中所述控制模块发出的控制动作对所述网络进行控制得 到的。
需要说明的是，上述所有装置实施例中提供的装置都可以用于实现上述所有方法实施例中的方法，同理，上述所有方法实施例中提供的方法都可以在上述所有装置实施例中提供的装置上运行。方法实施例中详述的方案中细节信息可以用于解释装置实施例中相应特征。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本发明的范围。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统、装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本发明各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本发明的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)或处理器(processor)执行本发明各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器 (ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述，仅为本发明的具体实施方式，但本发明的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本发明揭露的技术范围内，可轻易想到的变化或替换，都应涵盖在本发明的保护范围之内，因此本发明的保护范围应以权利要求的保护范围为准。

Claims (18)

  1. 一种网络控制策略的生成方法，应用于SDN网络，其特征在于，包括：
    获取网络的网络状态数据；
    根据所述网络状态数据估计网络状态信息，并根据所述网络状态信息预测所述网络下一时刻状态信息，所述网络状态信息用于描述所述网络的网络链路情况、所述网络中网元队列情况、所述网络的网络数据流分布情况；
    接收当前进入所述网络的数据流，对所述数据流进行预处理得到数据流预处理结果；
    应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作；
    根据所述控制动作控制所述网络，并得到经所述控制动作控制的所述网络的反馈信息；
    根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。
  2. 根据权利要求1所述的方法，其特征在于，所述网络状态数据包括网络当前状态数据，所述网络状态信息包括网络当前状态信息，在所述根据所述网络状态数据估计所述网络状态信息之前，所述方法还包括：
    发送状态采集命令；
    接收根据所述状态采集命令采集得到的网络当前状态数据；
    对应地，所述根据所述网络状态数据估计所述网络状态信息具体包括：
    根据所述网络当前状态数据估计所述网络当前状态信息，所述网络当前状态信息包括网络当前链路情况、网络中当前网元队列情况、网络当前数据流分布情况。
  3. 根据权利要求2所述的方法，其特征在于，所述网络状态信息还包括网络历史状态信息，所述根据所述网络状态信息预测所述网络下一时刻状态信息之前，所述方法还包括：
    从数据库获取网络历史状态信息；
    对应地，所述根据所述网络状态信息预测所述网络下一时刻状态信息包括：根据所述网络当前状态信息和所述网络历史状态信息预测所述网络下一时刻状态信息。
  4. 根据权利要求1至3任一所述的方法，其特征在于，所述根据所述网络状态信息预测所述网络下一时刻状态信息为：
    Figure PCTCN2016095696-appb-100001
    其中
    Figure PCTCN2016095696-appb-100002
    表示所述网络下一时刻状态信息，其中St-τ表示所述网络状态信息，且0≤τ≤L；其中L为所述网络状态信息的时间窗口长度，其中f为线性函数或非线性函数。
  5. 根据权利要求1至4任一所述的方法，其特征在于，所述对所述数据流进行预处理得到数据流预处理结果，包括：
    提取所述数据流的特征，所述特征包括数据包报头五元组、数据包长度序列或数据包到达时间间隔序列；
    通过所述特征对所述数据流进行时空数据挖掘处理和/或因果关系分析处理，得到处理结果；
    根据所述处理结果对所述数据流打标签得到所述数据流的标签，所述数据流的标签为所述数据流预处理结果。
  6. 根据权利要求5所述的方法，其特征在于，所述应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作为：
    Figure PCTCN2016095696-appb-100003
    其中
    Figure PCTCN2016095696-appb-100004
    表示所述网络下一时刻状态信息，其中St表示所述网络状态信息，其中li表示所述数据流的标签，其中i表示第i个数据流，其中πt表示所述当前网络控制策略，其中
    Figure PCTCN2016095696-appb-100005
    表示以
    Figure PCTCN2016095696-appb-100006
    St，li作为条件变量的当前网络控制策略下a的概率分布，其中a为控制变量，在所述控制变量a的概率分布中随机采样得到所述控制动作ai。
  7. 一种网络控制策略的生成方法，应用于SDN网络，其特征在于，包括：
    接收网络的网络状态信息，网络下一时刻状态信息，数据流预处理结果；其中，所述网络状态信息用于描述所述网络的网络链路情况、所述网络中网元队列情况、所述网络的网络数据流分布情况，所述网络下一时刻状态信息是根据所述网络状态信息进行预测得到的，所述数据流预处理结果是对当前进入网络的数据流进行预处理得到的；
    应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作；所述控制动作用于控制所述网络并得到经所述控制动作控制的所述网络的反馈信息；
    接收所述网络的反馈信息，并根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。
  8. 根据权利要求7所述的方法，其特征在于，所述网络下一时刻状态信息是根据所述网络状态信息进行预测得到的，其中所述预测的方法为：
    Figure PCTCN2016095696-appb-100007
    其中
    Figure PCTCN2016095696-appb-100008
    表示所述网络下一时刻状态信息，其中St-τ表示所述网络状态信息，且0≤τ≤L；其中L为所述网络状态信息的时间窗口长度，其中f为线性函数或非线性函数。
  9. 根据权利要求7或8所述的方法，其特征在于，所述数据流预处理结果是对当前进入网络的数据流进行预处理得到的，其中所述预处理的方法包括：
    提取所述数据流的特征，所述特征包括数据包报头五元组、数据包长度序列或数据包到达时间间隔序列；
    通过所述特征对所述数据流进行时空数据挖掘处理和/或因果关系分析处理，得到处理结果；
    根据所述处理结果对所述数据流打标签得到所述数据流的标签，所述数据流的标签 为所述数据流预处理结果。
  10. 一种网络控制策略的生成装置，应用于SDN网络，其特征在于，包括：
    获取模块，用于获取网络的网络状态数据；
    预测模块，用于根据所述网络状态数据估计网络状态信息，并根据所述网络状态信息预测所述网络下一时刻状态信息，所述网络状态信息用于描述所述网络的网络链路情况、所述网络中网元队列情况、所述网络的网络数据流分布情况；
    流数据预处理模块，用于接收当前进入所述网络的数据流，对所述数据流进行预处理得到数据流预处理结果；
    控制模块，用于应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作；
    执行模块，用于根据所述控制动作控制所述网络，并得到经所述控制动作控制的所述网络的反馈信息；
    所述控制模块还用于接收所述网络的反馈信息，并根据根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。
  11. 根据权利要求10所述的装置，其特征在于，所述网络状态数据包括网络当前状态数据，所述网络状态信息包括网络当前状态信息，所述预测模块具体用于：
    发送状态采集命令；
    接收根据所述状态采集命令采集得到的网络当前状态数据；
    根据所述网络当前状态数据估计所述网络当前状态信息，所述网络当前状态信息包括网络当前链路情况、网络中当前网元队列情况、网络当前数据流分布情况。
  12. 根据权利要求11所述的装置，其特征在于，所述网络状态信息还包括网络历史状态信息，所述预测模块还用于：从数据库获取网络历史状态信息；并根据所述网络当前状态信息和所述网络历史状态信息预测所述网络下一时刻状态信息。
  13. 根据权利要求10至12任一所述的装置，其特征在于，所述预测模块具体用于：
    Figure PCTCN2016095696-appb-100009
    其中
    Figure PCTCN2016095696-appb-100010
    表示所述网络下一时刻状态信息，其中St-τ表示所述网络状态信息，且0≤τ≤L；其中L为所述网络状态信息的时间窗口长度，其中f为线性函数或非线性函数。
  14. 根据权利要求10至13任一所述的装置，其特征在于，所述流数据预处理模块包括：
    提取单元，用于提取所述数据流的特征，所述特征包括数据包报头五元组、数据包长度序列或数据包到达时间间隔序列；
    挖掘分析单元，用于通过所述特征对所述数据流进行时空数据挖掘处理和/或因果关系分析处理，得到处理结果；
    标识单元，用于根据所述处理结果对所述数据流打标签得到所述数据流的标签，所述数据流的标签为所述数据流预处理结果。
  15. 根据权利要求14所述的装置，其特征在于，所述控制模块具体用于：
    Figure PCTCN2016095696-appb-100011
    其中
    Figure PCTCN2016095696-appb-100012
    表示所述网络下一时刻状态信息，其中St表示所述网络状态信息，其中li表示所述数据流的标签，其中i表示第i个数据流，其中πt表示所述当前网络控制策略，其中
    Figure PCTCN2016095696-appb-100013
    表示以
    Figure PCTCN2016095696-appb-100014
    St，li作为条件变量的当前网络控制策略下a的概率分布，其中a为控制变量，在所述控制变量a的概率分布中随机采样得到所述控制动作ai。
  16. 一种网络控制策略的生成装置，应用于SDN网络，其特征在于，包括：
    接收模块，用于接收网络的网络状态信息，网络下一时刻状态信息，数据流预处理结果；其中，所述网络状态信息用于描述所述网络的网络链路情况、所述网络中网元队列情况、所述网络的网络数据流分布情况，所述网络下一时刻状态信息是根据所述网络状态信息进行预测得到的，所述数据流预处理结果是对当前进入网络的数据流进行预处理得到的；
    控制模块，用于应用当前网络控制策略根据所述网络状态信息、所述网络下一时刻状态信息以及所述数据流预处理结果生成控制动作，所述控制动作用于控制所述网络并得到经所述控制动作控制的所述网络的反馈信息；
    所述控制模块还用于接收所述网络的反馈信息，并根据所述反馈信息对所述当前网络控制策略进行实时调整得到下一时刻网络控制策略。
  17. 根据权利要求16所述的装置，其特征在于，所述网络下一时刻状态信息是根据所述网络状态信息进行预测得到的具体为：
    Figure PCTCN2016095696-appb-100015
    其中
    Figure PCTCN2016095696-appb-100016
    表示所述网络下一时刻状态信息，其中St-τ表示所述网络状态信息，且0≤τ≤L；其中L为所述网络状态信息的时间窗口长度，其中f为线性函数或非线性函数。
  18. 根据权利要求16或17所述的装置，其特征在于，所述数据流预处理结果是对当前进入网络的数据流进行预处理得到的具体包括：
    提取所述数据流的特征，所述特征包括数据包报头五元组、数据包长度序列或数据包到达时间间隔序列；
    通过所述特征对所述数据流进行时空数据挖掘处理和/或因果关系分析处理，得到处理结果；
    根据所述处理结果对所述数据流打标签得到所述数据流的标签，所述数据流的标签为所述数据流预处理结果。
PCT/CN2016/095696 2015-08-24 2016-08-17 网络控制策略的生成方法、装置及网络控制器 WO2017032254A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP16838513.6A EP3297211B1 (en) 2015-08-24 2016-08-17 Method and apparatus for generating network control policy, and network controller
US15/903,051 US10749757B2 (en) 2015-08-24 2018-02-23 Method and apparatus for generating network control policy, and network controller

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201510524510.3A CN106487534B (zh) 2015-08-24 2015-08-24 网络控制策略的生成方法、装置及网络控制器
CN201510524510.3 2015-08-24

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US15/903,051 Continuation US10749757B2 (en) 2015-08-24 2018-02-23 Method and apparatus for generating network control policy, and network controller

Publications (1)

Publication Number Publication Date
WO2017032254A1 true WO2017032254A1 (zh) 2017-03-02

Family

ID=58099583

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/095696 WO2017032254A1 (zh) 2015-08-24 2016-08-17 网络控制策略的生成方法、装置及网络控制器

Country Status (4)

Country Link
US (1) US10749757B2 (zh)
EP (1) EP3297211B1 (zh)
CN (1) CN106487534B (zh)
WO (1) WO2017032254A1 (zh)

Families Citing this family (33)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10454714B2 (en) 2013-07-10 2019-10-22 Nicira, Inc. Method and system of overlay flow control
WO2017147808A1 (zh) * 2016-03-02 2017-09-08 华为技术有限公司 管理网络设备的方法和装置
US11706127B2 (en) 2017-01-31 2023-07-18 Vmware, Inc. High performance software-defined core network
US20180219765A1 (en) 2017-01-31 2018-08-02 Waltz Networks Method and Apparatus for Network Traffic Control Optimization
US10992568B2 (en) * 2017-01-31 2021-04-27 Vmware, Inc. High performance software-defined core network
US10530678B2 (en) * 2017-07-20 2020-01-07 Vmware, Inc Methods and apparatus to optimize packet flow among virtualized servers
US11102063B2 (en) 2017-07-20 2021-08-24 Vmware, Inc. Methods and apparatus to cross configure network resources of software defined data centers
US10841235B2 (en) 2017-07-20 2020-11-17 Vmware, Inc Methods and apparatus to optimize memory allocation in response to a storage rebalancing event
US10756967B2 (en) 2017-07-20 2020-08-25 Vmware Inc. Methods and apparatus to configure switches of a virtual rack
US11102032B2 (en) 2017-10-02 2021-08-24 Vmware, Inc. Routing data message flow through multiple public clouds
US11115480B2 (en) 2017-10-02 2021-09-07 Vmware, Inc. Layer four optimization for a virtual network defined over public cloud
US10999100B2 (en) 2017-10-02 2021-05-04 Vmware, Inc. Identifying multiple nodes in a virtual network defined over a set of public clouds to connect to an external SAAS provider
US11223514B2 (en) 2017-11-09 2022-01-11 Nicira, Inc. Method and system of a dynamic high-availability mode based on current wide area network connectivity
CN110048869A (zh) * 2018-01-16 2019-07-23 中国科学院沈阳自动化研究所 面向工业时间敏感软件定义网络的资源分配方法和系统
CN110266510B (zh) * 2018-03-21 2022-05-24 腾讯科技（深圳）有限公司 网络控制策略生成方法及装置、网络控制方法、存储介质
CN110941489B (zh) * 2018-09-21 2024-06-18 北京京东尚科信息技术有限公司 流处理引擎的伸缩方法和装置
CN111510313B (zh) 2019-01-30 2021-09-14 华为技术有限公司 通信方法、通信装置及存储介质
US11018995B2 (en) 2019-08-27 2021-05-25 Vmware, Inc. Alleviating congestion in a virtual network deployed over public clouds for an entity
US11489783B2 (en) 2019-12-12 2022-11-01 Vmware, Inc. Performing deep packet inspection in a software defined wide area network
US11438789B2 (en) 2020-01-24 2022-09-06 Vmware, Inc. Computing and using different path quality metrics for different service classes
CN111526096B (zh) * 2020-03-13 2022-03-15 北京交通大学 智融标识网络状态预测与拥塞控制系统
CN111756646B (zh) * 2020-07-08 2023-09-29 腾讯科技（深圳）有限公司 网络传输控制方法、装置、计算机设备及存储介质
US11363124B2 (en) 2020-07-30 2022-06-14 Vmware, Inc. Zero copy socket splicing
US11601356B2 (en) 2020-12-29 2023-03-07 Vmware, Inc. Emulating packet flows to assess network links for SD-WAN
US11792127B2 (en) 2021-01-18 2023-10-17 Vmware, Inc. Network-aware load balancing
US11979325B2 (en) 2021-01-28 2024-05-07 VMware LLC Dynamic SD-WAN hub cluster scaling with machine learning
US11582144B2 (en) 2021-05-03 2023-02-14 Vmware, Inc. Routing mesh to provide alternate routes through SD-WAN edge forwarding nodes based on degraded operational states of SD-WAN hubs
US12009987B2 (en) 2021-05-03 2024-06-11 VMware LLC Methods to support dynamic transit paths through hub clustering across branches in SD-WAN
US11729065B2 (en) 2021-05-06 2023-08-15 Vmware, Inc. Methods for application defined virtual network service among multiple transport in SD-WAN
US12015536B2 (en) 2021-06-18 2024-06-18 VMware LLC Method and apparatus for deploying tenant deployable elements across public clouds based on harvested performance metrics of types of resource elements in the public clouds
US11943146B2 (en) 2021-10-01 2024-03-26 VMware LLC Traffic prioritization in SD-WAN
US11909815B2 (en) 2022-06-06 2024-02-20 VMware LLC Routing based on geolocation costs
CN115174557B (zh) * 2022-07-01 2024-03-01 济南浪潮数据技术有限公司 一种日志下载的调度方法、装置以及介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE102004053608A1 (de) * 2003-12-19 2005-08-04 Lear Corporation, Southfield Fahrzeugenergie-Managementsystem, welches Vorhersagen verwendet
CN101808244A (zh) * 2010-03-24 2010-08-18 北京邮电大学 一种视频传输控制方法及系统
CN104283946A (zh) * 2014-09-26 2015-01-14 东北大学 一种单物理机下多虚拟机的资源自适应调整系统及方法
CN104468253A (zh) * 2013-09-23 2015-03-25 中兴通讯股份有限公司 一种深度包检测控制方法及装置

Family Cites Families (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7225250B1 (en) * 2000-10-30 2007-05-29 Agilent Technologies, Inc. Method and system for predictive enterprise resource management
US20060074875A1 (en) * 2004-09-30 2006-04-06 International Business Machines Corporation Method and apparatus for predicting relative selectivity of database query conditions using respective cardinalities associated with different subsets of database records
US7936762B2 (en) * 2004-12-09 2011-05-03 The Boeing Company QOS provisioning in a network having dynamic link states
CN101075917B (zh) * 2007-07-16 2010-08-25 华为技术有限公司 一种预测网络攻击行为的方法及装置
EP2184681A1 (en) * 2008-10-31 2010-05-12 HSBC Holdings plc Capacity control
US7917954B1 (en) * 2010-09-28 2011-03-29 Kaspersky Lab Zao Systems and methods for policy-based program configuration
US8621477B2 (en) * 2010-10-29 2013-12-31 International Business Machines Corporation Real-time monitoring of job resource consumption and prediction of resource deficiency based on future availability
US9137130B2 (en) * 2011-09-22 2015-09-15 Sap Se Dynamic network load forecasting
CN102510572B (zh) * 2011-11-18 2014-08-06 浙江工业大学 一种面向异构无线传感器网络的分簇路由控制方法
CN103731854B (zh) * 2012-10-10 2017-10-17 华为技术有限公司 基于自组网son的网络状态划分方法、装置和网络系统
EP2957071B1 (en) * 2013-02-18 2017-11-29 Tekelec, Inc. Method, system, and computer readable medium for providing a thinking diameter network architecture
US9736041B2 (en) * 2013-08-13 2017-08-15 Nec Corporation Transparent software-defined network management
CN103491571B (zh) * 2013-10-13 2016-05-25 西安电子科技大学 基于联合功率和速率调整的车辆自组织网络拥塞控制方法
WO2015061353A1 (en) * 2013-10-21 2015-04-30 Nyansa, Inc. A system and method for observing and controlling a programmable network using a remote network manager
WO2015069164A1 (en) * 2013-11-08 2015-05-14 Telefonaktiebolaget L M Ericsson (Publ) Handling of transport conditions
CN104734957B (zh) 2013-12-24 2018-03-23 中国移动通信集团公司 一种软件定义网络sdn中业务传输方法及装置
CN104104718B (zh) 2014-07-02 2017-05-17 北京邮电大学 一种基于软件定义网络的用户自主路由定制系统和方法
US9742690B2 (en) * 2014-08-20 2017-08-22 At&T Intellectual Property I, L.P. Load adaptation architecture framework for orchestrating and managing services in a cloud computing system

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE102004053608A1 (de) * 2003-12-19 2005-08-04 Lear Corporation, Southfield Fahrzeugenergie-Managementsystem, welches Vorhersagen verwendet
CN101808244A (zh) * 2010-03-24 2010-08-18 北京邮电大学 一种视频传输控制方法及系统
CN104468253A (zh) * 2013-09-23 2015-03-25 中兴通讯股份有限公司 一种深度包检测控制方法及装置
CN104283946A (zh) * 2014-09-26 2015-01-14 东北大学 一种单物理机下多虚拟机的资源自适应调整系统及方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3297211A4 *

Also Published As

Publication number Publication date
EP3297211B1 (en) 2021-03-17
CN106487534A (zh) 2017-03-08
EP3297211A4 (en) 2018-04-18
US20180183683A1 (en) 2018-06-28
US10749757B2 (en) 2020-08-18
EP3297211A1 (en) 2018-03-21
CN106487534B (zh) 2019-08-13

Similar Documents

Publication Publication Date Title
WO2017032254A1 (zh) 网络控制策略的生成方法、装置及网络控制器
US10797973B2 (en) Server-client determination
Li et al. Learning IoT in edge: Deep learning for the Internet of Things with edge computing
US20160091913A1 (en) Smart power management in switches and routers
CN113886001A (zh) 用于关于网络的预测的机器学习流水线
CN112532409B (zh) 网络参数配置方法、装置、计算机设备以及存储介质
US20180376105A1 (en) Latency mitigation through intelligent extrapolation in multimedia systems
CN112422443A (zh) 一种拥塞算法的自适应控制方法、存储介质、设备及系统
JP2022017588A (ja) 深層学習フレームワークのトレーニング方法、装置及び記憶媒体
CN116980569A (zh) 一种基于云计算的安全监控系统及方法
KR20200019727A (ko) 네트워크 코딩을 활성화하기 위한 네트워크 파라미터의 최적화
US20160191412A1 (en) Technologies for directed power and performance management
WO2023274304A1 (zh) 分布式路由确定方法、电子设备及存储介质
CN116743635A (zh) 一种网络预测与调控方法及网络调控系统
William et al. Edge Computing based Traffic Control Management for Distributed Environment
US10313470B2 (en) Hierarchical caching and analytics
Liang et al. DNN surgery: Accelerating DNN inference on the edge through layer partitioning
US10628201B2 (en) Analysis method and analysis apparatus
Guo et al. IEEE SA Industry Connections-IEEE 802 Nendica Report: Intelligent Lossless Data Center Networks
Xia et al. Learn to optimize: Adaptive VNF provisioning in mobile edge clouds
Du et al. Distributed in-network coflow scheduling
CN109302724B (zh) 一种分布式多节点无线电实时数据处理系统及方法
Salmeron-Garcia et al. Mobile robot motion planning based on cloud computing stereo vision processing
Zhang et al. DRL-based routing algorithm with guaranteed loss, latency and bandwidth in SDN networks: Application of online video conferencing
CN113179218B (zh) 模型训练方法、网络拥塞控制方法、装置及相关产品

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16838513

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

WWE Wipo information: entry into national phase

Ref document number: 2016838513

Country of ref document: EP