Method and apparatus for generating network control policy, and network controller

1. A method for use by a network controller of a network, comprising:

collecting network status data of a current moment;

estimating network status information of the current moment according to the network status data of the current moment, wherein the network status information of the current moment comprises network link status at the current moment, network element queue status at the current moment, and network data flow distribution status at the current moment;

predicting network status information of a next moment according to at least the network status information of the current moment, wherein the predicted network status information of the next moment comprises network link status at the next moment, network element queue status at the next moment, and network data flow distribution status at the next moment;

receiving and preprocessing a data flow entering the network at the current moment, to obtain a preprocessing result of the data flow;

generating a control action by using a network control policy of the current moment and according to the network status information of the current moment, the predicted network status information of the next moment, and the preprocessing result of the data flow;

obtaining feedback information of the network resulted from the control action; and

generating a network control policy of the next moment by adjusting the network control policy of the current moment according to the feedback information.

2. The method according to claim 1 , wherein before estimating the network status information of the current moment according to the network status data of the current moment, the method further comprises:

sending a status data acquisition command; and

receiving the network status data of the current moment acquired according to the status data acquisition command.

3. The method according to claim 2 , wherein before predicting the network status information of the next moment according to the network status information of the current moment, the method further comprises:

obtaining historical network status information from a database;

and wherein predicting the network status information of the next moment according to at least the network status information of the current moment comprises:

predicting the network status information of the next moment according to the network status information of the current moment and the historical network status information.

4. The method according to claim 1 , wherein the network status information of the next moment is predicted according to:

Ŝ t+1 =f ( S t−τ ,0≤τ≤ L ),

where Ŝ t+1 represents the network status information at the next moment t+1, S t−τ represents the network status information at a moment t−τ up to the current moment t, 0≤τ≤L, L is a time window length of the network status information, and f is a linear function or a nonlinear function.

5. The method according to claim 1 , wherein preprocessing the data flow entering the network at the current moment, to obtain a preprocessing result of the data flow comprises:

extracting a feature of the data flow;

performing spatial-temporal data mining processing and/or causal relationship analysis processing on the data flow by using the feature, to obtain a processing result; and

labeling the data flow according to the processing result, to obtain a label of the data flow, wherein the label of the data flow is the preprocessing result of the data flow;

wherein the feature of the data flow is one of: a packet header 5-tuple, a packet length sequence, or a packet arrival time interval sequence.

6. The method according to claim 5 , wherein the control action generated by using the network control policy of the current moment and according to the network status information of the current moment, the predicted network status information of the next moment, and the preprocessing result of the data flow is:

a i ˜π t ( a|s t ,ŝ t+1 ,l i ),

where Ŝ t+1 represents the network status information of the next moment, S t represents the network status information of the current moment, l i represents a label of the i th data flow, π t represents the network control policy at the current moment, a is a control variable, π t (a|s t , ŝ t+1 , l i ) represents probability distribution of a under the network control policy at the current moment using Ŝ t+1 , S t , and l i as conditional variables, and random sampling is performed in the probability distribution of the control variable a, to obtain the control action a i .

7. A method for use by a network controller of a network, comprising:

obtaining network status information of the network at a current moment, network status information of the network at a next moment, and a preprocessing result of a data flow entering the network at the current moment;

generating a control action by using a network control policy of the current moment and according to the network status information of the current moment, the network status information of the next moment, and the preprocessing result of the data flow;

receiving feedback information of the network resulted from the control action; and

generating a network control policy of the next moment by adjusting the network control policy of the current moment according to the feedback information;

wherein the network status information comprises network link status of the network, network element queue status in the network, and network data flow distribution status of the network, and

wherein the network status information of the next moment is obtained through prediction according to the network status information of the current moment.

8. The method according to claim 7 , wherein the network status information of the next moment is predicted according to:

Ŝ t+1 =f ( S t−τ ,0≤τ≤ L ),

where Ŝ t+1 represents the network status information at the next moment t+1, S t−τ represents the network status information at a moment t−τ up to the current moment t, 0≤τ≤L, L is a time window length of the network status information, and f is a linear function or a nonlinear function.

9. The method according to claim 7 , wherein the preprocessing result of the data flow is obtained by preprocessing the data flow centering the network at the current moment, and preprocessing the data flow comprises:

extracting a feature of the data flow;

performing spatial-temporal data mining processing and/or causal relationship analysis processing on the data flow by using the feature, to obtain a processing result; and

labeling the data flow according to the processing result, to obtain a label of the data flow, wherein the label of the data flow is the preprocessing result of the data flow;

wherein the feature of the data flow is one of: a packet header 5-tuple, a packet length sequence, or a packet arrival time interval sequence.

10. A network controller of a network, comprising:

a networking interface and a processor coupled to the networking interface;

wherein the networking interface is configured to:

receiving network status data of a current moment; and

receiving a data flow entering the network at the current moment;

wherein the processor is configured to:

estimating network status information of the current moment according to the network status data of the current moment, wherein the network status information of the current moment comprises network link status at the current moment, network element queue status at the current moment, and network data flow distribution status at the current moment;

predicting network status information of a next moment according to at least the network status information of the current moment, wherein the predicted network status information of the next moment comprises network link status at the next moment, network element queue status at the next moment, and network data flow distribution status at the next moment;

preprocessing the data flow entering the network at the current moment, to obtain a preprocessing result of the data flow;

generating a control action by using a network control policy of the current moment and according to the network status information of the current moment, the predicted network status information of the next moment, and the preprocessing result of the data flow;

obtaining feedback information of the network resulted from the control action; and

generating a network control policy of the next moment by adjusting the network control policy of the current moment according to the feedback information.

11. The network controller according to claim 10 , wherein before estimating the network status information of the current moment according to the network status data of the current moment the networking interface is configured to:

send a status data acquisition command; and

receive the network status data of the current moment acquired according to the status data acquisition command.

12. The apparatus according to claim 11 , wherein before predicting the network status information of the next moment according to the network status information of the current moment, the processor is further configured to:

obtain historical network status information from a database;

and wherein in predicting the network status information of the next moment according to at least the network status information of the current moment, the processor is configured to:

predict the network status information of the next moment according to the network status information of the current moment and the historical network status information.

13. The network controller according to claim 10 , wherein the network status information of the next moment is predicted according to:

Ŝ t+1 =f ( S t−τ ,0≤τ≤ L ),

where Ŝ t+1 represents the network status information at the next moment t+1, S t−τ represents the network status information at a moment t−τ up to the current moment t, 0≤τ≤L, L is a time window length of the network status information, and f is a linear function or a nonlinear function.

14. The network controller according to claim 10 , wherein in preprocessing the data flow entering the network at the current moment, to obtain a preprocessing result of the data flow, the processor is configured to:

extract a feature of the data flow;

perform spatial-temporal data mining processing and/or causal relationship analysis processing on the data flow by using the feature, to obtain a processing result; and

label the data flow according to the processing result, to obtain a label of the data flow, wherein the label of the data flow is the preprocessing result of the data flow;

wherein the feature of the data flow is one of: a packet header 5-tuple, a packet length sequence, or a packet arrival time interval sequence.

15. The network controller according to claim 14 , wherein the control action generated by using the network control policy of the current moment and according to the network status information of the current moment, the predicted network status information of the next moment, and the preprocessing result of the data flow is:

a i ˜π t ( a|s t ,ŝ t+1 ,l i ),

where Ŝ t+1 represents the network status information of the network of the next moment, S t represents the network status information of the next moment, l i represents a label of the i th data flow, π t represents the network control policy at the current moment, a is a control variable, π t (a|s t , ŝ t+1 , l i ) represents probability distribution of a under the network control policy at the current moment using Ŝ t+1 , S t , and l i as conditional variables, and random sampling is performed in the probability distribution of the control variable a, to obtain the control action a i .

16. A network controller of a network, comprising:

a networking interface;

a memory storing a program; and

a processor coupled to the networking interface and the memory;

wherein the program, when executed by the processor, causes the network controller to:

obtain network status information of the network at a current moment, network status information of the network at a next moment, and a preprocessing result of a data flow entering the network at the current moment;

generate a control action by using a network control policy of the current moment and according to the network status information of the current moment, the network status information of the next moment, and the preprocessing result of the data flow;

receive feedback information of the network resulted from the control action; and

generate a network control policy of the next moment by adjusting the network control policy of the current moment according to the feedback information;

wherein the network status information comprises network link status of the network, network element queue status in the network, and network data flow distribution status of the network, and

wherein the network status information of the next moment is obtained through prediction according to the network status information of the current moment.

17. The network controller according to claim 16 , wherein that the network status information of the next moment is predicted according to:

Ŝ t+1 =f ( S t−τ ,0≤τ≤ L ),

where Ŝ t+1 represents the network status information of the next moment t+1, S t−τ represents the network status information at a moment t−τ up to the current moment t, 0≤τ≤L, L is a time window length of the network status information, and f is a linear function or a nonlinear function.

18. The network controller according to claim 16 , wherein that the preprocessing result of the data flow is obtained by preprocessing the data flow entering the network at the current moment, and in preprocessing the data flow, the program causes the network controller to:

extract a feature of the data flow;

perform spatial-temporal data mining processing and/or causal relationship analysis processing on the data flow by using the feature, to obtain a processing result; and

label the data flow according to the processing result, to obtain a label of the data flow, wherein the label of the data flow is the preprocessing result of the data flow;

wherein the feature of the data flow is one of: a packet header 5-tuple, a packet length sequence, or a packet arrival time interval sequence.