Network service scheduling method and apparatus, storage medium, and program product

1. A method for network service scheduling, comprising:

obtaining, by processing circuitry of an information processing apparatus, historical network traffic data of network services that are provided in a content delivery network (CDN), the historical network traffic data including actual traffic data of the network services in past time units before a present moment;

predicting, by the processing circuitry, future network traffic of the network services in a next time unit after the present moment according to the historical network traffic data of the network services, wherein the future network traffic is predicted based on a machine learning model that is trained according to multiple groups of training data of the network services, the multiple groups of training data including at least information on service operators, regions, and service types of the network services; and

scheduling, by the processing circuitry, network resources for the network services according to the predicted future network traffic.

2. The method according to claim 1 , further comprising:

obtaining, for a network service, first feature data of the next time unit according to the historical network traffic data of the network service; and

predicting, by the processing circuitry, the future network traffic of the network service according to the first feature data of the next time unit and the machine learning model with a model parameter.

3. The method according to claim 2 , further comprising at least one of:

obtaining a time index of the next time unit that is indicative of a specific time duration in a day corresponding to the next time unit, the first feature data of the next time unit for the network service including the time index;

obtaining a day index of the next time unit that is indicative of a specific day in a week for the next time unit, the first feature data of the next time unit for the network service including the day index;

obtaining X network traffic measures of the network service respectively in latest X time units, wherein X is a positive integer, and the first feature data of the next time unit for the network service includes the X network traffic measures;

obtaining Y network traffic measures of the network service respectively for latest Y days, wherein Y is a positive integer, a network traffic measure in the Y network traffic measures for a day in the latest Y days is a traffic measure of the network service in a specific time unit of the day having the same time index as the next time unit, and the first feature data of the next time unit for the network service includes the Y network traffic measures;

obtaining Z network traffic measures of the network service respectively for latest Z weeks, wherein Z is a positive integer, a network traffic measure in the Z network traffic measures for a week is a traffic measure of the network service in a specific time unit in the week having the same time index and the same day index as the next time unit, and the first feature data of the next time unit for the network service includes the Z network traffic measures; and

determining an activeness of the network service, the first feature data of the next time unit for the network service including the activeness of the network service.

4. The method according to claim 2 , further comprising:

obtaining, the multiple groups of training data of the network service according to the historical network traffic data of the network service, a group in the multiple groups of training data comprising a network traffic measure of the network service in a past time unit and second feature data of the past time unit; and

training the machine learning model according to the multiple groups of training data of the network service, to determine the model parameter.

5. The method according to claim 4 , further comprising:

obtaining correction data of the network service after the next time unit and a network traffic measure of the network service in the next time unit is available, wherein the correction data comprises the network traffic measure in the next time unit and the first feature data of the next time unit; and

correcting the model parameter of the machine learning model according to the correction data.

6. The method according to claim 1 , further comprising:

scheduling network bandwidths respectively to the network services in the next time unit according to the predicted future network traffic of the network services.

7. The method according to claim 2 , further comprising:

training the machine learning model according to the historical network traffic data of the network services to determine the model parameter;

predicting the future network traffic of the network services according to the machine learning model; and

scheduling network bandwidths respectively to the network services according to the predicted future network traffic of the network services.

8. An apparatus, comprising:

interface circuitry configured to receive, from devices that provide network services in a content delivery network (CDN), network traffic measures of the network services; and

processing circuitry configured to:

obtain historical network traffic data of the network services, the historical network traffic data including actual traffic data of the network services in past time units before a present moment;

predict future network traffic of the network services in a next time unit after the present moment according to the historical network traffic data of the network services, wherein the future network traffic is predicted based on a machine learning model that is trained according to multiple groups of training data of the network services, the multiple groups of training data including at least information on service operators, regions, and service types of the network services; and

schedule network resources for the network services according to the predicted future network traffic.

9. The apparatus according to claim 8 , wherein the processing circuitry is configured to:

obtain, for a network service, first feature data of the next time unit according to the historical network traffic data of the network service; and

predict the future network traffic of the network service according to the first feature data of the next time unit and the machine learning model with a model parameter.

10. The apparatus according to claim 9 , wherein the processing circuitry is configured to obtain the first feature data that comprises at least one of:

a time index of the next time unit that is indicative of a specific time duration in a day corresponding to the next time unit;

a day index of the next time unit that is indicative of a specific day in a week for the next time unit;

X network traffic measures of the network service respectively in latest X time units, wherein X is a positive integer;

Y network traffic measures of the network service respectively for latest Y days, wherein Y is a positive integer, a network traffic measure in the Y network traffic measures for a day in the latest Y days is a traffic measure of the network service in a specific time unit of the day having the same time index as the next time unit;

Z network traffic measures of the network service respectively for latest Z weeks, wherein Z is a positive integer, a network traffic measure in the Z network traffic measures for a week is a traffic measure of the network service in a specific time unit in the week having the same time index and the same day index as the next time unit; and

an activeness of the network service.

11. The apparatus according to claim 9 , wherein the processing circuitry is configured to:

obtain, the multiple groups of training data of the network service according to the historical network traffic data of the network service, a group in the multiple groups of training data comprising a network traffic measure of the network service in a past time unit and second feature data of the past time unit; and

train the machine learning model according to the multiple groups of training data of the network service, to determine the model parameter.

12. The apparatus according to claim 11 , wherein the processing circuitry is configured to:

obtain correction data of the network service after the next time unit passes and a network traffic measure in the next time unit is available, wherein the correction data comprises the network traffic measure in the next time unit and the first feature data of the next time unit; and

correct the model parameter of the machine learning model according to the correction data.

13. The apparatus according to claim 8 , wherein the processing circuitry is configured to:

schedule network bandwidths respectively to the network services in the next time unit according to the predicted future network traffic of the network services.

14. The apparatus according to claim 9 , wherein the processing circuitry is configured to:

train the machine learning model according to the historical network traffic data of the network services to determine the model parameter;

predict the future network traffic of the network services according to the machine learning model; and

schedule network bandwidths respectively to the network services according to the predicted future network traffic of the network services.

15. A non-transitory computer-readable medium storing instructions which when executed by a computer cause the computer to perform:

obtaining historical network traffic data of network services that are provided in a content delivery network (CDN), the historical network traffic data including actual traffic data of the network services in past time units before a present moment;

predicting future network traffic of the network services in a next time unit after the present moment according to the historical network traffic data of the network services, wherein the future network traffic is predicted based on a machine learning model that is trained according to multiple groups of training data of the network services, the multiple groups of training data including at least information on service operators, regions, and service types of the network services; and

scheduling network resources for the network services according to the predicted future network traffic.

16. The non-transitory computer-readable medium according to claim 15 , wherein the instructions cause the computer to further perform:

obtaining, for a network service, first feature data of the next time unit according to the historical network traffic data of the network service; and

predicting the future network traffic of the network service according to the first feature data of the next time unit and the machine learning model with a model parameter.

17. The non-transitory computer-readable medium according to claim 16 , wherein the instructions cause the computer to further perform at least one of:

obtaining a time index of the next time unit that is indicative of a specific time duration in a day corresponding to the next time unit, the first feature data of the next time unit for the network service including the time index;

obtaining a day index of the next time unit that is indicative of a specific day in a week for the next time unit, the first feature data of the next time unit for the network service including the day index;

obtaining X network traffic measures of the network service respectively in latest X time units, wherein X is a positive integer, and the first feature data of the next time unit for the network service includes the X network traffic measures;

obtaining Y network traffic measures of the network service respectively for latest Y days, wherein Y is a positive integer, a network traffic measure in the Y network traffic measures for a day in the latest Y days is a traffic measure of the network service in a specific time unit of the day having the same time index as the next time unit, and the first feature data of the next time unit for the network service includes the Y network traffic measures;

obtaining Z network traffic measures of the network service respectively for latest Z weeks, wherein Z is a positive integer, a network traffic measure in the Z network traffic measures for a week is a traffic measure of the network service in a specific time unit in the week having the same time index and the same day index as the next time unit, and the first feature data of the next time unit for the network service includes the Z network traffic measures; and

determining an activeness of the network service, the first feature data of the next time unit for the network service including the activeness of the network service.

18. The non-transitory computer-readable medium according to claim 16 , wherein the instructions cause the computer to further perform:

obtaining, the multiple groups of training data of the network service according to the historical network traffic data of the network service, a group in the multiple groups of training data comprising a network traffic measure of the network service in a past time unit and second feature data of the past time unit; and

training the machine learning model according to the multiple groups of training data of the network service, to determine the model parameter.

19. The non-transitory computer-readable medium according to claim 18 , wherein the instructions cause the computer to further perform:

obtaining correction data of the network service after the next time unit passes and a network traffic measure in the next time unit is available, wherein the correction data comprises the network traffic measure in the next time unit and the first feature data of the next time unit; and

correcting the model parameter of the machine learning model according to the correction data.

20. The non-transitory computer-readable medium according to claim 15 , wherein the instructions cause the computer to further perform:

scheduling network bandwidths respectively to the network services in the next time unit according to the predicted future network traffic of the network services.