Data transmission method and apparatus

1. A data transmission method, the method comprising:

determining a first sending rate based on a network performance objective of first data and a network status of a first transmission control protocol (TCP) connection, wherein the first TCP connection is used to send the first data; and

sending the first data based on the first sending rate;

wherein the network performance objective of the first data is based on a characteristic of the first data, wherein the characteristic of the first data comprises a characteristic of an application providing the first data, and wherein the network performance objective indicates one or more requirements corresponding to the characteristic of the application;

wherein the network status of the first TCP connection comprises: a smoothed value of a queue delay, a rate of sending a TCP data packet within a decision period, a rate of receiving a TCP data packet within the decision period, a smoothed value of a round trip time (RTT), an interval of sending a TCP data packet, and an interval of sending an acknowledgement (ACK);

wherein the network status of the first TCP connection is used to evaluate a delay, a sending interval, a packet loss rate, and a bandwidth; and

wherein the network performance objective is an objective weight ratio of a goodput, a delay, and a packet loss rate of the first data.

2. The method according to claim 1 , wherein the network status of the first TCP connection is based on a received ACK corresponding to the first TCP connection.

3. The method according to claim 1 , wherein the determining the first sending rate comprises:

determining the first sending rate based on the network performance objective of the first data, the network status of the first TCP connection, and a first global measurement value, wherein the first global measurement value is a measurement value of one or more TCP connections corresponding to a transmit end.

4. The method according to claim 3 , wherein the global measurement value comprises calculation results from one or more of the following expressions using a logarithm function log ( ): log(goodput of the transmit end/maximum goodput of the transmit end), log(minimum RTT of the one or more TCP connections of the transmit end/smoothed value of a round trip time RTT of the one or more TCP connections of the transmit end), or log(packet loss rate of the transmit end).

5. The method according to claim 3 , further comprising:

storing the network performance objective of the first data, the network status of the first TCP connection, the first global measurement value, and the first sending rate in a database.

6. The method according to claim 3 , wherein a deep reinforcement learning-based rate model comprises a policy network and a sending rate decision network; and the determining the first sending rate comprises:

using the network status of the first TCP connection as an input of the policy network,

using an output of the policy network, the network performance objective of the first data, and the first global measurement value as an input of the sending rate decision network, and

using the first sending rate as an output.

7. The method according to claim 6 , wherein the deep reinforcement learning-based rate model further comprises a value network; and the method further comprising:

determining a loss function value over a value network based on a correspondence between a network performance objective of second data, a network status of a second TCP connection, a second global measurement value, and a second sending rate that are stored in a database;

updating the value network based on the loss function value, and determining an updated policy gradient; and

updating the policy network and the sending rate decision network based on the updated policy gradient.

8. A data transmission apparatus, comprising:

a processor, configured to determine a first sending rate based on a network performance objective of first data and a network status of a first transmission control protocol (TCP) connection, wherein the first TCP connection is used to send the first data; and

a transmitter, configured to send the first data based on the first sending rate;

wherein the network performance objective of the first data is based on a characteristic of the first data, wherein the characteristic of the first data comprises a characteristic of an application providing the first data, and wherein the network performance objective indicates one or more requirements corresponding to the characteristic of the application;

wherein the network status of the first TCP connection comprises: a smoothed value of a queue delay, a rate of sending a TCP data packet within a decision period, a rate of receiving a TCP data packet within the decision period, a smoothed value of a round trip time (RTT), an interval of sending a TCP data packet, and an interval of sending an acknowledgement (ACK);

wherein the network status of the first TCP connection is used to evaluate a delay, a sending interval, a packet loss rate, and a bandwidth; and

wherein the network performance objective is an objective weight ratio of a goodput, a delay, and a packet loss rate of the first data.

9. The apparatus according to claim 8 , wherein the network status of the first TCP connection is based on a received ACK corresponding to the first TCP connection.

10. The apparatus according to claim 8 , wherein the processor is configured to determine the first sending rate based on the network performance objective of the first data, the network status of the first TCP connection, and a first global measurement value, wherein the global measurement value is a measurement value of one or more TCP connections corresponding to a transmit end.

11. The apparatus according to claim 10 , wherein the global measurement value comprises calculation results from one or more of the following expressions using a logarithm function log ( ): log (a goodput of the transmit end/a maximum goodput of the transmit end), log (a minimum RTT of the one or more TCP connections of the transmit end/a smoothed value of a round trip time RTT of the one or more TCP connections of the transmit end), and log (a packet loss rate of the transmit end).

12. The apparatus according to claim 10 , wherein the processor is further configured to cause the network performance objective of the first data, the network status of the first TCP connection, the first global measurement value, and the first sending rate to be stored in a database.

13. The apparatus according to claim 10 , wherein a deep reinforcement learning-based rate model comprises a policy network and a sending rate decision network; and the processor is further configured to:

use the network status of the first TCP connection as an input of the policy network, and

use an output of the policy network, the network performance objective of the first data, and the first global measurement value as an input of the sending rate decision network, wherein an output is the first sending rate.

14. The apparatus according to claim 13 , wherein the deep reinforcement learning-based rate model further comprises a value network; and wherein the processor is further configured to:

determine a loss function value over a value network based on a correspondence between a network performance objective of second data, a network status of a second TCP connection, a second global measurement value, and a second sending rate that are stored in the database;

update the value network based on the loss function value, and determine an updated policy gradient; and

update the policy network and the sending rate decision network based on the updated policy gradient.