Uplink scheduling method and uplink scheduler

Abstract

The invention relates to an uplink scheduling method and an uplink scheduler. The uplink scheduling method is used for a computer communication network, and comprises the following steps of: utilizing a deep learning framework, and according to information of a media access control layer of the computer communication network, system parameters of the media access control layer and transmission data; a ratio of a quantity of transmission control protocol data to a quantity of transmission control protocol acknowledgement information is determined.

Description

Technical Field

[0001] The present invention relates to an uplink scheduling method and an uplink scheduler, and more particularly to an uplink scheduling method and an uplink scheduler for avoiding congestion collapse. Background Art

[0002] The physical layer (PHY) of computer communication networks provides unreliable packet transmission. Packets may be lost due to channels or collisions, packets may not arrive in order or arrive after a long delay when the packet switching system dynamically changes routing, or packets may be transmitted repeatedly. The transmission control protocol (TCP) is a communication protocol that provides reliable data stream transmission and transmission control protocol through positive acknowledgment (ACK) and retransmission technology.

[0003] However, the congestion control mechanism of the existing TCP communication protocol will decide whether to reduce the transmission rate based on the behavior of receiving TCP ACK. For example, if the sender continues to not receive TCP ACK, the sender will think that the network is congested, so it will continue to reduce the transmission rate and reduce the user's experience quality.

[0004] Since TCP ACK is a short packet, if the receiving end has to send back TCP ACK for each TCP data, the media access control (MAC) efficiency may be reduced. Assume that N refers to the ratio of TCP data to TCP ACK. For example, N = 1 means that when the sending end sends a TCP data, it will immediately request TCP ACK from the network interface card (NIC); N = 2 means that when the sending end sends two TCP data, it will request TCP ACK from the network interface card.

[0005] Figure 1 is a schematic diagram of a data quantity and a transmission control protocol logarithm. Figure 1 As shown in the figure, the horizontal axis is the number of TCP pairs and the vertical axis is the amount of data buffered in the MAC layer sequence. Figure 1 It can be seen that as the TCP logarithm increases, the amount of data cached in the MAC layer sequence corresponding to N=1 can approach a fixed value, but when N=2, the amount of data cached in the MAC layer sequence increases exponentially with the increase of the TCP logarithm, which is likely to trigger congestion collapse.

[0006] Therefore, how to select a correct N value for different user scenarios is an important issue in the art. Summary of the invention

[0007] In view of this, the present invention provides an uplink scheduling method and an uplink scheduler to solve the above problems through the generalization of artificial intelligence.

[0008] An embodiment of the present invention discloses an uplink scheduling method for a computer communication network, comprising using a deep learning architecture to determine a ratio of a quantity of transmission control protocol data to a quantity of transmission control protocol confirmation information based on information of a media access control layer of the computer communication network, system parameters of the media access control layer, and transmission data.

[0009] An embodiment of the present invention further discloses an uplink scheduler for a computer communication network, comprising a decision engine for determining a ratio of a quantity of transmission control protocol data to a quantity of transmission control protocol confirmation information based on information of a media access control layer of the computer communication network, system parameters of the media access control layer, and transmission data using a deep learning architecture; and a scheduler coupled to the decision engine for requesting a transmission control protocol confirmation information from a receiving end of the computer communication network based on the ratio. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A schematic diagram of a data quantity and a transmission control protocol logarithm.

[0011] Figure 2 FIG. 4 is a schematic diagram of an uplink scheduler according to an embodiment of the present invention.

[0012] Figure 3 Schematic diagram of a decision engine according to an embodiment of the present invention.

[0013] Figure 4 Schematic diagram of a ResNet-50 architecture according to an embodiment of the present invention.

[0014] Figure 5 FIG. 4 is a schematic diagram of an uplink scheduling method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0015] Please refer to Figure 2 , Figure 2Schematic diagram of an uplink scheduler 20 according to an embodiment of the present invention. The uplink scheduler 20 is used in a computer communication network, and the uplink scheduler 20 includes a decision engine 202 and a scheduler 204. The decision engine 202 is used to determine a ratio of a number of transmission control protocol (TCP) data to a number of TCP acknowledgment (ACK) information based on a media access control (MAC) layer of the computer communication network, system parameters of the MAC layer, and transmission data using a deep learning architecture to maximize a user experience (quality of experience, QoE) of the computer communication system. The scheduler 204 is coupled to the decision engine 202 and is used to request TCP acknowledgment information from a receiving end of the computer communication network based on the value of the TCP pair.

[0016] Please refer to Figure 3 , Figure 3 Schematic diagram of the decision engine 202 of the embodiment of the present invention. The decision engine 202 of the embodiment of the present invention learns the characteristics that can be used to describe the transmission control protocol layer according to the information of the media access control layer, and further predicts the ratio N. For example, when N=1, it means that when the transmitting end sends a transmission control protocol data, it immediately requests the receiving end to obtain the transmission control protocol confirmation information; when N=2, it means that when the transmitting end sends two transmission control protocol data, it immediately requests the receiving end to obtain the transmission control protocol confirmation information, and so on.

[0017] Depend on Figure 3 It can be seen that the input parameters of the decision engine 202 may include a sequence length Queue_size of the media access control layer cached in the transport layer of the computer communication network, a sequence state Queue_state (for example, an empty sequence state, a full sequence state), a packet discard probability Drop_rate, a MAC Protocol Data Unit maximum value (MaxPMDU) of the system parameters of the media access control layer, a MAC Service Data Unit maximum value (MaxMSDU) and a spectrum utilization efficiency Spectral_efficiency of the statistical transmission data after transmission, a user transmission speed Rate_per_STA, and a transmission error rate Packet_error_rate.

[0018] The decision engine 202 is based on a deep learning framework and transforms the above input parameters into a multi-class classification problem to determine the ratio N for maximizing the user experience of the computer communication system.

[0019] To avoid gradient exposure and identity mapping, the embodiment of the present invention learns data features based on residual learning. In one embodiment, a ResNet-50 architecture can be used to learn the features required by the uplink scheduler 202 to determine the optimal ratio N for different scenarios. In this way, the scheduler 204 can request transmission control protocol confirmation information from the receiving end of the computer communication network in different application scenarios according to the ratio N determined by the decision engine 202 of the uplink scheduler 10 of the embodiment of the present invention, thereby achieving N TCP data triggering a transmission control protocol confirmation information.

[0020] Please refer to Figure 4 , Figure 4 Schematic diagram of the ResNet-50 architecture of an embodiment of the present invention. The ResNet-50 architecture includes an input formatting module IF and stages Stage_1-Stage_7. The input formatting module IF formats the input, which is then processed by the convolution module CONV, batch normalization module Batch_Norm, rectified linear unit module ReLU and maximum pooling module MAXPOOL of stage Stage_1.

[0021] Stages Stage_2-Stage_6 further perform error convergence on the learning network, where the data output by stage Stage_1 is processed by the convolution block module CONV_BLOCKS and the identity block module ID_BLOCKS. Finally, stage Stage_7 processes the output of stage Stage_6 by the average pooling module AVG_POOL, the flattening module Flatten, and the fully connected layer FC to determine the optimal ratio N.

[0022] The above-mentioned operation of the uplink scheduler 20 can be summarized as an uplink scheduling method 50. The uplink scheduling method 50 comprises the following steps:

[0023] Step 502: Start;

[0024] Step 504: Determine the ratio of the amount of TCP data to the amount of TCP ACK information based on the information of the MAC layer of the computer communication network, the system parameters of the MAC layer, and the transmission data using a deep learning framework;

[0025] Step 506: End.

[0026] Regarding the operation flow of the uplink scheduling method 50 , please refer to the above-mentioned embodiment of the uplink scheduler 20 , which will not be described in detail here.

[0027] It is worth noting that the ResNet-50 architecture and the input parameters of the decision engine in the above-mentioned embodiments can be changed according to different requirements and all fall within the scope of the present invention.

[0028] In summary, the present invention discloses an uplink scheduling method and a related uplink scheduler, which determines the ratio of the amount of transmission control protocol data to the amount of transmission control protocol confirmation information according to different usage scenarios through deep learning of the features required by the uplink scheduler, thereby maximizing the user experience of the computer communication system.

[0029] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made according to the scope of the patent application of the present invention should fall within the scope of the present invention.

[0030]

Explanation of symbols

[0031] 20: Uplink Scheduler

[0032] 202: Decision Engine

[0033] 204: Scheduler

[0034] 50: Uplink Scheduling Method

[0035] 502-506: Methods

[0036] AVG_POOL: Average pooling module

[0037] Batch_Norm: Batch normalization module

[0038] CONV: Convolution module

[0039] CONV_BLOCKS: Convolutional block module

[0040] Drop_rate: packet drop probability

[0041] FC: Fully connected layer

[0042] Flatten: Flatten module

[0043] ID_BLOCKS: Identity Block Module

[0044] IF: Input formatting module

[0045] MaxMSDU: Maximum value of MAC service data unit

[0046] MaxPMDU: Maximum value of MAC protocol data unit

[0047] MAXPOOL: Maximum pooling module

[0048] N: Ratio

[0049] Packet_error_rate: Transmission error rate

[0050] Queue_size: sequence length

[0051] Queue_state: sequence state

[0052] Rate_per_STA: User transmission speed

[0053] ReLU: Rectified Linear Unit Module

[0054] Spectral_efficiency: Spectral efficiency

[0055] Stage_1-Stage_7: Stage.

Claims

1. An uplink scheduling method for a computer communication network, comprising: A ratio of a quantity of transmission control protocol data to a quantity of transmission control protocol confirmation information is determined based on information of a media access control layer of the computer communication network, system parameters of the media access control layer, and transmission data using a deep learning framework.

2. The uplink scheduling method as claimed in claim 1, wherein the information of the media access control layer at least includes one of a sequence length, a sequence status, and a packet discard probability cached in a transport layer of the computer communication network. 3 . The uplink scheduling method as claimed in claim 1 , wherein the system parameter of the MAC layer comprises at least one of a maximum MAC protocol data unit (PDU) and a maximum MAC service data unit (SPAU). 4 . The uplink scheduling method as claimed in claim 1 , wherein the transmission data comprises at least one of a spectrum utilization efficiency, a user transmission speed, and a transmission error rate.

5. The uplink scheduling method as claimed in claim 1, wherein the deep learning architecture determines the ratio based on a decision engine.

6. The uplink scheduling method as claimed in claim 5, wherein the decision engine converts the information of the media access control layer of the computer communication network, the system parameters of the media access control layer and the transmission data into a multi-class classification problem to determine the ratio.

7. The uplink scheduling method as claimed in claim 5, wherein the deep learning framework learns data features in a residual learning manner.

8. An uplink scheduler for a computer communication network, comprising: a decision engine for determining a ratio of a quantity of a transmission control protocol data to a quantity of a transmission control protocol confirmation message based on information of a media access control layer of the computer communication network, system parameters of the media access control layer, and transmission data using a deep learning framework; and A scheduler is coupled to the decision engine and is used to request a transmission control protocol confirmation message from a receiving end of the computer communication network according to the ratio.

9. The uplink scheduler of claim 8, wherein the information of the media access control layer comprises at least one of a sequence length, a sequence status, and a packet discard probability cached in a transport layer of the computer communication network.

10. The uplink scheduler of claim 8, wherein the system parameter of the MAC layer comprises at least one of a maximum MAC PDU size and a maximum MAC SDU size.

Images