WO2022022539A1

WO2022022539A1 - Network congestion control method and related apparatus

Info

Publication number: WO2022022539A1
Application number: PCT/CN2021/108785
Authority: WO
Inventors: 徐聪; 张海波; 袁庭球
Original assignee: 华为技术有限公司
Priority date: 2020-07-29
Filing date: 2021-07-28
Publication date: 2022-02-03
Also published as: CN114095434A

Abstract

The present application relates to the field of communications, and provides a network congestion control method and a related apparatus. In the technical solution provided in the present application, a source device sends a probe packet to a destination device before sending a data packet to the destination device; and by means of transmission of the probe packet by an intermediate network device and feedback of the probe packet by the destination device, available bandwidth on a transmission path from the source device to the destination device is tested. In this way, the source device sends the data packet on the basis of the available bandwidth tested by means of the probe packet, thus facilitating congestion free and packet loss free transmission of the data packet. Furthermore, in the technical solution of the present application, on the basis of the characteristic of the probe packet being allowed to be lost, the data pack can be transmitted at full throughput by transmitting the probe packet at a relatively aggressive speed. In addition, the technical solution of the present application provides a heavy loading mode and a light loading mode as well as congestion control methods under the modes; moreover, fast and fair convergence of a transmission rate can be achieved by selecting a suitable switching point for the light loading mode and the heavy loading mode.

Description

Method and related apparatus for controlling network congestion

This application claims the priority of the Chinese patent application with the application No. 202010742582.6 and the invention titled "Method and Related Apparatus for Controlling Network Congestion" filed with the China Patent Office on July 29, 2020, the entire contents of which are incorporated herein by reference middle.

technical field

The present application relates to the field of communications, and more particularly, to a method and related apparatus for controlling network congestion in a communications network.

Background technique

When data packets are transmitted in a communication network, the problem of network congestion usually occurs. Based on the problem of network congestion, congestion control methods have been proposed in the field of communication.

The traditional congestion control method is to detect the congestion status of the communication network based on congestion signals such as packet loss, and control the packet sending rate of the data packet from the source end of the data packet based on the congestion status of the communication network to control the congestion status of the communication network. . Specifically, when it is detected that the data packet is not lost, the data packet sending rate is increased; when it is detected that the data packet is lost, the data packet sending rate is reduced to relieve network congestion.

Although the traditional congestion control method can alleviate the congestion of the communication network to a certain extent, the congestion control method is a kind of "post-knowledge" mechanism. When the source receives the packet loss signal, the data packet has already occurred. Packet loss, that is, performance loss has been caused, and the subsequent slowdown adjustment has lagged behind the time point of network congestion. Therefore, under this mechanism, the congestion of the communication network becomes the norm, and the backlog of packets on the communication equipment becomes the norm, and the backlog of packets will seriously deteriorate the end-to-end delay of data packet transmission, and ultimately affect the transmission of data packets in the communication network. performance.

SUMMARY OF THE INVENTION

The present application provides methods and related apparatuses for controlling network congestion. The technical solution of the present application can detect the bandwidth situation of the network through detection packets that are not afraid of packet loss, and control the transmission rate of the data packets based on the detected network bandwidth.

In a first aspect, the present application provides a method for controlling network congestion. The method includes: a first network device sending a first detection packet, where the first detection packet refers to a detection packet in which the source device is the first network device and the destination device is the second network device, and the detection The message is used to detect the available bandwidth on the transmission path between the source device and the destination device; the first network device receives the first feedback information, and the first feedback information is used to indicate that the first detection message is in the The reception situation on the second network device; the first network device sends a first data packet based on the first feedback information, and the first data packet means that the source device is the first network device and The destination device is the data packet of the second network device.

In this method, because the first network device detects the network bandwidth currently available for transmitting data packets from the first network device to the second network device based on the probe packet before sending the data packet, and based on the detected bandwidth The available bandwidth is used to send data packets to the second network device, so it is helpful to reasonably utilize network bandwidth resources to send data packets to the second network device, thereby implementing network congestion control.

In some possible implementations, the first network device sends a first data packet based on the first feedback information, including:

The first network device sends, based on the first feedback information, the first data packet to the second network device through a third network device, where the third network device is configured to use a first proportion of bandwidth for sending A probe packet and a bandwidth for sending a data packet using a second ratio, wherein the first rate at which the first network device sends the first data packet and the second network device receiving the first probe The ratio of the second rate of packets is equal to the ratio of the second ratio to the first ratio.

In these implementation manners, the third network device presets to use the bandwidth of the first proportion to send the probe packets, and presets to use the bandwidth of the second proportion to send the data packets. In this way, after the first network device receives the first feedback information, it can know the rate at which the second network device receives the first probe packet according to the reception rate of the first feedback information, and can obtain the rate at which the second network device receives the first probe packet according to the rate at which the probe packet is received on the first network device. The receiving rate, the proportion of the bandwidth occupied by the probe packets, and the proportion of the bandwidth that can be occupied by the data packets control the sending rate of the data packets sent at this time, that is, the second rate.

The first network device sends the first data packet based on the first probe feedback information, and the rate of the first data packet is exactly the second rate, which can enable the third network device to just be able to handle the full throughput without packet loss. The second network device forwards the first data packet sent by the first network device.

In some possible implementation manners, the second network device is configured to send one piece of the first feedback information every time one of the first probe packets is received. The sending, by the first network device, a first data packet based on the first feedback information includes: sending a first data packet each time the first network device receives a piece of the first feedback information message.

In these implementation manners, the second network device sends a first feedback message every time it receives a first detection message, and the first network device sends a first data message every time it receives a first feedback message, so that the first data The rate of the packet is exactly the second rate, so that the first data packet can be transmitted to the second network device without packet loss and at full throughput.

In some possible implementations, the sending, by the first network device, the first detection packet includes: the first network device sends the first detection packet using a third rate; the first network device obtains the the first packet loss rate when the first network device uses the third rate to send the first probe packet; the first network device determines the second rate according to the first packet loss rate; the The first network device sends the first probe packet based on the second rate.

That is to say, in these implementations, according to the third rate used for sending the first probe packet and the packet loss rate when the first probe packet was previously sent using the third rate, it is known that the current network sends the first network device to the first network device. The sent probe packet is transmitted to the probe packet sending rate corresponding to the available bandwidth of the second network device, that is, the second rate, and the first probe packet is resent at the second rate. In this way, the sending rate of the current detection packet can be adjusted in time, so that the detection packet can be transmitted to the second network device at full throughput, so that the first data packet can be transmitted to the second network device at full throughput.

In some possible implementations, the sending, by the first network device, the first probe packet based on the second rate includes: when the second rate is less than or equal to a preset rate threshold, the The first network device sends the first probe packet using a preset target rate, where the target rate is greater than the second rate; or when the second rate is greater than the rate threshold, the first network The device sends the first probe packet using the second rate.

That is, when the second rate is small, it is determined that the network load is heavy and the available bandwidth is small. At this time, the preset fixed target rate may be used to send the first detection packet, so that the detection packet flows continuously, thereby making the data packet flow continuously.

In the case where the second rate is relatively high, it is determined that the network load is relatively light and the available bandwidth is relatively large. In this case, the adjusted second rate can be used to send the first probe packet, so as to make full use of network resources to send the first probe packet, so as to detect the most reasonable available bandwidth of the data packet, so as to make full use of network resources to send data message, to achieve full throughput and no packet loss transmission of the first data message.

In some implementation manners, the first network device is one of multiple network devices, wherein the target rates of detection packets of all network devices in the multiple network devices are the same.

Each network device uses the same target rate in overload mode, which can achieve fair convergence of probe packets from multiple network devices.

In some possible implementations, the target rate is the same as the rate threshold. In this way, when switching from the third rate to the target rate, the transmission rate of the probe packet in the communication network can quickly converge.

In some possible implementations, when the probe packet forwarded by the third network device on the transmission path from the first network device to the second network device only includes the first probe packet, the first probe The network device can successfully transmit a specified number of first probe packets from the first network device to the second network device within a specified period of time using the target rate.

In these implementation manners, the purpose rate of the first detection packet is relatively aggressive, that is, relatively large. Therefore, network resources can be fully utilized to send the first detection packet to detect the available bandwidth of the data packet, so that the network can be fully utilized. The resource sends the first data packet to achieve full throughput and no packet loss transmission of the first data packet.

In some possible implementations, the first network device determining the second rate according to the first packet loss rate includes: when the first packet loss rate is less than a preset packet loss rate threshold, The first network device adjusts the sending rate of the first probe packet based on the first packet loss rate and the third rate to obtain the second rate. The sending, by the first network device, the first detection packet based on the second rate includes: the first network device using the second rate to send the first detection packet.

That is, if the packet loss rate is small, it indicates that the light load mode is entered. At this time, a more appropriate second rate may be obtained by adjusting the packet loss rate and the previous third rate. In this way, the resources can be fully utilized to transmit the probe packets, so that the resources can be fully utilized to transmit the data packets.

In some possible implementations, the packet loss rate threshold is greater than the packet loss rate of the first probe packet when the first network device sends the first probe packet at the second rate.

Because the packet loss rate threshold is greater than the packet loss rate at the second rate, when the first network device switches from the heavy-load mode to the light-load mode, it will adjust the speed so as not to destroy the sending rates of the original multiple source devices fairness between.

In some possible implementations, the second rate is obtained based on the first packet loss rate and the third rate using a structure-behavior-performance SCP strategy or a fast-path strategy.

In this way, the convergence of the sending rate of the probe packets of each sending device can be achieved quickly and fairly, thereby realizing the fast and fair convergence of the sending rate of the data packets of each network device.

In a second aspect, the present application provides a method for controlling network congestion. The method includes: a second network device receives a first detection packet, where the first detection packet refers to a detection packet in which the source device is the first network device and the destination device is the second network device, and the detection The message is used to detect the available bandwidth on the transmission path between the source device and the destination device; the second network device sends first feedback information, where the first feedback information is used to instruct the second network device to Reception of the first detection packet; the second network device receives the first data packet sent by the first network device based on the first feedback information, where the first data packet refers to the source device being the The first network device and the destination device is the data packet of the second network device.

In this method, after receiving the detection packet, the second network device feeds back the reception status of the detection packet to the source device that sent the detection packet, so that the source device can know the available bandwidth of the network according to the reception status, Further, the source device can send data packets based on the available bandwidth.

In some possible implementation manners, the sending, by the second network device, the first feedback information includes: sending one piece of the first feedback information each time the second network device receives one of the first probe packets.

In some other possible implementation manners, the first feedback information is specifically used to indicate that the second network device receives the first probe packet.

In a third aspect, the present application provides a method for controlling network congestion. The method includes: a third network device receives a detection packet, where the detection packet is used to detect available bandwidth on a transmission path between a source device and a destination device; the third network device forwards the detection packet; The third network device receives feedback information, where the feedback information is used to indicate that the destination device receives the probe packet; the third device forwards the feedback information; the third device receives the the data packet sent by the source device based on the feedback information; the third device forwards the data packet.

In this method, the third network device acts as an intermediate device between the source device that sends the probe packet and the data packet and the destination device that receives the probe packet and the data packet, and assists in sending the probe packet, so that the probe can be used for sending the probe packet. The available bandwidth of the data packets, so that the source device can control the sending of the data packets based on the available bandwidth, so as to control the congestion of the data packets.

In some possible implementations, the third network device forwarding the probe packet includes: the third network device sends the probe packet by using a bandwidth of a first proportion; and the third network device forwards the data packet, including: the third network device sends the data packet by using the bandwidth of the second ratio; wherein the rate at which the third network device receives the first data packet is the same as the rate at which the third network device receives the first data packet The ratio of the rates of the first probe packets is equal to the ratio of the second ratio to the first ratio.

In this implementation manner, the third network device uses a preset ratio to send probe packets, so that the source device that sends the probe packets can know the actual use for sending probe packets based on the reception of the probe packets at the destination device. Then, based on the bandwidth and the preset first ratio and second ratio, the available bandwidth that can be used for sending data packets can be obtained, and finally, based on the available bandwidth, no packet loss and full throughput transmission of data packets can be realized.

In some possible implementations, the third network device forwarding the feedback information includes: the third network device forwards the feedback information with the highest priority.

In this implementation manner, the third network device preferentially transmits the feedback information, that is, preferentially feeds back the reception status of the detection packet, so that the success rate and the accuracy rate of the source device sending the detection packet to know the reception status of the detection packet can be improved. Improving the success rate and accuracy rate of the source device in learning the available bandwidth for transmitting data packets can ultimately improve the success rate and accuracy rate of network congestion control.

In a fourth aspect, the present application provides an apparatus for controlling network congestion. The apparatus includes a module for executing the method in the first aspect or any one of the implementation manners.

In a fifth aspect, the present application provides an apparatus for controlling network congestion. The apparatus includes a module for executing the method in the second aspect or any one of the implementation manners.

In a sixth aspect, the present application provides an apparatus for controlling network congestion. The apparatus includes a module for executing the method in the third aspect or any one of the implementation manners.

In a seventh aspect, a device for controlling network congestion is provided, the device comprising: a memory, a processor and a transceiver; the memory is used to store a program; the processor is used to execute the program stored in the memory; The processor is used for communicating with other apparatuses or devices; when the program stored in the memory is executed, the processor and the transceiver are used for executing the method in the first aspect or any one of the implementation manners.

In an eighth aspect, a network congestion control device is provided, the device comprising: a memory, a processor and a transceiver; the memory is used for storing a program; the processor is used for executing the program stored in the memory; The processor is used to communicate with other apparatuses or devices; when the program stored in the memory is executed, the processor and the transceiver are used to execute the method in the second aspect or any one of the implementation manners.

In a ninth aspect, a device for controlling network congestion is provided, the device comprising: a memory, a processor, and a transceiver; the memory is used to store a program; the processor is used to execute the program stored in the memory; The processor is used to communicate with other apparatuses or devices; when the program stored in the memory is executed, the processor and the transceiver are used to execute the method in the third aspect or any one of the implementation manners.

In a tenth aspect, a computer-readable medium is provided, where the computer-readable medium stores program codes for device execution, where the program codes are used to execute the method in the first aspect or any one of the implementation manners thereof.

In an eleventh aspect, a computer-readable medium is provided, where the computer-readable medium stores program code for execution by a device, where the program code is used for executing the method in the second aspect or any one of the implementation manners thereof.

A twelfth aspect provides a computer-readable medium, where the computer-readable medium stores program code for device execution, where the program code is used to execute the method in the third aspect or any one of the implementation manners thereof.

A thirteenth aspect provides a computer program product containing instructions, when the computer program product runs on a computer, the computer causes the computer to execute the method in the first aspect or any one of the implementation manners.

A fourteenth aspect provides a computer program product comprising instructions, which when the computer program product is run on a computer, causes the computer to execute the method of the second aspect or any one of the implementation manners.

A fifteenth aspect provides a computer program product comprising instructions, which when the computer program product is run on a computer, causes the computer to execute the method in the third aspect or any one of the implementation manners.

A sixteenth aspect provides a chip, the chip includes a processor and a data interface, the processor reads an instruction stored in a memory through the data interface, and executes the first aspect or any one of the implementation manners. Methods.

Optionally, as an implementation manner, the chip may further include a memory, in which instructions are stored, the processor is configured to execute the instructions stored in the memory, and when the instructions are executed, the The processor is configured to execute the method in the first aspect or any one of the implementation manners thereof.

A seventeenth aspect provides a chip, the chip includes a processor and a data interface, the processor reads an instruction stored in a memory through the data interface, and executes the second aspect or any one of the implementation manners. Methods.

Optionally, as an implementation manner, the chip may further include a memory, in which instructions are stored, the processor is configured to execute the instructions stored in the memory, and when the instructions are executed, the The processor is configured to execute the method in the second aspect or any one of the implementations.

In an eighteenth aspect, a chip is provided, the chip includes a processor and a data interface, the processor reads an instruction stored in a memory through the data interface, and executes the third aspect or any one of the implementation manners. Methods.

Optionally, as an implementation manner, the chip may further include a memory, in which instructions are stored, the processor is configured to execute the instructions stored in the memory, and when the instructions are executed, the The processor is configured to execute the method in the third aspect or any one of the implementation manners.

A nineteenth aspect provides a computing device, the computing device comprising: a memory for storing a program; a processor for executing the program stored in the memory, when the program stored in the memory is executed, the The processor is configured to execute the method in the first aspect or any one of the implementation manners.

In a twentieth aspect, a computing device is provided, the computing device comprising: a memory for storing a program; a processor for executing the program stored in the memory, when the program stored in the memory is executed, the The processor is configured to execute the method in the second aspect or any one of the implementation manners.

A twenty-first aspect provides a computing device, the computing device comprising: a memory for storing a program; a processor for executing the program stored in the memory, when the program stored in the memory is executed, the The processor is configured to execute the method in the third aspect or any one of the implementation manners.

Description of drawings

FIG. 1 is an example diagram of an application scenario of the technical solution of the present application.

FIG. 2 is an exemplary interaction flow chart of the control method of the present application.

FIG. 3 is an example diagram of transmitting probe packets and data packets according to the proportional bandwidth in the control method of the present application.

FIG. 4 is another exemplary flowchart of the control method of the present application.

FIG. 5 is another exemplary flowchart of the control method of the present application.

FIG. 6 is still another exemplary flowchart of the control method of the present application.

FIG. 7 is an exemplary structural diagram of the control device of the present application.

FIG. 8 is another exemplary structural diagram of the control device of the present application.

FIG. 9 is an exemplary structural diagram of the computer program product of the present application.

detailed description

The technical solutions in the present application will be described below with reference to the accompanying drawings.

The technical solutions of the embodiments of the present application can be applied to various communication networks, such as a wide area network (WAN), a metropolitan area network (MAN), a local area network (LAN), or a data center network, etc.

The source device in this embodiment of the present application may also be referred to as a source host, a source end, a source node, or a source side, etc., and refers to the sending end of data. Further, the source device that sends the data can be understood as the network device indicated by the source Internet Protocol (IP) address of the data. In the embodiments of the present application, the source device that sends the data is simply referred to as the source device of the probe packet or the data packet.

The destination device in the embodiments of the present application may also be referred to as a destination host, a destination end, a destination node, or a destination side, etc., and refers to the final receiving end of the data, that is, the destination of the data. Further, the destination device that receives the data can be understood as the network device indicated by the destination IP address of the data. In the embodiments of the present application, the destination device that receives the data is simply referred to as the destination device of the data.

The data in this embodiment of the present application may include data packets and probe packets. For the meaning of the data packet, reference may be made to the definition of the data packet in the prior art, which will not be repeated here.

The probe packet in this embodiment is used to detect the load of the transmission path between the source device and the destination device of the probe packet. The load condition of the transmission path may include the available bandwidth condition of the transmission path, for example, how much the available bandwidth of the transmission path is. Alternatively, it can be understood that the probe packet is used to detect or reserve the bandwidth that can be used for transmitting the data packet in the next transmission period.

The length of the detection packet may be preset, for example, a detection packet may be encapsulated into an Ethernet frame of 64 bytes (byte, B). Generally, the probe packet may only include a header, and the content included in the header may be the same as the content in the header of the data packet sent in the subsequent steps, for example, the same quintuple information.

The intermediate device in this embodiment may also be referred to as an intermediate node, a forwarding device, a transit device, a transit node, a switch, a routing device, a router, or a routing node, and refers to a communication device capable of forwarding data sent by a source device to a destination device.

FIG. 1 is a schematic structural diagram of a communication network to which the technical solutions of the embodiments of the present application can be applied. As shown in FIG. 1 , a communication network to which the technical solutions of the embodiments of the present application can be applied may include a source host 110 , an intermediate node 120 and a destination host 130 .

It can be understood that only one source host, one intermediate node and one destination host are shown in FIG. 1 as an example. The communication network to which the technical solutions of the embodiments of the present application can be applied may include more source hosts, more intermediate nodes, and more destination hosts.

In the communication network shown in FIG. 1 , the source host 110 serves as the sender of the data packet, and sends the data packet at a certain rate. After the data packet reaches the intermediate node 120, the intermediate node 120 sends the data packet to the destination host 130 using the available bandwidth.

Wherein, if the sending rate of the data packets is too fast, when the intermediate node 120 does not have time to forward the received data packets, the data packets that are not forwarded in time will be backlogged in the intermediate node 120 . When the backlog of data packets in the cache of the intermediate node 120 reaches a certain level, the intermediate node 120 will discard the received data packets that have not been forwarded in time, thereby causing packet loss of data packets. At this point, it can be considered that the network is congested.

In order to avoid packet loss, that is, to avoid congestion, the present application proposes a new technical solution to control the congestion of the communication network. FIG. 2 is an exemplary flowchart of a method for controlling network congestion according to an embodiment of the present application. The method may include S210, S220, S230, S240, S250 and S260.

S210, the first network device sends a first detection packet. The first network device is the source device of the first detection packet, for example, the source IP address in the first detection packet is the IP address of the first network device.

The first network device may be any source device in the communication network, and the size of the first detection packet may be preset.

In some possible implementations of this embodiment, the first network device may send the first detection packets packet by packet at equal intervals.

In this embodiment, sending a detectable packet by the first network device may be understood as the first network device sending a first probe packet to a second network device, where the second network device is the destination device of the first probe packet, for example, The network device indicated by the destination IP address in the first detection packet is the second network device.

In this embodiment, after the first network device sends the first detection packet, the third network device on the transmission path from the first network device to the second network device may receive the first detection packet.

S220, the third network device forwards the first detection packet.

In some examples, the third network device may receive probe packets sent by one or more network devices to one or more other network devices, and forward these probe packets, where these probe packets may include the first network device to The first detection packet sent by the second network device.

One or more third network devices may be included on the transmission path from the first network device to the second network device. If the transmission path from the first network device to the second network device includes a third network device, after the third network device sends the probe packet, the second network device receives the first probe packet in the probe packet; If the transmission path from the first network device to the second network device includes multiple third network devices, each third network device forwards the probe packet to the next device after receiving the probe packet sent by the previous hop device. , the detection packet includes the first detection packet until the second network device receives the first detection packet.

After the first detection packet is forwarded by one or more third network devices, the second network device may receive the first detection packet.

S230, the second network device sends the first feedback information. The first feedback information is used to indicate the reception status of the first probe packet on the second network device.

For example, the first feedback information may be specifically used to indicate that the second network device has received the first probe packet. For another example, the first feedback information may be specifically used to indicate the time when the second network device receives the first probe packet. For another example, the first feedback information may be specifically used to indicate that the second network device receives the first probe packet and the time when the second network device receives the first probe packet.

In some possible implementations, after receiving the first probe packet, the second network device may modify the destination address in the first probe packet to the source address of the first probe packet, and modify the new probe packet obtained by modification. The message is sent out.

Optionally, the detection packet in this embodiment may include a trip identifier, and the trip identifier may include an outbound identifier and a return identifier. The outbound identifier is used to indicate that the detection packet is sent from the source device to the destination device, and the return identifier Used to identify the destination device of the probe packet returned to the source device. In this case, the second network device not only modifies the destination address of the first probe packet, but also modifies the trip identifier of the first probe packet from the outgoing identifier to the return identifier.

After the second network device sends the first feedback information, the third network device on the transmission path from the second network device to the first network device may receive the first feedback information.

In this embodiment, each time the second network device receives a detection packet, it may send a piece of feedback information corresponding to the detection packet. It can be understood that this is only an example, and the second network device can also use other methods to send the first feedback information. For example, after receiving multiple probe packets, the same message can be carried in multiple probe packets. corresponding feedback information.

S240, the third network device sends the first feedback information.

It can be understood that the transmission path from the first network device to the second network device and the transmission path from the second network device to the first network device may be the same or different. That is to say, the third network device in this step and the third network device in S220 may be the same device, or may be different devices.

For example, when the transmission path from the first network device to the second network device is different from the transmission path from the second network device to the first network device, the third network device in this step and the third network device in S220 are different devices .

In some implementation manners, the third network device may send the feedback information as a data packet. That is, the manner in which the third network device forwards the feedback information may be the same as the manner or resource for forwarding the data packet.

One or more third network devices may be included on the transmission path from the second network device to the first network device. If the transmission path from the second network device to the first network device includes a third network device, after the third network device sends the first feedback information, the first network device receives the first feedback information; if the second network device to The transmission path of the first network device includes multiple third network devices. After receiving the first feedback information sent by the previous hop device, each third network device forwards the first feedback information to the next hop device until the third network device receives the first feedback information sent by the previous hop device. A network device receives the first feedback information.

S250, the first network device sends a first data packet based on the first feedback information. The source device of the first data packet is the first network device, and the destination device of the first data packet is the second network device.

Generally speaking, the path traversed by the first data packet is the same as the path traversed by the first probe packet. Therefore, after the first network device sends the first data packet, it forwards the first data packet from the first network device to the third network device of the second network device, and forwards the first detection packet from the first network device to the third network device of the second network device. The third network device of the second network device is identical.

S260, the third network device receives the data packet, and forwards the data packet, where the data packet includes the first data packet.

Usually, the third network device in this step is the third network device in S220, that is, the path traversed by the first data packet is the same as the path traversed by the first detection packet.

After the third network device sends the data packet, the second network device may receive the first data packet in the data packet.

In this embodiment, the network status of the communication network is sensed by sending probe packets in advance, and the sending of data packets is guided, so that congestion and packet loss will not occur on the data packet channel, that is, the lossless "foresight" is achieved. congestion control.

In the embodiment of the present application, the transmission of the probe packet and the data packet may be isolated in various manners.

In some manners, the third network device may allocate bandwidth for the probe channel for sending probe packets and the data channel for sending data packets in the same physical link according to a certain ratio, so as to logically allocate bandwidth to the two types of data packets. The channel is isolated, so as to logically isolate probe packets and data packets. For example, the ratio may be the ratio of the respective packet lengths of probe packets and data packets to the total packet lengths of these two types of packets.

In other manners, the source device may transmit the probe packet by constructing an independent signaling network, so as to realize the physical isolation of the probe packet and the data packet.

In an example in which the probe packet and the data packet are transmitted in a logically isolated manner, the third network device may use the bandwidth of the first proportion to send the probe packet, where the probe packet includes the first probe packet, and, The third network device uses the bandwidth of the second ratio to send a data packet, where the data packet includes the first data packet.

The first ratio and the second ratio may be preset. That is to say, the proportion of the bandwidth used by the third network device to send the probe packets and the proportion of the bandwidth used to send the data packets are preset.

In some possible implementations, the first ratio and the second ratio on the third network device may be set based on the sizes of the probe packets and the data packets.

For example, in the scenario where a probe packet is constructed with a minimum Ethernet frame of 84 bytes, and the source device is driven to transmit data packets with a maximum Ethernet frame of 1538 bytes, the bandwidth allocated by the third network device for the probe packet accounts for the third The ratio of the total bandwidth of the network device is about: 84/(84+1538)≈5% The ratio of the bandwidth allocated for the data message to the total bandwidth is about 95%.

FIG. 3 is a schematic diagram of a method for proportionally sending probe packets and data packets according to an embodiment of the present application. As shown in FIG. 3 , the intermediate node 330 between the source host 310 and the destination host 320 may reserve 5% of the bandwidth to send probe packets, and use the remaining 95% of the bandwidth to send data packets. No matter how much the current available bandwidth of the intermediate node 330 is, it can only use 5% of the available bandwidth to send probe packets, and can only use 95% of the available bandwidth to send data packets.

If the bandwidth of the outgoing port of the intermediate node 330 is 10 gigabits per second (Gbps), the intermediate node 330 uses a bandwidth of 9.5 Gbps to send data packets and a bandwidth of 0.5 Gbps to send probe packets. Wherein, if the rate at which the source host 310 sends the first detection packet is 1 Gbps, it means that the transmission quantity of the first detection packet is too fast (the packet interval is too small). The packet rate is limited.

Specifically, the intermediate node 330 will limit the transmission rate of the first probe packet to 0.5 Gbps, that is, 500 megabits per second (Mbps). The transmission rate of 500 Mbps is just the optimal transmission rate of the first probe packet. After receiving the probe packet transmitted by the intermediate node 330 at a transmission rate of 500 Mbps, the destination host 320 sends the first feedback information to the source host 310 . After receiving the first feedback information, the source host 310 can learn the load of the intermediate node 330 according to the first feedback information, that is, the current network can just transmit the first probe packet with a transmission rate of 500 Mbps to the destination host 320 . According to the ratio of the first ratio of 5% and the second ratio of 95%, it can be seen that the current network can just use the rate of 9.5Gbps to send the first data packet, so that the network resources can be fully utilized without packet loss and congestion. .

In a scenario where the intermediate node sends the probe packet and the data packet according to a certain proportion of bandwidth, in some implementation manners, the third network device may use a bandwidth other than the first proportion to send the first feedback information. For example, in the case where the first feedback information is a probe packet obtained by modifying the first probe packet, the third network device may determine, based on the backhaul identifier in the probe packet, that the probe packet is fed back by the second network device to the first probe packet. of the network device, so the probe packet is sent using the bandwidth used to send the data packet.

In the embodiment of the present application, when the first network device sends the first detection packet in the first round, it may send the packet in a relatively aggressive manner. In this way, network resources can be fully utilized.

Generally speaking, when the first network device sends the first probe packet in the first round, the sending rate of the first probe packet may be higher than the bandwidth requirement of the application on the first network device. For example, the sending rate of the first probe packet may be initialized to be 1.5 times the bandwidth requirement of the application.

In the embodiment of the present application, during the non-first-round sending process of the first detection packet, the first network device may adjust the transmission rate of the first detection packet, so as to achieve a higher transmission rate of the first detection packet. The purpose of control, so as to achieve the purpose of controlling the sending rate of the first data packet, thereby achieving the purpose of controlling network congestion.

In some implementation manners of adjusting the sending rate of the first probe packet, if the current rate used by the first network device to send the first probe packet is called the third rate, the first network device can obtain the first probe packet. The first packet loss rate when the first detection packet is sent at the third rate is used, the second rate is determined according to the first packet loss rate, and the first detection packet is sent based on the second rate.

In the embodiments of the present application, a structure-conduct-performance (SCP) strategy may be used to obtain the second rate based on the third rate and the first packet loss rate. For example, the online learning-based congestion control method in the SCP strategy can be adopted.

In the SCP strategy, the first network device can calculate the utility function fw _,b (r _C ,l) based on the current transmission rate r _C and the packet loss rate l in real time, and according to the change direction of the utility function (increase or decrease) to adjust the change direction (increase or decrease) of the transmission rate of the first probe packet, that is, increase the third rate to obtain the second rate or decrease the third rate to obtain the second rate.

An example of a utility function f _w,b (r _C ,l) is as follows:

Among them, s, a, b and w are parameters, and the specific values can be obtained through online learning; r _P represents the sending rate of probe packets in the previous cycle, and l ₀ represents the target packet loss rate.

In the embodiment of the present application, it may be determined whether the transmission of the first probe packet is about to be in a light-load mode or a heavy-load mode based on the first packet loss rate and/or the second rate. For example, if the second rate is greater than the preset rate threshold, it is considered that the transmission of the first detection packet is about to be in a light-load mode, otherwise, it is considered that it is about to be in a heavy-load mode. For another example, if the first packet loss rate is less than the preset packet loss rate threshold, it is considered that the transmission of the first detection packet is about to be in a light-load mode; otherwise, it is considered that the transmission of the first detection packet is about to be in a heavy-load mode.

If the transmission of the first detection packet is about to be in the light-load mode, the first detection packet is sent at the second rate; if the transmission of the first detection packet is about to be in the heavy-load mode, the first detection packet is sent at the preset target rate Probe packets.

In some implementations, the set target rate should meet the following conditions: when the probe packet forwarded by the third network device on the transmission path from the first network device to the second network device only includes the first probe packet, the first network device The device can successfully transmit a specified number of first probe packets from the first network device to the second network device within a specified period of time using the target rate. In this way, when the transmission of the first detection packet is switched from the light-load mode to the heavy-load mode, for example, when the third rate is greater than the target rate and the second rate is less than or equal to the target rate, the first detection packet is not sent at a reduced speed. , so that the interruption of the flow of the first detection packet can be avoided, so that the interruption of the flow of the first data packet can be avoided. Especially when multiple source devices send data packets at high speed at the same time, the problem of current interruption can be avoided.

In some implementation manners, the target rates of probe packets of all source devices in the multiple source devices may be the same. This is to achieve fair convergence of the sending rates of probe packets of these source devices.

In some implementations, the preset rate threshold may be equal to the preset target rate, which enables the first network device to achieve rapid convergence of the transmission rate of the first probe packet when switching from the light-load mode to the heavy-load mode.

In the light load mode, after obtaining the second rate based on the third rate and the first packet loss rate, it may be determined whether to continue to maintain the light load mode or switch to the heavy load mode based on the second rate and the preset target rate.

In some implementation manners of the embodiments of the present application, the first network device may determine, according to the third rate, whether the transmission of the first probe packet has been in a light-load mode or a heavy-load mode. For example, if the third rate is greater than the preset rate threshold, it means that it is currently in the light-load mode; otherwise, the specification is currently in the heavy-load mode.

If it is currently in the light load mode, the first detection packet is sent using the adjusted second rate; if it is in the heavy load mode, the first detection packet is sent using the preset target rate. For the characteristics of the rate threshold and the target rate in this implementation manner, reference may be made to the foregoing related content of the rate threshold and the target rate.

In some implementations of the embodiments of the present application, after acquiring the first packet loss rate, the first network device may first determine whether to continue using the target rate to send the first packet loss rate based on the first packet loss rate and a preset packet loss rate threshold. The probe packet still adjusts the third rate and uses the adjusted second rate to send the first probe packet.

That is to say, according to the first packet loss rate, after judging that the first detection packet is transmitted at the third rate, whether the transmission of the first detection packet is currently in the light-load mode or the heavy-load mode, if it is currently in the light-load mode, use the adjustment The obtained second rate is used to send the first detection packet; if currently in the overload mode, the first detection packet is transmitted using the preset target rate. In this way, network resources can be fully utilized to send detection packets, and thus network resources can be fully utilized to transmit data packets.

In this implementation manner, when the third rate is adjusted to obtain the second rate, the SCP policy may be used to obtain the second rate based on the third rate and the first packet loss rate, and reference may be made to the foregoing manner for the acquisition method.

Wherein, optionally, the first packet loss rate may be equal to the parameter l ₀ in the utility function. This enables the first network device to continue the iterative convergence process of the transmission rate of the first probe packet when the first network device switches to the light-load mode, and realize fast and fair convergence of the first probe packet.

In some implementations of the embodiments of the present application, each time the second network device receives a first probe packet, it sends a piece of first feedback information, the destination device of the first feedback information is the first network device, and the The source device is the second network device. For example, the first feedback information may be a probe packet obtained by exchanging and modifying the destination address and source address of the first probe packet. In this implementation manner, each time the first network device receives a piece of first feedback information, it sends a first data packet.

In this implementation manner, if the second network device can successfully receive a first detection packet, it can accordingly successfully receive a first data packet. Therefore, this implementation manner can avoid congestion caused by the first data packet and packet loss of the first data packet.

In some implementations of the embodiments of the present application, the operating system kernel of the source host needs to be modified accordingly, and a new transport layer protocol can be written, for example, a traditional transmission control protocol (transmission control protocol, TCP,), user data A protocol in which transport layer protocols such as user datagram protocol (UDP) are juxtaposed.

In addition, the intermediate node also needs to make relevant modifications at the software level. For example, the intermediate node needs to reserve a certain buffer for the detection packet to store the burst of the detection packet, and the outbound port of the intermediate node needs to reserve a certain proportion of bandwidth. It is specially used to send probe packets, and the probe packets that exceed the reserved bandwidth are discarded; normal data packets can be sent using a certain percentage of bandwidth, for example, using a link other than the bandwidth reserved for probe packets. bandwidth to transmit data packets.

Since the probe packet is adjusted by the rate of the intermediate node, the data packet transmission rate guided by the probe packet rate is just right. In this way, through the cooperation between the source host and the intermediate node, lossless transmission of data packets, that is, transmission without packet loss, can be realized.

The technical solution of the present application can be used for congestion control of high concurrent data streams in a low-latency network. In some examples, dual congestion signals are introduced to adjust the transmission rate of the first probe packet in light-load and heavy-load modes, respectively. For example, whether to switch from the light-load mode to the heavy-load mode is determined according to whether the transmission rate of the first detection packet is lower than a preset rate threshold; by judging whether the current packet loss rate of the first detection packet is lower than the preset rate threshold The packet loss rate threshold determines whether congestion control switches from heavy load mode to light load mode.

In an example of congestion control in the present application, when a source host has data to transmit, it first transmits probe packets in the control channel at equal intervals according to the initialization rate until it receives probe packets returned by the destination host. After the first round of data transmission, in the next data transmission process of the source host, the sending rate of data packets will be regulated based on the sending rate and receiving conditions of probe packets, while the sending rate of probe packets is based on the RTT period, with The packet loss rate of detection packets is regulated.

For example, the source host initializes the timer to zero, and each time the source host receives a probe packet, it sends a data packet, and at the same time updates the timer time t and the packet loss rate l of the probe packet. After that, the source host compares the size of the current time domain RTT.

If the timer time t does not reach the duration of one RTT, continue to execute the operation of "receive the probe packet - send the data packet - update the timer - update the packet loss rate of the probe packet"; if the timer time t reaches a The duration of the RTT is compared with the size of the current detection packet sending rate r _C and the target rate r ₀ to determine whether to enter the heavy load mode.

If r _C >r ₀ , implement the congestion control method in light load mode, that is, increase or decrease the sending rate of probe packets in the next round according to the specified algorithm, and reset the packet loss rate l and timer time t ; otherwise, it enters the overload congestion control mode.

In overload congestion control mode, the source host sends probe packets at a fixed rate r ₀ . In this mode, the sending rate of probe packets will not be adjusted, and the source host always sends probe packets at equal intervals at the rate of r ₀ . Similarly, the source host cyclically executes the operations of "receiving probe packets - sending data packets - updating the timer - updating the packet loss rate of probe packets". When the timer time t reaches an RTT duration, the source host compares the current detection packet loss rate 1 with the preset packet loss rate threshold 1 ₀ to determine whether to switch back to the light-load congestion control mode.

If l<l ₀ , switch back to the previous light-load congestion control mode, and adjust the sending rate of probe packets according to the specified congestion control algorithm; otherwise, continue the congestion control process in the heavy-load mode, that is, continue at a fixed rate r ₀ sends probe packets at equal intervals, and resets the probe packet loss rate l and timer time t.

The third network device reserves a part of the link bandwidth for transmitting the probe packets from the source host to the destination host. When the third network device receives the data, it first performs analysis and judgment. If it is determined that the data is a probe packet from the source host to the destination host, the packet is transmitted using the reserved link bandwidth; otherwise, the data is transmitted using the unreserved link bandwidth.

Once the destination host receives the message sent by the source host, it will parse and judge. If it is determined that the packet is a detection packet, the source and destination addresses of the packet are changed, and the detection packet is sent back to the source host. If the packet is a data packet, normal parsing and processing operations are performed.

In order to ensure that the detection packet returned by the destination host to the source host can be smoothly transmitted back to the source host, the third network device may additionally increase the transmission priority of the detection packet.

FIG. 4 is an exemplary flowchart of a method for controlling network congestion according to an embodiment of the present application. For example, in a low-latency network, the network congestion control method shown in FIG. 4 can be applied.

Each step of the control method is described in detail below. The control method can be executed by the source host from the start of data transmission until the data transmission is completed.

S401, the source host needs to transmit data.

S402, determine whether the round of data transmission is the first round of transmission, and enter into different branch steps according to the judgment result. If the transmission is the first round of data transmission, go to S403; if the transmission is not the first round of data transmission, go to S404.

S403: Initialize the first-round sending rate of the detection packet. For example, the first-round sending rate of probe packets can be initialized to 1.5 times the bandwidth requirement of the application.

S404, sending a detection packet at an initial rate. For example, probe packets are sent on the control channel at equal intervals at the initial rate. The logical channel for sending probe packets is called the control channel.

S405, the source host judges whether the detection packet returned by the destination host is received, and enters different branch steps according to the judgment result. If the detection packet returned by the destination host is received, the first round of transmission is completed, and the process proceeds to S406; if the detection packet returned by the destination host is not received, the process proceeds to S404, and the transmission of the first round of detection packets is continued.

After the first round of transmission, the transmission rate of probe packets and data packets can be adjusted. Among them, the transmission rate of data packets is adjusted according to the rate of returned probe packets; and the transmission rate of probe packets is adjusted to ensure the transmission rate of probe packets from multiple different streams (for example, from multiple different source hosts). The purpose of fairness convergence can be achieved on the shared bottleneck link. For example, the sending rate of the probe packets can be adjusted at the RTT cycle.

S406, initialize the target rate r ₀ and initialize the target packet loss rate l ₀ . The target packet loss rate in this embodiment may also be referred to as a packet loss rate threshold.

S407: Calculate the sending rate r _C of the detection packet of the current round according to the feedback information of the previous round. For example, the sending rate r _C of the current round of detection packets is obtained according to the transmission rate and packet loss rate l of the previous round of detection packets.

In light load mode, the sending rate of probe packets can be adjusted according to the value of the utility function. Specifically, the value of the utility function can be calculated, and the sending rate of the probe packet can be adjusted according to the SCP scheme according to the change of the utility function.

S408 , setting the timer t=0 and setting the packet loss rate l=0 of the detection packet.

S409 , compare the size of the calculated probe packet sending rate r _C and the target rate r ₀ , and enter different branches according to the comparison result. If r _C >r ₀ , go to step S410 , that is, the light-load mode; if r _C ≦r ₀ , go to step S414 , that is, the rate adjustment of the heavy-load mode.

S410, using the latest calculated r _C to send a detection packet.

S411, each time a backhaul detection packet is received, a data packet is sent.

S412 , record the time interval Δt between the current and last time when the backhaul probe packet is received, and update the timer t=t+Δt; and update the probe packet loss rate l according to the sequence number of the backhaul probe packet.

S413, determine whether the current timer duration reaches the RTT size, and enter different branch steps according to the determination result. If the timer duration reaches the RTT size, enter S414, that is, the rate adjustment in the light load mode. If the timer duration does not reach the RTT size, continue to S410.

S414 , send a probe packet at a fixed target rate r ₀ . Since packets are detected at a fixed sending rate, and the sending rate is not adjusted, fairness convergence is naturally satisfied.

S415, each time a returned probe packet is received, a data packet is sent.

S416, update the detection packet loss rate l, and update the timer t=t+Δt.

S417, determine whether the current timer duration reaches the RTT size, and enter into different branch steps according to the determination result. If the timer duration reaches the RTT size, proceed to S418; if the timer duration does not reach the RTT size, proceed to S414.

S418, it is determined that l<l0, that is, it is determined whether to switch the light and heavy load mode. Specifically, compare the packet loss rate l of the current round of detection packets and the target packet loss rate l0, and enter different branch steps according to the comparison result. If l<l0, go to S407, that is, the rate control of the light load mode is performed; if l≥l0, go to S419.

S419 , set the timer t=0 and set the packet loss rate l=0 of the detection packet, and continue to perform S414 to continue the rate control in the reload mode.

S407 to S419 are continuously executed after the first round of transmission ends until the entire data transmission process from the source end to the destination end ends.

In this embodiment, after the pre-rehearsal transmission of the detection packets, the transmission rate of the data packets can ensure that the network bandwidth is fully utilized and absolutely no packet loss is ensured. Among them, in the light-load mode, the control algorithm of the SCP policy can achieve fast and fair convergence; in the heavy-load mode, all source nodes send probe packets at a uniform rate, which naturally satisfies the fairness convergence and further improves the overall fairness convergence of the policy. speed. In addition, the criterion for switching from heavy load mode to light load mode is l≥l0, while the convergence point of rate regulation in light load mode is l=0. Therefore, after switching back to light load mode, the whole strategy will continue to iteratively adjust the rate. Switching does not break the fairness convergence process in light-load mode.

A schematic flowchart of another embodiment of the method for controlling network congestion of the present application is shown in FIG. 5 . For example, the control method shown in FIG. 5 can be applied in a low-latency network.

This method can be executed by the destination host. Each time the destination host receives a probe packet or data packet, it executes the operations in the method.

S510, the destination host receives data.

S520, after receiving the data, the destination host first determines whether the data is a probe packet or a data packet. If the received packet is a data message, go to S530; if the received packet is a probe message, go to S540.

S530: Perform parsing processing on the data packet. The processing of data packets can use the working mode of traditional transport layer protocols to decapsulate and process data packets.

S540, returning a detection packet. For example, the source IP address of the probe packet is exchanged with the destination IP address, and the probe packet is sent back to the source host.

Each time the destination host receives a probe packet or a data packet, it executes the method shown in FIG. 5 , parses and processes the data packet, and returns the probe packet.

A schematic flowchart of another embodiment of the method for controlling network congestion of the present application is shown in FIG. 6 . For example, the control method shown in Figure 6 can be applied in a low-latency network.

The method can be performed by a third network device of the network, and lasts throughout the entire end-to-end transmission process. The third network device allocates a certain percentage of bandwidth to each link, which is dedicated to transmitting probe packets, for example, allocating 5% of the bandwidth to transmit probe packets; using other link bandwidths other than the allocated bandwidth to transmit data packets, such as Use the other 95% of the bandwidth to transmit datagrams.

S610, the third network device receives the data transmitted from the previous hop device. The last hop device of the third network device may be the source host, or may be another third network device.

S620, after receiving the data from the previous hop device, the third network device first determines whether the received packet is a data packet or a probe packet. If the received message is a data message, the process goes to step S630; if the received message is a probe message, the process goes to S640.

S630: Use the bandwidth other than the bandwidth allocated for the probe packet to transmit the data packet. In other words, use unreserved bandwidth to transmit data packets.

S640: Determine whether the detection packet is an outgoing detection packet or a return detection packet, and enter different branch steps according to the judgment result. The outgoing probe packet refers to the probe packet sent from the source host to the destination host, and the backhaul probe packet refers to the probe packet sent from the destination host to the source host.

If the detection packet is an outgoing detection packet, proceed to step S650; if the detection packet is a return detection packet, proceed to step S660.

S650. Use the pre-allocated bandwidth ratio to transmit outgoing detection packets. The third network device may reserve a small amount of buffers, for example, 8 buffers of the size of the detection packets are reserved to deal with the burst transmission of the detection packets. When the probe packet buffer is full, the redundant probe packets received are discarded.

S660, using the bandwidth used for transmitting the data packet to transmit the backhaul detection packet. Among them, the backhaul detection packet is returned with the highest priority.

Executing the methods in Figures 5 and 6 is equivalent to using the detection packet as a "stand-in" for the data packet, rehearsing the data transmission process in the current network environment, and detecting the packet loss and destination host of the packet through a third network device The return of the probe packet conveys the exact load status of the network to the source host. In this way, the source host in FIG. 4 can precisely control the transmission rate of data packets according to the network conditions reflected in the detection packets, thereby realizing the rapid convergence and lossless of the entire congestion control process. The entire congestion control process is coordinated by end-to-end devices (source host, destination host, and a third network device in the network).

The embodiments of FIG. 4 to FIG. 5 specially design the congestion control scheme in the overload mode, so that the end-to-end transmission with low delay, high throughput and no congestion can still be maintained in the overload mode. Among them, taking advantage of the dual-channel (the channel for transmitting probe packets and the channel for transmitting data packets) architecture, in overload mode, all source hosts can send probe packets at a fixed rate, and the sending rate is relatively aggressive, so that the The control channel that sends the probe packets is continuously in a state of packet loss, so that the data channel can ensure full throughput and no packet loss when transmitting data packets. In addition, in the reload mode, the transmission of probe packets does not slow down, and the source host can always receive the echo probe packets and send data packets, thus avoiding the problem of current interruption.

In addition, the embodiment shown in FIG. 4 selects an appropriate mode switching point. When the congestion control is switched from the heavy-load mode to the light-load mode, the light-load mode will not immediately determine that the convergence point is reached, but will continue to move towards the convergence point. The iterative control of the transmission rate, and finally converges to the established steady-state transmission rate, ensures that the switching between the light-load mode and the heavy-load mode will not affect the fairness convergence process of the light-load mode; while the congestion control switches from the light-load mode to the heavy-load mode. In the load mode, the entire transmission will immediately converge to the steady-state rate, which speeds up the convergence process of the entire congestion control scheme to a certain extent.

The congestion control method proposed in this application can also be applied to a data center network. In the light load mode of the data center network, the data center network congestion control algorithm in the academic paper "ExpressPass" published by ACM SIGCOMM in 2019 can be selected.

SIGCOMM is the flagship conference of the Association for Computing Machinery (ACM) in the field of communication networks. In the "ExpressPass" scheme, the congestion control algorithm regulates the sending rate of each data stream until the packet loss rate of all probe packets reaches the target packet loss rate (target_loss). In terms of rate control, the next cycle is calculated by the current rate (cur_rate), control weight (w), maximum rate (max_rate), current detection packet loss rate (credit_loss) and target packet loss rate (target_loss). Probe packet sending rate.

Based on the "ExpressPass" algorithm, this application introduces a rate adjustment method in the overload mode, and selects a suitable transition point for the light-heavy-load mode, wherein, when cur_rate≤r ₀ , the light-load mode is switched to the heavy-load mode; credit_loss< l When ₀ , switch from heavy load mode to light load mode.

Assuming that the network environment is still a low-latency network with an RTT of 10 microseconds (us) and a bottleneck link bandwidth of 100 Gbps, an example of implementing congestion control in a data center network is as follows:

Step 701, the source host initializes the target packet loss rate (target_loss), the maximum rate (max_rate), the weight (w=0) and the maximum weight (w _max =0.5).

Step 702, the source host initializes the performance parameters of the light-heavy-load mode conversion: r ₀ =100Mbps, 0<l ₀ <target_loss.

Step 703, the source host sends probe packets at equal intervals according to the application requirements and the target rate, and requests bandwidth. The third network device of the network reserves a certain percentage (for example, 5%) of the bandwidth as a control channel to transmit the detection message.

Step 704: During the outgoing journey, the hop-by-hop outgoing port of the third network device performs a Committed Access Rate (CAR) rate limit on all probe packets according to a certain bandwidth ratio (eg, 5%), and discards probe packets exceeding the port capability.

Step 705, the destination host exchanges the source address and destination address of the received probe packet, and returns the probe packet.

Step 706, the third network device returns the detection packet to the source host with the highest priority. During the backhaul, the third network device does not limit the rate of detection packets.

Step 707, the source host immediately sends a data packet every time it receives a probe packet.

After an RTT period, the source host determines whether to switch between the light-load mode and the heavy-load mode according to the packet loss rate and transmission rate of the probe packets.

If cur_rate≤r ₀ , switch to heavy load mode, and send probe packets at a fixed rate r ₀ until the next state switch; if cur_rate>r ₀ , perform rate regulation in light load mode.

In light load mode, if the current packet loss rate is less than the target packet loss rate, the speed will be increased, and the rate is calculated as follows: cur_rate=(1-w)*cur_rate+w*max_rate*(1+target_loss); if the current packet loss rate is If the rate is greater than the target packet loss rate, the speed is reduced, and the rate is calculated as follows: cur_rate=cur_rate*(1-credit_loss)*(1+target_loss).

In heavy load mode, if credit_loss<l ₀ , switch to light load mode, and perform rate convergence according to the target packet loss rate; if credit_loss≥l ₀ , continue to send probe packets at a fixed rate r ₀ .

It can be known from the above method that the technical solution of the present application can be applied to a data center network that is more sensitive to delay. Among them, the transition point from the heavy load mode to the light load mode is credit_loss<l ₀ . Since l ₀ <target_loss, after the mode switching, the process of iterative speed adjustment will still be performed, and the fairness convergence process of the original scheme will not be destroyed. . Similarly, when the congestion control is switched from the light load mode to the heavy load mode, the entire transmission will immediately converge to the steady-state rate (the convergence rate of the heavy load mode is r ₀ ), which accelerates the convergence process of the entire congestion control scheme to a certain extent.

This embodiment further illustrates that the technical solution of the present application can be perfectly compatible with any other congestion control solutions based on transmission rate and packet loss rate.

In addition, in the data center network, the "ExpressPass" scheme with fewer iterations is selected as the light-load scheme. Compared with the online learning-based scheme, the "ExpressPass" scheme has relatively few performance detection operations, that is, it does not need to repeatedly calculate the utility function. In order to determine the change of the transmission rate, the transition from the light load mode to the heavy load mode will be faster, and it is more suitable for the data center network with sensitive delay and strong burst.

FIG. 7 is a schematic structural diagram of an apparatus 700 for controlling network congestion according to an embodiment of the present application. The apparatus 700 may include a sending module 710 and a receiving module 720 .

In some examples, the apparatus 700 may be used to perform the relevant steps or operations performed by the source host in the method shown in FIG. 2 , for example, the sending module 710 may be used to perform the relevant operations in S210 and S250, and the receiving module 720 may be used to perform Relevant operations in S240.

In another example, the apparatus 700 can be used to perform the relevant steps or operations performed by the intermediate node in the method shown in FIG. 2 , for example, the sending module 710 can be used to perform the relevant operations in S220 and S260, and the receiving module 720 can be used to perform the relevant operations in S220 and S260. Relevant operations in S210, S230 and S250 are performed.

In still other examples, the apparatus 700 can be used to perform the relevant steps or operations performed by the destination host in the method shown in FIG. 2 , for example, the sending module 710 can be used to perform the relevant operations in S230, and the receiving module 720 can be used to perform S220 and related operations in S260.

In some other examples, the apparatus 700 may be used to perform the method shown in FIG. 4 , or the apparatus 700 may be used to perform the method shown in FIG. 5 , or the apparatus 700 may be used to perform the method shown in FIG. 6 .

FIG. 8 is a schematic structural diagram of an apparatus 800 for controlling network congestion according to an embodiment of the present application. The apparatus 800 includes a processor 802 , a communication interface 803 and a memory 804 . One example of the apparatus 800 is a chip, and another example of the apparatus 800 is a computing device.

The processor 802, the memory 804 and the communication interface 803 can communicate through a bus. Executable code is stored in the memory 804, and the processor 802 reads the executable code in the memory 804 to execute the corresponding method. The memory 804 may also include other software modules required for running processes such as an operating system. The operating system can be LINUX ^™ , UNIX ^™ , WINDOWS ^™ and the like.

For example, the executable code in the memory 804 is used to implement the steps or operations performed by the source host, the intermediate node or the destination host in FIG. 2; the processor 802 reads the executable code in the memory 804 to execute the steps or operations in FIG. A step or action performed by a host, intermediate node, or destination host.

For another example, the executable code in the memory 804 is used to execute the method shown in FIG. 4 , FIG. 5 or FIG. 6 ; the processor 802 reads the executable code in the memory 804 to execute the method shown in FIG. 4 , FIG. 5 or FIG. 6 . method shown.

The processor 802 may be a CPU. Memory 804 may include volatile memory, such as random access memory (RAM). The memory 804 may also include non-volatile memory (2non-volatile memory, 2NVM), such as 2read-only memory (2ROM), flash memory, hard disk drive (HDD) or solid state drive ( solid state disk, SSD).

In some embodiments of the present application, the disclosed methods may be implemented as computer program instructions encoded on a computer-readable storage medium in a machine-readable format or on other non-transitory media or articles of manufacture. 9 schematically illustrates a conceptual partial view of an example computer program product including a computer program for executing a computer process on a computing device, arranged in accordance with at least some embodiments presented herein. In one embodiment, example computer program product 900 is provided using signal bearing medium 901 . The signal bearing medium 901 may include one or more program instructions 902 that, when executed by one or more processors, may provide for the methods described above with respect to any one of FIGS. 2 , 4 , 5 and 6 . The described function or part of the function. Thus, for example, in the embodiment shown in FIG. 4 , one or more of the features of S401 to S419 may be undertaken by one or more instructions associated with the signal bearing medium 901 . As another example, referring to the embodiment shown in FIG. 6 , one or more of the features of S610 to S660 may be undertaken by one or more instructions associated with the signal bearing medium 901 .

In some examples, the signal bearing medium 901 may include a computer readable medium 903 such as, but not limited to, a hard drive, a compact disc (CD), a digital video disc (DVD), a digital tape, a memory, a read only memory (read only memory) -only memory, ROM) or random access memory (RAM), etc. In some implementations, the signal bearing medium 901 may include a computer recordable medium 904, such as, but not limited to, memory, read/write (R/W) CDs, R/W DVDs, and the like. In some embodiments, signal bearing medium 901 may include communication medium 905, such as, but not limited to, digital and/or analog communication media (eg, fiber optic cables, waveguides, wired communication links, wireless communication links, etc.). Thus, for example, the signal bearing medium 901 may be conveyed by a wireless form of communication medium 905 (eg, a wireless communication medium conforming to the IEEE 802.11 standard or other transmission protocol). The one or more program instructions 902 may be, for example, computer-executable instructions or logic-implemented instructions. In some examples, the aforementioned computing devices may be configured to, in response to program instructions 902 communicated to the computing device via one or more of computer-readable media 903 , computer-recordable media 904 , and/or communication media 905 , Provides various operations, functions, or actions. It should be understood that the arrangements described herein are for illustrative purposes only. Thus, those skilled in the art will understand that other arrangements and other elements (eg, machines, interfaces, functions, sequences, and groups of functions, etc.) can be used instead and that some elements may be omitted altogether depending on the desired results . Additionally, many of the described elements are functional terms that can be implemented as discrete or distributed components, or in conjunction with other components in any suitable combination and position.

Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of this application.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described systems, devices and units may refer to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

In the several embodiments provided in this application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored, or not implemented. On the other hand, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and 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 in this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit.

The functions, if implemented in the form of software functional units and sold or used as independent products, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be embodied in the form of a software product in essence, or the part that contributes to the prior art or the part of the technical solution. The computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a removable hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk and other media that can store program codes.

The above are only specific embodiments of the present application, but the protection scope of the present application is not limited to this. should be covered within the scope of protection of this application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

A method for controlling network congestion, the method comprising:

The first network device sends a first detection packet, where the first detection packet refers to a detection packet in which the source device is the first network device and the destination device is the second network device, and the detection packet is used for detection The available bandwidth on the transmission path between the source device and the destination device;

receiving, by the first network device, first feedback information, where the first feedback information is used to indicate the reception status of the first probe packet on the second network device;

The first network device sends a first data packet based on the first feedback information, where the first data packet refers to a source device being the first network device and the destination device being the second network device. data message.
The method according to claim 1, wherein, based on the first feedback information, the first network device sends the first data packet, comprising:

The first network device sends, based on the first feedback information, the first data packet to the second network device through a third network device, where the third network device is configured to send using a first proportion of bandwidth Probing packets and sending data packets using a second ratio of bandwidth, wherein the first rate at which the first network device sends the first data packets and the second network device receiving the first probe The ratio of the second rate of packets is equal to the ratio of the second ratio to the first ratio.
The method of claim 2, wherein each time the second network device receives one of the first detection packets, it sends one piece of the first feedback information;

Wherein, based on the first feedback information, the first network device sends the first data packet, including:

The first network device sends one first data packet each time it receives one piece of the first feedback information.
The method according to any one of claims 1 to 3, wherein the sending, by the first network device, the first detection packet comprises:

sending, by the first network device, the first probe packet using a third rate;

obtaining, by the first network device, a first packet loss rate when the first network device uses the third rate to send the first probe packet;

The first network device determines the second rate according to the first packet loss rate;

The first network device sends the first probe packet based on the second rate.
The method according to claim 4, wherein the sending, by the first network device, the first detection packet based on the second rate comprises:

When the second rate is less than or equal to a preset rate threshold, the first network device sends the first probe packet using a preset target rate, and the target rate is greater than the second rate; or

When the second rate is greater than the rate threshold, the first network device sends the first probe packet by using the second rate.
The method of claim 5, wherein the first network device is one of a plurality of network devices, wherein all network devices in the plurality of network devices use the same target rate for sending detection packets .
6. The method of claim 5 or 6, wherein the target rate is the same as the rate threshold.
The method according to any one of claims 4 to 7, wherein when the probe packet forwarded by the third network device only includes the first probe packet, the first network device uses the The target rate is capable of successfully transmitting a specified number of first probe packets from the first network device to the second network device within a specified time period.
The method of claim 4, wherein the first network device determines the second rate according to the first packet loss rate, comprising:

When the first packet loss rate is less than a preset packet loss rate threshold, the first network device sends the first probe packet based on the first packet loss rate and the third rate performing adjustment to obtain the second rate;

The sending, by the first network device, the first detection packet based on the second rate includes:

The first network device sends the first probe packet using the second rate.
The method of claim 9, wherein the packet loss rate threshold is greater than the loss rate of the first detection packet when the first network device sends the first detection packet at the second rate package rate.
The method of any one of claims 4 to 10, wherein the second rate is based on the first packet loss rate and the second rate using a structure-behavior-performance SCP strategy or a fast-track strategy obtained at three rates.
The method according to any one of claims 1 to 11, wherein the first feedback information is specifically used to indicate that the second network device has received the first probe packet.
A method for controlling network congestion, the method comprising:

The second network device receives a first detection packet, where the first detection packet refers to a detection packet in which the source device is the first network device and the destination device is the second network device, and the detection packet is used for detection The available bandwidth on the transmission path between the source device and the destination device;

sending, by the second network device, first feedback information, where the first feedback information is used to indicate the reception status of the first probe packet by the second network device;

The second network device receives a first data packet sent by the first network device based on the first feedback information, where the first data packet means that the source device is the first network device and the destination device is the data packet of the second network device.
The method of claim 13, wherein the sending, by the second network device, the first feedback information comprises:

Each time the second network device receives one of the first probe packets, it sends one piece of the first feedback information.
The method according to claim 13 or 14, wherein the first feedback information is specifically used to indicate that the second network device receives the first probe packet.
A method for controlling network congestion, the method comprising:

The third network device receives the detection packet, where the detection packet is used to detect the available bandwidth on the transmission path between the source device and the destination device;

the third network device forwards the detection packet;

receiving, by the third network device, feedback information, where the feedback information is used to indicate that the destination device receives the probe packet;

the third device forwards the feedback information;

receiving, by the third device, a data packet sent by the source device based on the feedback information;

The third device forwards the data packet.
The control method according to claim 16, wherein forwarding the detection packet by the third network device comprises: the third network device sends the detection packet by using a bandwidth of a first proportion;

The third network device forwarding the data packet includes: the third network device sends the data packet by using a second proportion of bandwidth;

The ratio of the rate at which the third network device receives the first data packet to the rate at which the third network device receives the first probe packet is equal to the second ratio to the first ratio ratio.
The control method according to claim 16 or 17, wherein the third network device forwarding the feedback information comprises:

The third network device forwards the feedback information with the highest priority.
A device for controlling network congestion, comprising:

A sending module, configured to send a first detection packet, where the first detection packet refers to a detection packet in which the source device is the first network device to which the device belongs and the destination device is the second network device, and the detection packet is a detection packet. This document is used to detect the available bandwidth on the transmission path between the source device and the destination device;

a receiving module, configured to receive first feedback information, where the first feedback information is used to indicate the reception status of the first probe packet on the second network device;

The sending module is further configured to send a first data packet based on the first feedback information, where the first data packet means that the source device is the first network device and the destination device is the second network device data packets.
The apparatus of claim 19, wherein the sending module is specifically configured to send the first data packet to the second network device based on a third network device, and the third network device is configured to use The bandwidth of the first proportion is used to send the probe packet and the bandwidth of the second proportion is used to send the data packet, wherein the first rate at which the first network device sends the first data packet is the same as the second network device. The ratio of the second rate for receiving the first probe packet is equal to the ratio of the second ratio to the first ratio.
The apparatus of claim 20, wherein the second network device is configured to: send one piece of the first feedback information every time one of the first detection packets is received;

Wherein, the sending module is specifically configured to: send one first data message each time one piece of the first feedback information is received.
The device according to any one of claims 19 to 21, wherein the sending module is specifically configured to:

sending the first probe packet using a third rate;

acquiring the first packet loss rate when the first network device sends the first probe packet using the third rate;

determining the second rate according to the first packet loss rate;

The first probe packet is sent based on the second rate.
The apparatus according to claim 22, wherein the sending module is specifically configured to:

When the second rate is less than or equal to a preset rate threshold, use a preset target rate to send the first probe packet, where the target rate is greater than the second rate; or

When the second rate is greater than the rate threshold, the first probe packet is sent using the second rate.
The apparatus according to claim 23, wherein the first network device is one of multiple network devices, wherein all network devices in the multiple network devices use the same target rate for sending probe packets .
24. The apparatus of claim 23 or 24, wherein the target rate is the same as the rate threshold.
The apparatus according to any one of claims 22 to 25, wherein when the probe packet forwarded by the third network device only includes the first probe packet, the first network device uses the The target rate is capable of successfully transmitting a specified number of first probe packets from the first network device to the second network device within a specified time period.
The apparatus according to claim 22, wherein the sending module is specifically configured to:

When the first packet loss rate is less than a preset packet loss rate threshold, adjust the sending rate of the first detection packet based on the first packet loss rate and the third rate, to obtain the second rate;

The first probe packet is sent using the second rate.
The apparatus of claim 27, wherein the packet loss rate threshold is greater than the loss of the first detection packet when the first network device sends the first detection packet at the second rate package rate.
The apparatus of any one of claims 22 to 28, wherein the second rate is based on the first packet loss rate and the second rate using a structure-behavior-performance SCP strategy or a fast-track strategy obtained at three rates.
The apparatus according to any one of claims 19 to 29, wherein the first feedback information is specifically used to indicate that the second network device has received the first probe packet.
A device for controlling network congestion, comprising:

a receiving module, configured to receive a first detection packet, where the first detection packet refers to a detection packet in which the source device is a first network device and the destination device is a second network device to which the control device belongs, and the detection packet The message is used to detect the available bandwidth on the transmission path between the source device and the destination device;

a sending module, configured to send first feedback information, where the first feedback information is used to indicate the reception status of the first probe packet by the second network device;

The receiving module is further configured to receive a first data packet sent by the first network device based on the first feedback information, where the first data packet means that the source device is the first network device and the destination device is a data packet of the second network device.
The apparatus according to claim 31, wherein the sending module is specifically configured to send one piece of the first feedback information every time one of the first probe packets is received.
The apparatus according to claim 31 or 32, wherein the first feedback information is specifically used to indicate that the second network device receives the first probe packet.
A device for controlling network congestion, comprising:

a receiving module, configured to receive a detection packet, where the detection packet is used to detect the available bandwidth on the transmission path between the source device and the destination device;

a sending module, configured to forward the detection message;

The receiving module is further configured to receive feedback information, where the feedback information is used to indicate that the destination device receives the probe packet;

The sending module is further configured to forward the feedback information;

The receiving module is further configured to receive a data packet sent by the source device based on the feedback information;

The sending module is further configured to forward the data message.
The apparatus of claim 34, wherein the sending module is specifically configured to:

sending the probe packet using the bandwidth of the first proportion;

sending the data packet using the second proportional bandwidth;

The ratio of the rate at which the third network device to which the control device belongs receives the first data packet to the rate at which the third network device receives the first probe packet is equal to the second ratio to the rate at which the first probe packet is received. the ratio of the first ratio.
The apparatus according to claim 34 or 35, wherein the sending module is specifically configured to: forward the feedback information with the highest priority.
An apparatus for controlling network congestion, comprising: a processor coupled to a memory;

the memory is used to store instructions;

The processor is configured to execute instructions stored in the memory to cause the network device to implement the method of any one of claims 1 to 18.
A computer readable medium comprising instructions which, when executed on a processor, cause the processor to implement the method of any one of claims 1 to 18.