WO2021143913A1

WO2021143913A1 - Congestion control method, apparatus and system, and storage medium

Info

Publication number: WO2021143913A1
Application number: PCT/CN2021/072574
Authority: WO
Inventors: 杨文斌; 王小忠; 李广
Original assignee: 华为技术有限公司
Priority date: 2020-01-19
Filing date: 2021-01-18
Publication date: 2021-07-22
Also published as: CN113141313A

Abstract

Disclosed are a congestion control method, apparatus and system, and a storage medium, belonging to the technical field of networks. The method comprises: a receiving-end device monitoring a plurality of queues of the receiving-end device, and notifying, when a first queue of the plurality of queues of the receiving-end device satisfies a preset start index, a sending-end device to start congestion control for a target queue, which corresponds to the first queue, in the sending-end device, wherein the preset start index comprises the amount of a cache space of the receiving-end device occupied by a plurality of queues of the receiving-end device reaching or exceeding a start back-pressure waterline value of the first queue. The present application facilitates solving the congestion problem of the current communication network, and facilitates an improvement in the resource utilization rate and the transmission quality of the communication network.

Description

Congestion control method, device, system and storage medium

This application requires a patent application filed on January 19, 2020 with an application number of 202010060644.5 and an invention title of "Congestion Control Method, Device and System, and Storage Medium", and the application number filed on January 22, 2020 as 202010075761.9 1. The priority of the Chinese patent application titled "Congestion Control Method, Device and System, Storage Medium", the entire content of which is incorporated into this application by reference.

Technical field

This application relates to the field of network technology, and in particular to a congestion control method, device and system, and storage medium.

Background technique

A communication network usually includes multiple network devices, and the network devices implement communication between different workstations by forwarding messages from workstations (for example, a host or a server, etc.). Among them, a network device usually has multiple egress ports, each egress port has multiple queues with different priorities, and one queue of each egress port is used to buffer a message of a priority that needs to be forwarded via the egress port. For each received message, the network device usually buffers the message into a queue matching the message, and sends the messages in the queue at a certain rate.

With the expansion of the scale of communication networks and the popularization of various applications such as video, high-performance cloud computing and consumer cloud services, the traffic (such as the number of packets) that network devices need to forward has skyrocketed, and network congestion has become a major problem faced by communication networks. . Network congestion will cause the resource utilization rate and transmission quality of the communication network to decrease, and even cause the communication network to be paralyzed. Therefore, a congestion control scheme is urgently needed to control the congestion of the communication network.

Summary of the invention

The embodiments of the present application provide a congestion control method, device, system, and storage medium. The technical solutions are as follows:

In a first aspect, a congestion control method is provided. The method includes: monitoring multiple queues of a receiving end device, each queue of the receiving end device corresponds to one or more queues in the sending end device; when the receiving end device When the first queue of the multiple queues of the device meets the preset opening index, the sending end device is notified to enable congestion control for the target queue corresponding to the first queue in the sending end device; wherein the preset opening index includes: The occupancy of the buffer space of the receiving end device by the multiple queues of the receiving end device reaches or exceeds the opening back pressure waterline value of the first queue.

In the technical solution provided by the embodiment of the present application, the receiving end device notifies the sending end device to enable congestion control for the target queue corresponding to the first queue in the sending end device when the first queue meets the preset opening index. The congestion control solution has It helps to solve the congestion problem of the current communication network, and helps to improve the resource utilization rate and transmission quality of the communication network. In addition, the congestion control scheme has high flexibility and can ensure the throughput of the communication network.

Optionally, the preset opening indicator further includes one or more of the following:

The first queue belongs to the N longest queues of the first priority among the multiple queues of the receiving end device, the priority of the first queue is the first priority, and when the first priority is the multiple queues When the lowest priority among the priorities of, N is an integer greater than 0, and when the first priority is the non-lowest priority among the priorities of the multiple queues, N is an integer greater than or equal to 0;

The length of the first queue reaches or exceeds the preset length threshold.

Wherein, the preset length threshold may be the average value of the lengths of all queues of the receiving end device, and the N longest queues of the first priority are the topN longest queues among all the queues of the first priority. The lengths of the N longest queues may be equal or unequal. The meaning of the N longest queues of the first priority may be: if all the queues of the first priority are sorted according to the order of length from largest to smallest , The N longest queues of the first priority can be the first N queues among all the queues of the first priority; if all the queues of the first priority are sorted according to the order of length from smallest to largest , The N longest queues of the first priority may be the last N queues among all the queues of the first priority.

In the technical solution provided by the embodiment of the present application, when the preset opening index includes: the buffer space occupied by the multiple queues of the receiving end device on the receiving end device reaches or exceeds the open backpressure waterline value of the first queue, the first queue When a queue belongs to the N longest queues of the first priority among the multiple queues of the receiving end device, and the length of the first queue reaches or exceeds the preset length threshold, the longest queue can be guaranteed to be back pressured, thereby ensuring Reliability of congestion control.

In the embodiment of the present application, it is possible to determine whether to perform back pressure control according to the priority level and determine the number of queues that need back pressure according to the priority level. For the queue with the highest priority, back pressure may not be performed. Back pressure multiple queues with different priorities, the number of queues with multiple priorities that need back pressure can be reduced from low to high according to the multiple priorities, that is, the lower the priority, the lower the priority. The greater the number of queues, the higher the priority and the higher the number of back-pressure queues. For example, the receiving end device has 8 priority queues of pri_1 to pri_8 (that is, pri_1, pri_2, pri_3, pri_4, pri_5, pri_6, pri_7, and pri_8), and the 8 priorities of pri_1 to pri_8 are sequentially increased. For the queue with priority pri_8, the back pressure waterline value may not be set, so that the pri_8 queue may not be back pressured (of course, the back pressure waterline value can also be set for the queue with priority pri_8 for back pressure) . For another example, suppose that two queues with different priorities, pri_2 and pri_3, need to be back-pressured at the same time, and pri_3 is higher than pri_2, then in the queue that needs back-pressure, the number of queues with priority pri_2 can be greater than the queue with priority pri_3 quantity. For example, the number of queues with priority pri_2 may be g, and the number of queues with priority pri_3 may be 0.8×g. In some embodiments, among the queues that require back pressure, the lowest priority queue includes 4 longest queues, the medium priority queue includes 4 longest queues, and the highest priority queue may include 2 or 0 longest queues. Long queue. In this scenario, congestion control (that is, back pressure) can be turned on first for the longest queue with a low priority to ensure the delay and throughput of the queue with a higher priority.

Optionally, notifying the sending end device to enable congestion control for the target queue corresponding to the first queue in the sending end device includes: sending an opening control message to the sending end device, the opening control message instructing the sending end device to open the target queue Congestion control. Wherein, the opening control message may carry the identifier of the target queue and an enable flag bit, the identifier of the target queue indicates the target queue, and the enable flag bit instructs the sending end device to enable congestion control, so that the opening control message instructs the sending end device Turn on congestion control for the target queue. Optionally, the receiving end device may send the activation control message to the sending end device in a broadcast form.

Optionally, after notifying the sending end device to enable congestion control on the target queue corresponding to the first queue in the sending end device, the method further includes: when the first queue meets a preset closing index, notifying the sending end device The congestion control is turned off for the target queue of the sending end device; wherein the preset closing index includes: the multiple queues of the receiving end device occupy less than the closing back pressure of the first queue for the buffer space of the receiving end device Watermark value.

In the technical solution provided by the embodiment of the present application, the receiving end device notifies the sending end device to close the congestion control for the target queue corresponding to the first queue in the sending end device when the first queue meets the preset closing index, which can improve the congestion control performance. Flexibility to ensure the throughput of the communication network.

Optionally, the preset closing indicator further includes one or more of the following:

The multiple queues of the receiving end device occupy the buffer space of the receiving end device less than the closing backpressure waterline value of the first queue for a duration that reaches or exceeds a preset duration threshold;

The first queue does not belong to the N longest queues of the first priority among the multiple queues of the receiving end device, the priority of the first queue is the first priority, and when the first priority is the multiple When the priority of the queue is the lowest priority, N is an integer greater than 0, and when the first priority is the non-lowest priority among the priorities of the multiple queues, N is an integer greater than or equal to 0;

The length of the first queue is less than the preset length threshold.

Optionally, after notifying the sending end device to enable congestion control for the target queue corresponding to the first queue in the sending end device, the method further includes: allocating a sending credit value to the target queue; Sending a message with a credit value, so that the sending end device sends the message in the target queue to the receiving end device according to the sending credit value.

According to the technical solution provided by the embodiments of the present application, the sending end device allocates sending credit values for the target queue according to the receiving end device, and sending the packets in the target queue, which can actively control congestion in the sending end device.

Optionally, the method further includes: adjusting the distribution coefficient of the sending credit value of the target queue according to the occupancy of the multiple queues of the receiving-end device on the buffer space of the receiving-end device, and the distribution coefficient is different from the distribution coefficient of the receiving-end device. The multiple queues are negatively correlated with the occupancy of the buffer space of the receiving device.

According to the technical solution provided by the embodiment of the present application, the receiving end device adjusts the distribution coefficient of the sending credit value of the target queue according to the amount of the buffer space occupied by the receiving end device by the multiple queues of the receiving end device. In this way, the receiving end device The target queue can be assigned a sending credit value based on the adjusted distribution coefficient. Each time the receiving end device allocates a sending credit value to the target queue, the target queue is assigned a sending credit value based on the distribution coefficient of the sending credit value of the target queue at the current moment. , Can improve the flexibility of sending credit allocation, so that the sending end device can flexibly control congestion according to the occupied amount of the buffer space of the receiving end device.

Optionally, each queue of the multiple queues of the receiving end device has a back pressure waterline value, and the back pressure waterline values of queues of different priorities are different, and the method further includes: when multiple queues of the receiving end device When the occupancy of the buffer space of the receiving end device by the queue reaches or exceeds the minimum watermark threshold of the receiving end device, the back pressure waterline value of the queue of each priority of the receiving end device is adjusted.

In the technical solution provided by the embodiments of the present application, the receiving end device adjusts the backpressure waterline value of each priority queue, so that the adjusted backpressure waterline value can be used to determine whether the receiving end device is congested, that is, Each time the receiving end device judges whether it is congested, it uses the current back pressure waterline value to make the judgment, which helps to improve the flexibility of congestion control.

Optionally, the backpressure waterline value includes an open backpressure waterline value and a backpressure waterline value off, and the adjustment of the backpressure waterline value of each priority queue of the receiving end device includes: for any priority Queue: According to the maximum watermark threshold of the receiving end device, the minimum watermark threshold of the receiving end device, the occupancy of the buffer space of the receiving end device by all priority queues lower than any priority, and the receiving end device Adjust the open back pressure waterline value of the queue of any priority for the occupancy of all queues of the end device to the buffer space of the receiving end device; adjust according to the open back pressure waterline value of the queue of any priority The closed backpressure waterline value of the queue of any priority.

Optionally, the occupancy of the buffer space of the receiving end device by the queue of any priority is negatively correlated with the occupancy of the buffer space of the receiving end device by all priority queues lower than the any priority. The opening backpressure waterline value of a priority queue is negatively related to the amount of buffer space occupied by the receiving end device by the queue of any priority; the opening backpressure waterline value of the queue of any priority is greater than or It is equal to the open back pressure waterline value of all priority queues lower than any one of the priority levels. That is, if queues of all priority levels lower than any priority occupies a large amount of buffer space of the receiving end device, the queue of any priority level occupies a relatively large amount of buffer space of the receiving end device. Smaller, the queues of any priority are less likely to be back pressured, the higher the back pressure waterline value of the queue of any priority is, the higher the priority queue, the higher the back pressure waterline value is, the priority is The lower the queue, the smaller the opening back pressure waterline value. In some embodiments, the number of longest queues with high priority is less than the number of longest queues with low priority. In other embodiments, the number of longest queues with high priority may be zero.

The technical solution provided by the embodiment of the present application is due to the occupancy of the buffer space of the receiving end device by a queue of any priority and the occupancy of the buffer space of the receiving end device by all priority queues lower than the any priority. The opening back pressure waterline value of the queue of any priority is negatively correlated with the amount of buffer space occupied by the receiving end device by the queue of any priority. The opening of the queue of any priority is negatively correlated. The pressure waterline value is greater than or equal to the open backpressure waterline value of all priority queues lower than the any priority. Therefore, the open backpressure waterline value of the queue of any priority is the same as the one that is lower than any one of the priority queues. All priority queues of priority are positively correlated to the amount of buffer space occupied by the receiving device, and the open backpressure waterline value of any priority queue is related to all priority queues lower than any priority. The open back pressure waterline value of is positively correlated. In this way, it can prevent the high priority queue from being controlled too early to affect the performance of the high priority queue, and it can also prevent the high priority queue from being controlled too late.

Optionally, according to the maximum watermark threshold of the receiving end device, the minimum watermark threshold of the receiving end device, the occupancy of the buffer space of the receiving end device by all priority queues lower than any one of the priorities, and Adjusting the open back pressure waterline value of the queue of any priority for the occupancy of all queues of the receiving end device to the buffer space of the receiving end device, including: according to the maximum waterline threshold and the minimum waterline threshold, The occupancy of the buffer space of the receiving-end device by all queues with a priority lower than any priority, and the occupancy of the buffer space of the receiving-end device by all queues of the receiving-end device, using the watermark formula Calculate the opening backpressure waterline value of the queue of any priority; adjust the opening backpressure waterline value of the queue of any priority according to the calculation result; where the waterline value formula is:

Vo_pri_k represents the open back pressure waterline value of the queue with priority pri_k, B_pri_i represents the amount of buffer space occupied by all queues of the priority pri_i on the receiving end device, and B_total represents the buffer of all queues of the receiving end device on the receiving end device The amount of space occupied, V_max represents the maximum watermark threshold, and V_min represents the minimum watermark threshold.

Optionally, the backpressure waterline value includes an open backpressure waterline value and a backpressure waterline value off, and the adjustment of the backpressure waterline value of each priority queue of the receiving end device includes: for any priority Queue: According to the occupancy of the buffer space of the receiving end device by all queues of any priority and the size of the buffer space, adjust the opening back pressure waterline value of the queue of any priority. The open backpressure waterline value of the queue of any priority is negatively related to the amount of buffer space occupied by all queues of any priority level on the receiving end device; adjust according to the open backpressure waterline value of the queue of any priority The closed backpressure waterline value of the queue of any priority.

Optionally, the backpressure waterline value includes an open backpressure waterline value and a backpressure waterline value off, and the adjustment of the backpressure waterline value of each priority queue of the receiving end device includes: for any priority Queue: According to the average delay of the M longest queues of any priority and the average delay of all the longest queues of all priorities of the receiving end device, adjust the opening backpressure of the queue of any priority Watermark value, M is an integer greater than 0; according to the open backpressure waterline value of the queue of any priority, the closed backpressure waterline value of the queue of any priority is adjusted.

Optionally, the method further includes: the maximum occupancy of the buffer space of the receiving end device by all queues of the receiving end device within the target duration is greater than the minimum watermark threshold and less than the maximum watermark threshold and the first threshold value Or, when the maximum occupancy is greater than the difference between the maximum watermark threshold and the second threshold, the minimum watermark threshold is adjusted.

In the technical solution provided by the embodiments of the present application, by adjusting the minimum watermark threshold, active congestion control can be avoided or reduced as much as possible, and high throughput and low latency can be ensured.

In a second aspect, a congestion control method is provided. The method includes: enabling congestion control on a target queue of a sending end device according to a notification from a receiving end device, each queue of the receiving end device and one of the sending end device or Corresponding to multiple queues, and the target queue corresponds to the first queue of the receiving end device; wherein the receiving end device notifies the sending end device to enable congestion control for the target queue when the first queue meets a preset opening index, and The preset opening index includes: the amount of the buffer space of the receiving end device occupied by the multiple queues of the receiving end device reaches or exceeds the open back pressure waterline value of the first queue.

The length of the first queue reaches or exceeds the preset length threshold.

Optionally, enabling congestion control on the target queue of the sending end device according to the notification of the receiving end device includes: receiving an open control message sent by the receiving end device, the open control message instructing the sending end device to open congestion control on the target queue; The enable control message enables congestion control on the target queue. Wherein, the opening control message may carry the identifier of the target queue and an enable flag bit, the identifier of the target queue indicates the target queue, and the enable flag bit instructs the sending end device to enable congestion control, so that the opening control message instructs the sending end device Turn on congestion control for the target queue. The sending end device can parse the enable control message to obtain the enable flag and the target queue identifier, determine the need to enable congestion control according to the enable flag bit, determine the target queue according to the target queue identifier, and then determine the target queue according to the enable control message. The queue turns on congestion control.

Optionally, after enabling congestion control on the target queue of the sending end device according to the notification of the receiving end device, the method further includes: turning off congestion control on the target queue of the sending end device according to the notification of the receiving end device; wherein, the receiving end device When the first queue satisfies the preset closing index, the sending end device is notified to close the congestion control of the target queue. The preset closing index includes: the multiple queues of the receiving end device occupy less than the first queue buffer space of the receiving end device. The closing back pressure waterline value of the queue.

The length of time that the multiple queues of the receiving end device occupy the buffer space of the receiving end device is less than the closing back pressure waterline value of the first queue reaches or exceeds the preset duration threshold;

The length of the first queue is less than the preset length threshold.

Optionally, after the congestion control is enabled on the target queue of the sending end device according to the notification of the receiving end device, the method further includes: receiving a message containing a sending credit value sent by the receiving end device; and sending a message to the receiving end device according to the sending credit value. The end device sends the packets in the target queue.

In the technical solution provided by the embodiment of the present application, the sending end device sends the packets in the target queue according to the sending credit value allocated to the target queue by the receiving end device, and the congestion can be actively controlled in the sending end device.

In a third aspect, a congestion control device is provided. The congestion control device includes one or more modules, and the one or more modules are used to implement the congestion control provided in the first aspect or any optional manner of the first aspect. method.

In a fourth aspect, a congestion control device is provided. The congestion control device includes one or more modules, and the one or more modules are used to implement the congestion control provided by the second aspect or any optional manner of the second aspect. method.

In a fifth aspect, a congestion control device is provided. The congestion control device includes a processor and a memory, the memory storing a program, and the processor is configured to execute the program stored in the memory to implement the above-mentioned first aspect or the first aspect. Congestion control method provided by any optional method.

In a sixth aspect, a congestion control device is provided. The congestion control device includes a processor and a memory. The memory is stored with a program. The processor is configured to execute the program stored in the memory to implement the above-mentioned second aspect or the second aspect. Congestion control method provided by any optional method.

In a seventh aspect, a computer-readable storage medium is provided, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the above-mentioned first aspect or any of the optional methods of the first aspect. Or the congestion control method provided in the foregoing second aspect or any optional manner of the second aspect.

In an eighth aspect, a computer program product containing instructions is provided, when the computer program product runs on a computer, it causes the computer to execute the congestion control method provided in the first aspect or any one of the optional methods of the first aspect, or , To implement the congestion control method provided by the foregoing second aspect or any optional manner of the second aspect.

In a ninth aspect, a chip is provided. The chip includes a programmable logic circuit and/or program instructions. When the chip is running, it is used to implement the congestion control method provided in the first aspect or any of the optional methods of the first aspect. , Or, implement the congestion control method provided in the second aspect or any optional manner of the second aspect.

In a tenth aspect, a congestion control system is provided, including: a sending end device and a receiving end device,

The receiving end device is used to monitor multiple queues of the receiving end device, and each queue of the receiving end device corresponds to one or more queues in the sending end device; when the multiple queues of the receiving end device are When the first queue satisfies the preset opening index, the sending end device is notified to enable congestion control for the target queue corresponding to the first queue in the sending end device; wherein, the preset opening index includes: multiple numbers of the receiving end device The occupation of the buffer space of the receiving end device by the queue reaches or exceeds the opening back pressure waterline value of the first queue;

The sending end device is configured to enable congestion control for the target queue corresponding to the first queue in the sending end device according to the notification of the receiving end device.

Optionally, in a possible implementation manner, the sending end device may include the congestion control apparatus provided in the foregoing fourth aspect, and the receiving end device may include the congestion control apparatus provided in the foregoing third aspect; or,

In another possible implementation manner, the sending end device may include the congestion control apparatus provided in the sixth aspect, and the receiving end device may include the congestion control apparatus provided in the fifth aspect.

For the technical effects of the third aspect to the tenth aspect described above, reference may be made to the first aspect to the second aspect, which will not be repeated here.

The beneficial effects brought about by the technical solutions provided by the embodiments of the present application are:

In the technical solution provided by the embodiments of the present application, the receiving end device monitors multiple queues of the receiving end device, and when the first queue of the multiple queues of the receiving end device meets the preset opening index, the sending end device is notified to the sending end The target queue corresponding to the first queue in the device turns on congestion control, and the sending end device turns on congestion control on the target queue according to the notification of the receiving end device. The congestion control solution helps to solve the current congestion problem of the communication network, and helps to improve the resource utilization rate and transmission quality of the communication network. The congestion control solution combines active congestion control and passive congestion control. After using passive congestion control to sense congestion, through active control of congestion, congestion control can be carried out from the entire communication network, and end-to-end congestion can be realized for specific priority congestion. Control, in the process of congestion control, you can flexibly adjust parameters such as the back pressure waterline and the distribution coefficient of the sending credit value, and can make full use of the buffer space of the receiving end device to avoid back pressure of the instantaneous congested message on the sending end device , To ensure the throughput of the sending end device, thereby ensuring the throughput of the workstation connected to the sending end device, and improving the flexibility of congestion control. In this congestion control scheme, before the sender device starts congestion control, there is no need to consider the backpressure waterline value and the sending credit value, so that the traffic of the sender device can be pushed as much as possible, and all queues are kept as much as possible before the congestion control is enabled Huff.

Description of the drawings

FIG. 1 is a schematic diagram of an implementation environment involved in an embodiment of the present application;

FIG. 2 is a schematic diagram of a processing flow of a sending end device in a congestion control solution provided by an embodiment of the present application;

FIG. 3 is a schematic diagram of a processing flow of a receiving end device in a congestion control solution provided by an embodiment of the present application;

FIG. 4 is a schematic diagram of a congestion control solution provided by an embodiment of the present application;

Fig. 5 is a method flowchart of a congestion control method provided by an embodiment of the present application;

FIG. 6 is a schematic diagram of congestion control provided by an embodiment of the present application;

FIG. 7 is a schematic diagram of the amount of buffer space occupied by queues with different priorities according to an embodiment of the present application; FIG.

FIG. 8 is a flowchart of a method for adjusting the back pressure waterline value of a queue of any priority level according to an embodiment of the present application;

FIG. 9 is a flowchart of another method for adjusting the back pressure waterline value of a queue of any priority level provided by an embodiment of the present application;

FIG. 10 is a flowchart of yet another method for adjusting the back pressure waterline value of a queue of any priority according to an embodiment of the present application;

FIG. 11 is a schematic diagram of another congestion control provided by an embodiment of the present application;

FIG. 12 is a schematic diagram of the logical structure of a congestion control device provided by an embodiment of the present application;

FIG. 13 is a schematic diagram of the logical structure of another congestion control apparatus provided by an embodiment of the present application;

FIG. 14 is a schematic diagram of the logical structure of yet another congestion control apparatus provided by an embodiment of the present application;

15 is a schematic diagram of the logical structure of yet another congestion control device provided by an embodiment of the present application;

16 is a schematic diagram of the hardware structure of a congestion control device provided by an embodiment of the present application;

Fig. 17 is a schematic diagram of a congestion control system provided by an embodiment of the present application.

Detailed ways

The implementation environment of the congestion control solution provided in the embodiments of the present application may include a communication network and multiple workstations accessing the communication network. Wherein, the communication network may include multiple network devices, the workstations access the communication network through the network devices, and the network devices implement communication between different workstations by forwarding messages of the workstations.

Among them, the workstation can be various types of devices such as a host, a terminal, a server, or a virtual machine. The network device can be a switch, router, virtual switch or virtual router that is used to forward messages in a communication network, and the network devices in the same communication network can be the same network device or different Internet equipment. For example, all network devices in the same communication network are routers, or some of the network devices are routers, and the other part of the network devices are switches.

In the embodiment of the present application, the communication network may be a data center network (DCN), a metropolitan area network, a wide area network, or a campus network, etc. The embodiment of the present application does not limit the type of the communication network. Optionally, the communication network may be a deep cross network (DCN).

For example, please refer to FIG. 1, which shows a schematic diagram of an implementation environment involved in an embodiment of the present application. The implementation environment includes a communication network 01 and multiple workstations 021 to 028 (that is, Is workstation 021, workstation 022, workstation 023, workstation 024, workstation 025, workstation 026, workstation 027, and workstation 028), the communication network 01 includes network devices 01a1 to 01a2 (that is, network devices 01a1 and network devices 01a2) and network Devices 01b1 to 01b4 (that is, network device 01b1, network device 01b2, network device 01b3, and network device 01b4), and the communication network 01 may include an aggregation layer and an access layer, and the aggregation layer is the high-speed switching backbone of the communication network 01 The access layer is used to connect workstations to the communication network 01. As shown in Figure 1, network devices 01a1 to 01a2 are located at the convergence layer, network devices 01b1 to 01b4 are located at the access layer, and workstations 021 to 022 are accessed through network device 01b1 The communication network 01, workstations 023-024 access the communication network 01 through the network device 01b2, the workstations 025-026 access the communication network 01 through the network device 01b3, and the workstations 027-028 access the communication network 01 through the network device 01b4.

Optionally, the communication network 01 shown in FIG. 1 may be a leaf-spine topology network, and the network devices 01a1 to 01a2 at the convergence layer may be spine switches, and network devices at the access layer 01b1 to 01b4 can be leaf switches, each spine switch is connected to all leaf switches, and each leaf switch is connected to all spine switches (that is, the spine switch and the leaf switch are fully interconnected).

The implementation environment shown in Figure 1 is only used as an example, and is not used to limit the technical solutions of the embodiments of the present application. During the implementation process, the implementation environment may also include other devices. The number of network devices and the number of workstations can be configured as required. The number and the connection relationship between the network devices, the connection relationship between the network devices and the workstation, and the topology of the communication network can be other topologies. For example, the spine switch and the leaf switch may not be fully interconnected. For another example, the network devices in the convergence layer can be interconnected. For example, the communication network may also include a core layer located above the convergence layer. For example, as shown in Figure 1 The illustrated communication network 01 is a level 2 Clos network, and the communication network may also be a level 3 Clos network, which will not be repeated in this embodiment of the present application.

In the embodiment of the present application, each network device has one or more outgoing ports, each outgoing port has one or more queues with different priorities, and one queue of each outgoing port is used for buffering needs to be sent via the outgoing port. For each received packet, the network device usually buffers the packet in a queue that matches the packet, and sends the packets in the queue at a certain rate. Optionally, the number of queues for each egress port is equal to the number of priorities. For example, taking the network device 01b1 in FIG. 1 as an example, the network device 01b1 may have egress port 1 to egress port 8 (that is, egress port 1, egress port 2, egress port 3, egress port 4, and egress port 5. , Outgoing port 6, outgoing port 7 and outgoing port 8) these 8 outgoing ports, outgoing port 1 can have 8 queues with 8 priorities from pri_1 to pri_8. The 8 queues are queues 1 to 8 (that is, Queue 1, Queue 2, Queue 3, Queue 4, Queue 5, Queue 6, Queue 7 and Queue 8), the priority of the queues 1 to 8 are pri_1, pri_2, pri_3, pri_4, pri_5, pri_6, pri_7 and pri_8, among the 8 queues, queue 1 is used to buffer the packets with priority pri_1 that need to be sent via the egress port 1. The network device 01b1 responds to the received packets with priority pri_1 that need to be sent via the egress port 1. Text 1, buffer the message 1 to the queue 1. Queue 2 is used to buffer the messages with priority pri_2 that need to be sent via the egress port 1, and the network device 01b1 needs to send the received messages via the egress port 1. The message 2 with priority pri_2 is buffered to the queue 2. Queue 3 is used to buffer the message with priority pri_3 that needs to be sent through the egress port 1. The network device 01b1 needs to pass the received message through The message 3 with the priority of pri_3 sent by the egress port 1 is buffered in the queue 3, and so on.

Optionally, according to the sending direction of the message, each network device may be a sending end device or a receiving end device. For example, taking Figure 1 as an example, when the network device 01b1 sends a message to the network device 01b4, the network device 01b1 is the sender device, and the network device 01b4 is the receiver device; for another example, when the network device 01b4 sends a message to the network device 01b1 For example, when the network device 01b4 is the sender device, and the network device 01b1 is the receiver device; for example, when the network device 01b1 sends a message to the network device 01b4, and the network device 01b4 sends a message to the network device 01b2, the network device 01b1 It is the sending end device, and the network device 01b2 is the receiving end device, and the network device 01b4 is both the sending end device and the receiving end device; for example, when the network device 01b1 and the network device 01b2 both send a message to the network device 01b4, the network device 01b1 Both the network device 01b2 and the network device 01b2 are the sending end device, and the network device 01b4 is the receiving end device.

In the embodiment of the present application, each queue of the sending end device can be mapped to an outgoing port of the receiving end device, each queue of the sending end device can correspond to the queue of the receiving end device, and each queue of the sending end device When the message arrives at the receiving end device, it is first buffered by the receiving end device into the corresponding queue in the receiving end device, and the receiving end device sends the messages in the queue at a certain rate through the corresponding out port. For example, the sending end device is the network device 01b1, the receiving end device is the network device 01b4, the queue 1 of the network device 01b1 can be mapped to an out port 1 of the network device 01b4, and the queue 1 of the network device 01b1 is the same as that of the network device 01b4. Queue 1 corresponds. When the message 1 in the queue 1 of the network device 01b1 arrives at the network device 01b4, it is first buffered by the network device 01b4 in the queue 1 of the network device 01b4, and the network device 01b4 passes through the outgoing port of the network device 01b4 1 Send the message at a certain rate.

With the expansion of the scale of communication networks and the popularization of various applications such as video, high-performance cloud computing and consumer cloud services, the traffic that network devices need to forward has skyrocketed, the network load rate (that is, the network utilization rate), network latency, and Network congestion has become a major problem faced by communication networks. For example, communication networks are faced with probabilistic and long-term packet loss problems, as well as the problems of burst traffic discarding caused by insufficient buffer space of network devices. For another example, remote direct memory access (RDMA) protocol and Ethernet-based remote direct memory access (RDMA overconverged ethernet, RoCE) protocol communication networks are extremely important for high-performance cloud computing. These communication networks currently exist Many problems such as network delay and network congestion. Among the above-mentioned problems, network congestion is the main problem of current communication networks. Network congestion not only requires bandwidth expansion of the communication network, but more importantly, reasonable traffic scheduling is required to control network congestion. Therefore, there is an urgent need for a congestion control scheme to control the congestion of the communication network.

The network congestion control scheme provided by the embodiment of the present application can combine the active congestion control scheme and the passive congestion control scheme to control the congestion of the communication network, which not only helps the congestion problem of the communication network, but also has better flexibility. Among them, the active congestion control scheme is an end-to-end probe estimation or request response method to estimate the degree of congestion of the link to estimate the corresponding bandwidth or distribute the credit value (credit) to ensure that the communication network is not congested. Traffic transmission schemes, for example, remote procedure call (RCP) scheme, high precision congestion control (HPCC) scheme, and push and pull (PPOP) scheme in Venus algorithm are all active congestion control Program. The passive congestion control scheme is a scheme that passively senses and controls network congestion at the receiving end device. For example, the explicit congestion notification (ECN) scheme, the policy feature card scheme, and the discard backpressure scheme Long queue (drop and backpressure the longest queue, DBLQ) solutions are all passive congestion control solutions. The following uses the PPOP solution and the DBLQ solution as examples to introduce the active congestion control solution and the passive congestion control solution provided by the embodiments of the present application.

First, the active congestion control solution provided by the embodiment of the present application is introduced.

In the active congestion control scheme, the queue of the sending device can be a virtual output queue (VOQ), for example, it can be a dynamic virtual output queue (DVOQ), and the queue of the receiving device can be a request flow Queue (request flow queue, RFQ). Figures 2 and 3 are schematic diagrams of active congestion control solutions provided by embodiments of the present application, in which Figure 2 is a schematic diagram of the processing flow of the sending end device, and Figure 3 is a schematic diagram of the processing flow of the receiving end device. 2 and 3 take DVOQ as an example for the queue of the sending end device.

Referring to Fig. 2, for the received message, the sending end device buffers the message to a DVOQ matching the message, and updates the queue information of the DVOQ stored in the DVOQ module. For multiple DVOQs of the sender device, the sender device schedules DVOQs of different priorities according to strict priority (strict priority, SP), and schedules DVOQs of the same priority according to round-robin (round-robin, RR). For the scheduled DVOQ, the sending end device sends a request packet (request packet, REQ) carrying the queue information of the DVOQ to the receiving end device to request the receiving end device to allocate a sending credit value for the DVOQ.

Among them, the sender device can query the queue mapping table (such as DVOQ mapping table) to determine the DVOQ that matches the message. If the DVOQ that matches the message cannot be determined according to the queue mapping table, the sender device can apply for DVOQ resources and create a new DVOQ. , The newly created DVOQ is determined as the DVOQ matching the message. If the DVOQ resource cannot be applied for, the sending end device determines the highest priority DVOQ of the sending end device as the DVOQ matching the message. The queue information of the DVOQ may include one or more of the priority of the DVOQ, the length of the DVOQ, and the identifier of the DVOQ. The length of the DVOQ may be a virtual queue length (VQL). The sending credit value can be the number of packets in the target queue that the receiving device can schedule this time, or the number of packets in the target queue that the sending device can send this time. In some implementation scenarios, the sending credit value is again It is called queue credit (QC).

As shown in Figure 2, for the multiple DVOQs of the sending end device, the SP1 scheduler preferentially schedules the multiple DVOQs with the highest priority, and assigns the bandwidth of the outgoing port of the sending end device to the highest priority DVOQ first. RR calls the multiple DVOQs with the highest priority; the SP1 scheduler allocates the remaining bandwidth after deducting the bandwidth occupied by the highest priority DVOQ to the SP2 scheduler, and the SP2 scheduler prioritizes the scheduling of multiple DVOQs with high priority. The remaining bandwidth is preferentially allocated to the high-priority DVOQ, and multiple DVOQs of the high-priority are called according to the RR; the SP2 scheduler allocates the remaining bandwidth after deducting the bandwidth occupied by the high-priority DVOQ from the remaining bandwidth to the low-priority DVOQ. Multiple DVOQs of priority are used, and multiple DVOQs of low priority are called according to RR.

Referring to FIG. 3, for the received REQ, the receiving end device buffers the REQ to an RFQ matching the REQ, and updates the queue information of the RFQ stored in the RFQ module. For multiple RFQs of the receiving end device, the receiving end device schedules RFQs of different priorities according to SP, and RFQs of the same priority according to RR. For any REQ in the scheduled RFQ, the receiving end device allocates a sending credit value for the DVOQ according to the DVOQ queue information carried by the REQ, and sends a reply packet (acknowledge packet, which carries the sending credit value) to the sending end device. ACK), so that the sending end device sends the message in the DVOQ according to the sending credit value.

As shown in Figure 3, for multiple RFQs of the receiving end device, the SP1 scheduler preferentially schedules the multiple RFQs with the highest priority, and assigns the bandwidth of the output port of the receiving end device to the highest priority RFQ first, and Call the multiple RFQs with the highest priority according to the RR; the SP1 scheduler allocates the remaining bandwidth after deducting the bandwidth occupied by the RFQ with the highest priority to the SP2 scheduler, and the SP2 scheduler preferentially schedules multiple RFQs with high priority. The remaining bandwidth is allocated to the high-priority RFQ first, and multiple high-priority RFQs are called according to the RR; the SP2 scheduler allocates the remaining bandwidth after deducting the bandwidth occupied by the high-priority RFQ from the remaining bandwidth It is used for multiple RFQs with low priority, and calls the multiple RFQs with low priority according to RR.

It should be noted that in the active congestion control scheme, each network device is configured with a certain amount of initial credit value. After the communication network is started, the sending end device first pushes to the receiving end device according to the initial credit value of the sending end device. After a certain flow rate, REQ is sent to the receiving end device according to the queue information of the DVOQ to request the receiving end device to allocate a sending credit value for the receiving end device.

The active congestion control scheme is an end-to-end congestion control scheme. The active congestion control scheme tries to avoid congestion and can control traffic bursts from the sending end device to the receiving end device. In this active congestion control scheme, the sending end The device sends REQ to the receiving end device, and the receiving end device replies to the sending end device with an ACK carrying the sending credit value according to the scheduling capability, so that the sending end device performs controlled packet sending according to the sending credit value, thereby realizing congestion control.

The above is an introduction to the active congestion control solution provided by the embodiments of this application. After introducing the active congestion control solution, the passive congestion control solution provided in the embodiments of this application is then introduced.

Please refer to FIG. 4, which shows a schematic diagram of a passive congestion control solution provided by an embodiment of the present application. In the passive congestion control scheme, for each message (such as REQ) received by the receiving end device, the message is buffered in the receiving end device’s queue that matches the message, according to all queues of the receiving end device The occupancy of the buffer space of the receiving end device determines whether the receiving end device is congested. When the receiving end device is congested, back pressure is performed on the longest queue among the priority queues in the order of priority from low to high. (That is, the messages in the longest queue in the queues of each priority are discarded in the order of priority from low to high), so as to perform passive congestion control on the communication network. For example, the receiving device preferentially discards the packets in the longest queue among the multiple queues with the lowest priority. If the packets in the longest queue with the lowest priority are discarded, the congestion of the receiving device is not effectively controlled. Discard packets in the longest queue among multiple queues with the second lowest priority, and so on. In the embodiment of the present application, the buffer space of the receiving end device may be a buffer. According to the occupied amount of the buffer space, the buffer space may include a safe area and a prohibited area, and the receiving end device may be based on the amount of the buffer space. The occupied area is used to determine whether the receiving end device is congested. If the occupied amount of the buffer space is in the safe zone, the receiving end device determines that the receiving end device is not congested, and if the occupied amount of the buffer space is in the forbidden zone , The receiving end device determines that the receiving end device is congested. Optionally, the safe area may be an area where the occupied amount of the cache space is less than or equal to 90% of the size of the cache space, and the prohibited area may be an area where the occupied amount of the cache space is greater than 90% of the size of the cache space.

The passive congestion control scheme allows the receiving device to perceive network congestion. When the amount of buffer space occupied is high (that is, the network congestion is sensed), the highest priority in the queue of each priority is discarded in the order of priority from low to high. The messages in the long queue can maximize the use of buffer space resources to reduce the waiting delay of traffic, and the utilization of buffer space is high, which can reduce the packet loss rate caused by congestion and ensure the throughput of the communication network.

The above is an introduction to the active congestion control scheme and the passive congestion control scheme provided by the embodiments of this application. After introducing the active congestion control and passive congestion control schemes, the congestion control is performed on the combination of active congestion control and passive congestion control schemes provided by the embodiments of this application. The program is introduced. The embodiments of the present application combine the active congestion control scheme and the passive congestion control scheme to perform congestion control on the communication network, which can not only solve the congestion problem of the communication network, but also has better flexibility.

Please refer to FIG. 5, which shows a method flow chart of a congestion control method provided by an embodiment of the present application. The congestion control method may be applied to a communication network. The communication network may include multiple network devices. The multiple network devices including the sending end device and the receiving end device are taken as an example for description. Referring to Figure 5, the method may include the following steps:

Step 501: The receiving end device sets a back pressure waterline value for the queue of the receiving end device, and the back pressure waterline value includes the opening back pressure waterline value and the closing back pressure waterline value. Wherein, each queue of the receiving end device corresponds to one or more queues in the sending end device.

In the embodiment of the present application, each network device in the communication network may have one or more outgoing ports, and each outgoing port may have one or more priority queues. If each outgoing port has multiple priority queues, For queues, the multiple priorities may be different, and each queue of each network device corresponds to one or more queues of one or more other network devices in the communication network. Each network device can set a back pressure waterline value for the queue in the network device, so as to be used for back pressure control when the network device is congested. In each network device, each queue can have a priority, and the network device can set the back pressure waterline value for each priority queue of the network device, or the network device is not the highest priority of the network device The backpressure waterline value is set for the queue of, and the backpressure waterline value is set for queues of other priority levels except the highest priority, which is not limited in this embodiment of the application. In the embodiment of this application, the backpressure waterline values of queues of different priority levels can be different, the backpressure waterline value of all queues of the same priority level can be the same, and the backpressure waterline value of each priority queue can be called Is the back pressure waterline value corresponding to this priority. Wherein, the back pressure waterline value may include the open back pressure waterline value and the off back pressure waterline value. The open back pressure waterline value of each priority queue can be referred to as the open back pressure waterline corresponding to the priority. Value, the closing backpressure waterline value of each priority queue can be called the closing backpressure waterline value corresponding to the priority.

Optionally, in each network device, all queues with the same priority have the same backpressure waterline value, queues with different priorities have different backpressure waterline values, and queues with different priority levels have different backpressure waterline values. According to the different priorities, increase from low to high, and the closing backpressure waterline values of queues of different priorities increase from low to high according to the different priorities. That is, the lower the priority of the queue, the smaller the open back pressure waterline value and the lower the back pressure waterline value off, and the higher the priority queue, the higher the open back pressure waterline value and the off back pressure waterline value. Big.

The receiving end device in the embodiment of the present application may be any network device in the communication network. Therefore, the receiving end device has one or more outgoing ports, and each outgoing port can have one or more priority queues. The receiving end device may set a back pressure waterline value for the queue of the receiving end device, and the back pressure waterline value includes an open back pressure waterline value and a closed back pressure waterline value. Wherein, each queue of the receiving end device corresponds to one or more queues in the sending end device.

Optionally, in the embodiment of the present application, the receiving end device sets the back pressure waterline value for each priority queue of the receiving end device as an example for description. All queues of the receiving end device can be stored in the buffer space of the receiving end device, and the receiving end device can set the maximum waterline threshold and the minimum waterline threshold for the receiving end device according to the size of the buffer space of the receiving end device. The watermark threshold and the minimum watermark threshold are set to open the backpressure waterline value for each priority queue of the receiving end device, and the backpressure waterline value for each priority queue is set according to each priority. The queue setting closes the back pressure waterline value. Among them, the difference between the open back pressure waterline value and the closed back pressure waterline value of each priority queue can be a fixed value. For example, the fixed value may be 10.

Optionally, the receiving end device setting the maximum watermark threshold and the minimum watermark threshold for the receiving end device according to the size of the buffer space of the receiving end device may include: the receiving end device compares the size of the buffer space of the receiving end device with the first The product of a coefficient is determined as the maximum watermark threshold of the receiving end device, and the product of the size of the buffer space of the receiving end device and the second coefficient is determined as the minimum watermark threshold of the receiving end device, and is the receiving end device The maximum waterline threshold and the minimum waterline threshold are set, wherein the first coefficient is greater than the second coefficient, the first coefficient is greater than 0 and less than 0.85, and the second coefficient is greater than 0 and less than 0.7. For example, the first coefficient may be 0.8, and the second coefficient may be 0.5.

Optionally, setting the receiving end device to turn on the back pressure waterline value according to the maximum watermark threshold and the minimum watermark threshold value for each priority queue of the receiving end device may include: the receiving end device sets the maximum watermark threshold value Set the open back pressure waterline value of the highest priority queue of the receiving end device, and set the minimum watermark threshold to the open back pressure waterline value of the lowest priority queue of the receiving end device. For the highest priority queue For any priority between the priority level and the lowest priority, the receiving end device can open the back pressure waterline value of the queue of the previous priority lower than the any priority, the maximum waterline threshold, the minimum waterline Line threshold and the number of priorities of the receiving end device (that is, the number of priorities of all queues of the receiving end device), set the open back pressure waterline value for the queue of any priority (that is, setting The back pressure waterline value corresponding to any priority). For example, the receiving end device can determine a certain value based on the maximum watermark threshold, the minimum watermark threshold, and the priority of the receiving end device. The value will be lower than The sum of the opening back pressure waterline value of the queue of the previous priority of the any priority and the fixed value is determined as the opening back pressure waterline value of the any priority.

For example, taking the first coefficient of 0.8 and the second coefficient of 0.5 as an example, the receiving end device can use the initial value configuration formula to set the open back pressure waterline value for each priority queue of the receiving end device, where the The initial value configuration formula is:

B_size represents the size of the buffer space of the receiving end device, V_max represents the maximum waterline threshold of the receiving end device, V_min represents the minimum waterline threshold of the receiving end device, and Vo_pri_min represents the opening back pressure waterline of the lowest priority queue of the receiving end device Value, Vo_pri_max represents the open back pressure waterline value of the highest priority queue of the receiving end device, Vo_pri_k represents the open back pressure waterline value of the queue with priority pri_k, Vo_pri_(k-1) represents the priority of pri_(k -1) The opening back pressure waterline value of the queue, that is, the opening back pressure waterline value of the queue of the previous priority lower than the priority pri_k, and Q_pri represents the priority number of the receiving end device.

For example, please refer to FIG. 6, which shows a schematic diagram of congestion control provided by an embodiment of the present application. Referring to FIG. 6, the receiving end device has 8 priority queues from pri_1 to pri_8, and the priority is pri_1. The open back pressure waterline is Vo_pri_1, the closed back pressure water line is Vc_pri_1, the open back pressure water line of the queue with priority pri_2 is Vo_pri_2, the closed back pressure water line is Vc_pri_2, and the open back pressure of the queue with priority pri_3 The water line is Vo_pri_3, the closed back pressure water line is Vc_pri_3, and so on, the open back pressure water line of the queue with priority pri_8 is Vo_pri_8, the closed back pressure water line is Vc_pri_8, and the priority levels pri_1～pri_8 increase in turn, Vo_pri_1～ Vo_pri_8 increases in sequence, and Vc_pri_1～Vc_pri_8 increases in sequence, that is, the lower the priority of the queue, the smaller the value of opening and closing the back pressure waterline, the higher the priority of the queue, the opening of the anti The higher the pressure waterline value and the closing back pressure waterline value.

Step 502: The receiving end device monitors multiple queues of the receiving end device.

Optionally, according to step 501, it can be known that each of the multiple queues of the receiving end device has a priority, and there may be multiple queues belonging to the same priority among the multiple queues of the receiving end device. The receiving-end device can monitor multiple queues of the receiving-end device, determine the length of each of the multiple queues of the receiving-end device, and how much the multiple queues of the receiving-end device occupy the buffer space of the receiving-end device, The occupancy of the buffer space of the receiving end device by all queues of each priority in the multiple queues of the receiving end device, and so on.

Optionally, the receiving end device may record the length of each queue in the multiple queues of the receiving end device, the occupation of the buffer space of the receiving end device by all queues of each priority, and the receiving end device All queues (all queues of all priorities) occupy the buffer space of the receiving end device. Each time the receiving end device receives a message, it can change it according to the destination port number and priority of the message. The message is buffered in the corresponding queue in the receiving device (for example, the queue with the same priority as the message in the queue of the target port), and the length of the corresponding queue is updated according to the size of the message, and the update priority is this The occupancy of the buffer space of the receiving end device by all the queues of the message priority, and the occupancy of the buffer space of the receiving end device by all the queues of the receiving end device.

Optionally, the first queue may be one of multiple queues of the receiving end device, the priority of the first queue may be the first priority, the receiving end device may record the length of the first queue, and the first queue The occupancy of the buffer space of the receiving-end device by all queues of one priority, and the occupancy of the buffer space of the receiving-end device by all the queues of the receiving-end device, the receiving-end device for the first report received The message can be cached in the first queue according to the destination port number and priority of the first message, the length of the first queue is updated according to the size of the first message, and the first priority is updated. The occupancy of the buffer space of the receiving end device by all the queues of the level, and the occupancy of the buffer space of the receiving end device by all the queues of the receiving end device.

For example, please refer to FIG. 7, which shows a schematic diagram of the occupancy of the buffer space of the receiving-end device by queues of different priorities in the receiving-end device provided by an embodiment of the present application. As shown in FIG. 7, B_size Indicates the size of the buffer space of the receiving end device. The receiving end device has 8 priority queues from pri_1 to pri_8. All queues with priority pri_1 occupy the buffer space of the receiving end device B_pri_1, and the priority is pri_2 The occupancy B_pri_2 of the buffer space of the receiving end device by all the queues, the occupancy B_pri_3 of the buffer space of the receiving end device by all queues with priority pri_3, and so on. Assuming that the number of queues for each of the 8 priority levels of pri_1 to pri_8 is 10, the receiving end device can record the length of each queue in the 80 queues of the 8 priority levels, and the 80 queues The buffer space occupied by the receiving end device B_total, the buffer space occupied by the 10 queues with priority pri_1 for the receiving end device B_pri_1, and the buffer space occupied by the 10 queues with priority pri_2 for the receiving end device The occupancy of B_pri_2, the occupancy B_pri_3 of the buffer space of the receiving end device by the 10 queues with priority pri_3, and so on. The first queue can be queue 1 of the 80 queues, and the first priority can be pri_1. For the received message 1, the receiving end device can set the destination port number and priority of the message 1 according to the information such as the destination port number and priority of the message 1. The message 1 is cached in the queue 1, and the length of the queue 1 is updated according to the size of the message 1, the 10 queues with priority pri_1 and the buffer space occupied by the receiving end device B_pri_1 are updated, and the 80 is updated. The occupancy B_total of each queue of the buffer space of the receiving end device.

Step 503: When the first queue of the multiple queues of the receiving end device meets the preset opening index, the receiving end device notifies the sending end device to enable congestion control for the target queue corresponding to the first queue in the sending end device.

In the embodiment of the present application, in the process of monitoring the multiple queues of the receiving end device, the receiving end device can detect whether each of the multiple queues of the receiving end device meets the preset opening index. When the queue meets the preset opening index, the receiving end device notifies the sending end device to enable congestion control for the queue corresponding to each queue in the sending end device.

In the embodiment of the present application, the first queue may be one of the multiple queues of the receiving end device. Therefore, the receiving end device can detect whether the first queue is satisfied during the process of monitoring the multiple queues of the receiving end device. A preset opening index is set. When the first queue meets the preset opening index, the receiving end device notifies the sending end device to enable congestion control for the target queue corresponding to the first queue in the sending end device. Wherein, the preset opening index includes: the amount of the buffer space occupied by the multiple queues of the receiving end device for the receiving end device reaches or exceeds the opening back pressure waterline value of the first queue.

Optionally, the preset opening indicator may further include one or more of the following:

The length of the first queue reaches or exceeds the preset length threshold.

Optionally, the preset opening index includes: the amount of the buffer space occupied by the multiple queues of the receiving end device on the receiving end device reaches or exceeds the open backpressure waterline value of the first queue, and the first queue belongs to the first queue. The N longest queues of the first priority among the multiple queues of the receiving end device, and the length of the first queue reaches or exceeds the preset length threshold. That is, the first queue satisfies at the same time: the amount of the buffer space occupied by the multiple queues of the receiving end device on the receiving end device reaches or exceeds the opening backpressure waterline value of the first queue, and the first queue belongs to the first queue. The N longest queues of the first priority among the multiple queues of the receiving end device, and the length of the first queue reaches or exceeds the preset length threshold, the receiving end device notifies the sending end device of the first priority among the sending end device The target queue corresponding to a queue turns on congestion control.

The preset length threshold may be the average value of the lengths of all queues of the receiving end device. For example, the receiving end device always has 80 queues, and the preset length threshold may be the average value of the lengths of the 80 queues. The N longest queues of the first priority are the topN longest queues among all the queues of the first priority. The length of the N longest queues may be equal or unequal. The N of the first priority The meaning of the longest queues can be: if all the queues of the first priority are sorted according to the order of their length, the N longest queues of the first priority can be all the queues of the first priority The N queues that are sorted at the top of the list; if all the queues of the first priority are sorted in ascending order of length, the N longest queues of the first priority can be all the queues of the first priority The N queues sorted in the bottom. For example, N=5, there are a total of 10 queues of the first priority, and the 10 queues are queue 1, queue 2, queue 3, queue 4, queue 5, queue 6, queue 7, queue 8, queue 9, and Queue 10, the results obtained by sorting the 10 queues in order of length from largest to smallest can be: Queue 1, Queue 2, Queue 5, Queue 4, Queue 3, Queue 6, Queue 7, Queue 8, Queue 9, Queue 10, the N longest queues of the first priority can be the top 5 queues, which are queue 1, queue 2, queue 5, queue 4, and queue 3, in the order of ascending length The result of sorting the 10 queues can be: queue 10, queue 9, queue 8, queue 7, queue 6, queue 3, queue 4, queue 5, queue 2, queue 1, then the N of the first priority The longest queues can be the 5 last queues, which are queue 1, queue 2, queue 5, queue 4, and queue 3.

Optionally, the receiving-end device notifying the sending-end device to enable congestion control for the target queue corresponding to the first queue in the sending-end device may include: the receiving-end device sends an enabling control message to the sending-end device, where the enabling control message indicates The sending end device enables congestion control for the target queue. Optionally, the turn-on control message may carry an identifier of the target queue and an enable flag bit, the identifier of the target queue indicates the target queue, and the enable flag bit instructs the sending end device to turn on congestion control, so that the turn-on control message instructs to send The end device enables congestion control on the target queue. Optionally, the opening control message may also carry the identification of the sending end device, so that the sending end device determines according to the identification of the sending end device that the opening control message is sent to the sending end device (that is, the sending end device). The identifier indicates the recipient of the turn-on control message). Optionally, the receiving end device may send the opening control message to the sending end device in a broadcast form. For example, the receiving end device broadcasts the opening control message to all sending end devices of the receiving end device.

Optionally, in the embodiment of the present application, the receiving end device may determine whether to perform back pressure control according to the priority level, and determine the number of queues that need back pressure according to the priority level. For the queue with the highest priority, it is not necessary to perform back pressure control. For back pressure, if you need to back pressure multiple queues with different priorities at the same time, the number of queues with multiple priorities that need back pressure can be reduced from low to high according to the multiple priorities, that is, priority The lower the level, the greater the number of backpressure queues, and the higher the priority, the higher the number of backpressure queues.

For example, please continue to refer to FIG. 6, the receiving end device has 8 priority queues pri_1 to pri_8, and the 8 priorities of pri_1 to pri_8 are successively increased, and the back pressure on the pri_8 queue may not be performed. As another example, suppose that 8 queues with different priorities, pri_1 to pri_8, need to be back-pressured at the same time. In the queues that need back-pressure, the number of priority queues decreases in order from low to high, for example, pri_3 is high. For pri_2, the number of queues with priority pri_2 can be greater than the number of queues with priority pri_3, the number of queues with priority pri_2 can be g, and the number of queues with priority pri_3 in the queue that needs back pressure It can be 0.8×g, and the g queues with priority pri_2 belong to the N longest queues among all the queues with priority pri_2. In some embodiments, the longest queue of the high-priority queue is less than the longest queue of the low-priority queue. In other embodiments, the longest queue of the high-priority queue may be zero. Optionally, in some embodiments, among the queues that require back pressure, the lowest priority queue includes 4 longest queues, the medium priority queue includes 4 longest queues, and the highest priority queue may include 2 Or 0 longest queues. In this scenario, the longest queue with low priority can be back pressured first to ensure the delay and throughput of the queue with higher priority.

Step 504: The sending end device starts congestion control for the target queue of the sending end device according to the notification of the receiving end device.

Optionally, according to step 503, the receiving end device notifies the sending end device to enable congestion control on the target queue when the first queue of the receiving end device meets the preset opening index, and the target queue corresponds to the first queue of the receiving device . Wherein, the preset opening index includes: the amount of the buffer space occupied by the multiple queues of the receiving end device for the receiving end device reaches or exceeds the open back pressure waterline value of the first queue.

Optionally, the preset opening indicator may also be one or more of the following:

The length of the first queue reaches or exceeds the preset length threshold.

Optionally, the sending-end device enabling congestion control on the target queue of the sending-end device according to the notification of the receiving-end device may include: the sending-end device receives an opening control message sent by the receiving-end device, and according to the opening control message to the target queue Enable congestion control, where the enable control message instructs the sending end device to enable congestion control for the target queue. Optionally, the turn-on control message may carry an identifier of the target queue and an enable flag bit, the identifier of the target queue indicates the target queue, and the enable flag bit instructs the sending end device to turn on congestion control, so that the turn-on control message instructs to send The end device enables congestion control on the target queue. Optionally, the turn-on control message may also carry an identifier of the sending-end device, and the identifier of the sending-end device indicates the recipient of the turn-on control message. The sender device can parse the start control message to obtain the sender device's identity, enable flag bit, and target queue identity. According to the sender device’s identity, it is determined that the switch on control message is sent to the sender device. The capability flag determines that congestion control needs to be turned on, and the target queue is determined according to the identifier of the target queue, so that congestion control is turned on for the target queue according to the turn-on control message.

Step 505: The receiving end device allocates a sending credit value to the target queue.

After the receiving-end device notifies the sending-end device to enable congestion control for the target queue in the sending-end device, it can assign a sending credit value to the target queue, so that the sending-end device sends the target queue to the receiving-end device according to the sending credit value. Message. Among them, the sending credit value may be the number of packets in the target queue that the receiving end device can schedule this time, or the number of packets in the target queue that the sending end device can send this time, and the receiving end device can be based on its own The scheduling capability allocates sending credits to the target queue. In some implementation scenarios, the sending credit value may also be referred to as the queue credit value.

Optionally, after the sending-end device starts congestion control for the target queue in the sending-end device according to the notification of the receiving-end device, it can send to the receiving-end device a message containing the queue information (such as queue length and priority, etc.) of the target queue. Message, the receiving end device allocates a sending credit value to the target queue according to the queue information in the message. Wherein, the message containing the queue information of the target queue may be REQ.

Optionally, the receiving end device may determine a distribution coefficient according to the open backpressure waterline of the first queue, and use the distribution coefficient to allocate the transmission credit value to the target queue according to the queue information in the message. Wherein, when the opening backpressure waterline of the first queue is the initial backpressure waterline value set by the system, after the sending end device starts congestion control on the target queue in the sending end device, the receiving end device schedules the target queue for the first time In the case of a message, the distribution coefficient may be the initial distribution coefficient, and the initial distribution coefficient may be 1, for example.

Step 506: The receiving end device sends a message containing the sending credit value to the sending end device.

Optionally, the receiving end device may generate a message containing the sending credit value according to the sending credit value of the target queue, and sending the message containing the sending credit value to the sending end device through a communication connection with the sending end device. Wherein, the message containing the sending credit value may be ACK.

Step 507: The sending end device sends the packet in the target queue to the receiving end device according to the sending credit value.

Corresponding to the receiving end device sending a message containing the sending credit value of the target queue to the sending end device, the sending end device may receive the message containing the sending credit value of the target queue. After that, the sending end device may send the message in the target queue to the receiving end device according to the sending credit value.

For example, the sending credit value of the target queue may be 5 (for example, 5 packets), and the sending end device may send 5 packets in the target queue to the receiving end device according to the sending credit value.

Step 508: When the occupancy of the buffer space of the receiving end device by the multiple queues of the receiving end device reaches or exceeds the minimum waterline threshold of the receiving end device, the receiving end device adjusts the queues of each priority of the receiving end device. The back pressure waterline value.

In the embodiment of the present application, the receiving end device may have a minimum watermark threshold. In the process of monitoring the multiple queues of the receiving end device, the receiving end device may detect that the multiple queues of the receiving end device have a positive effect on the receiving end device. Whether the occupancy of the buffer space reaches or exceeds the minimum water mark threshold of the receiving end device, when the multiple queues of the receiving end device occupies the receiving end device’s buffer space reach or exceed the minimum water mark threshold of the receiving end device When the line threshold is set, the receiving end device can adjust the back pressure waterline value of each priority queue of the receiving end device to perform congestion perception and judgment according to the adjusted back pressure waterline value, thereby flexibly controlling. Optionally, the receiving end device can dynamically adjust the minimum watermark threshold of the receiving end device, and the receiving end device detects each time whether the multiple queues of the receiving end device occupy or exceed the buffer space of the receiving end device. When the minimum watermark threshold of the receiving end device is used, the minimum watermark threshold used may be the minimum watermark threshold of the receiving end device at the current moment (that is, the minimum watermark threshold after the most recent adjustment).

In the embodiment of the present application, the receiving end device adjusting the backpressure waterline value of each priority queue of the receiving end device may include the following three possible implementation manners:

The first implementation mode: the receiving end device may have a maximum watermark threshold, and the receiving end device is based on the maximum watermark threshold and minimum watermark threshold of the receiving end device, and the buffer space of the receiving end device's queue for the receiving end device Adjust the backpressure waterline value of each priority queue of the receiving end device.

Optionally, please refer to FIG. 8, which shows a flow chart of a method for a receiving end device to adjust the back pressure waterline value of any priority queue of the receiving end device according to an embodiment of the present application. For each For the priority queue, the receiving end device can adjust the back pressure waterline value of the priority queue according to the method shown in FIG. 8. Referring to Figure 8, the method may include:

Sub-step 5081a, according to the maximum watermark threshold of the receiving end device, the minimum watermark threshold of the receiving end device, the occupancy of the buffer space of the receiving end device by all priority queues lower than any priority, and the Regarding the occupancy of all queues of the receiving end device to the buffer space of the receiving end device, the opening back pressure waterline value of the queue of any priority is adjusted.

Wherein, the occupancy of the buffer space of the receiving end device by the queue of any priority is negatively correlated with the occupancy of the buffer space of the receiving end device by all priority queues lower than the any priority. The open back pressure waterline value of the queue of any priority is negatively related to the amount of buffer space occupied by the receiving end device by the queue of any priority; the open back pressure waterline value of the queue of any priority is greater than or equal to low The open backpressure waterline value of all priority queues at any priority level. That is, if queues of all priority levels lower than any priority occupies a large amount of buffer space of the receiving end device, the queue of any priority level occupies a relatively large amount of buffer space of the receiving end device. Smaller, the queues of any priority are less likely to be back pressured, the higher the back pressure waterline value of the queue of any priority is, the higher the priority queue, the higher the back pressure waterline value is, the priority is The lower the queue, the smaller the opening back pressure waterline value. Optionally, the open back pressure waterline value of the queue of any priority is positively correlated with the occupancy of the buffer space of the receiving end device by the queues of all priorities lower than the any priority.

Optionally, in this embodiment of the present application, the receiving end device may respond to the maximum watermark threshold of the receiving end device, the minimum watermark threshold of the receiving end device, and all priority queues lower than any priority. The occupancy of the buffer space of the receiving end device, and the occupancy of all queues of the receiving end device to the buffer space of the receiving end device, the waterline value formula is used to calculate the open backpressure waterline value of the queue of any priority , Adjust the opening back pressure waterline value of the queue of any priority according to the calculation result. Among them, the waterline value formula can be:

In this waterline value formula, Vo_pri_k represents the opening back pressure waterline value of the queue with priority pri_k, B_pri_i represents the occupation of the buffer space of the receiving end device by all queues with priority pri_i, and B_total represents the receiving end device's buffer space All queues occupy the buffer space of the receiving end device, V_max represents the maximum watermark threshold of the receiving end device, and V_min represents the minimum watermark threshold of the receiving end device.

Optionally, the receiving end device can set the maximum watermark threshold of the receiving end device, the minimum watermark threshold of the receiving end device, and all priority queues lower than any priority to the buffer space of the receiving end device. The occupancy and the occupancy of all queues of the receiving end device to the buffer space of the receiving end device are substituted into the waterline value formula to calculate the opening backpressure waterline value of the queue of any priority. For example, there are 8 priority queues of pri_1 to pri_8 in the receiving end device, and the priorities pri_1 to pri_8 increase in sequence. When pri_k=pri_1, there is no priority lower than pri_1 in the receiving end device, so the above formula B_pri_i in does not exist, Vo_pri_1=Vo_pri_min, when pri_k=pri_8, Vo_pri_8=Vo_pri_max.

It should be noted that in this embodiment of the present application, the receiving end device can dynamically adjust the minimum watermark threshold of the receiving end device. In the first implementation manner, the receiving end device adjusts the opening response of any priority queue. When setting the watermark value, the minimum watermark threshold used may be the minimum watermark threshold of the receiving end device at the current moment (that is, the minimum watermark threshold after the most recent adjustment).

Sub-step 5082a: According to the open back pressure waterline value of the queue of any priority, adjust the closed back pressure waterline value of the queue of any priority.

In the embodiment of this application, the difference between the open back pressure waterline value and the closed back pressure waterline value of each priority queue can be a fixed value. Therefore, for any priority queue, the receiving end device The difference between the open backpressure waterline of the queue of any priority and the fixed value may be determined as the closed backpressure waterline value of the queue of any priority. Optionally, Vc_pri_k=Vo_pri_k-a, a represents the fixed value, for example, a=10.

In the embodiment of the present application, the occupancy of the buffer space of the receiving end device by a queue of any priority is negatively related to the occupancy of the buffer space of the receiving end device by all priority queues below the any priority. , The opening backpressure waterline value of the queue of any priority is negatively related to the amount of buffer space occupied by the receiving end device by the queue of any priority; the opening backpressure waterline of the queue of any priority The value is greater than or equal to the open back pressure waterline value of all priority queues lower than any priority. That is, if queues of all priority levels lower than any priority occupies a large amount of buffer space of the receiving end device, the queue of any priority level occupies a relatively large amount of buffer space of the receiving end device. Smaller, the queues of any priority are less likely to be back pressured, the higher the back pressure waterline value of the queue of any priority is, the higher the priority queue, the higher the back pressure waterline value is, the priority is The lower the queue, the smaller the opening back pressure waterline value. In this way, it can prevent the high-priority queue from being controlled too early to affect the performance of the high-priority queue, and it can also avoid the high-priority queue being controlled too late. .

The second implementation mode: the receiving end device adjusts the back pressure waterline of the queues of each priority according to the amount of the buffer space occupied by all the queues of the receiving end device and the size of the buffer space of the receiving end device. value.

Optionally, please refer to FIG. 9, which shows a flowchart of another method for a receiving end device to adjust the back pressure waterline value of a queue of any priority of the receiving end device according to an embodiment of the present application. For a priority queue, the receiving end device can adjust the back pressure waterline value of the priority queue according to the method shown in FIG. 9. Referring to Figure 9, the method may include:

Sub-step 5081b: Adjust the opening back pressure waterline value of the queue of any priority according to the occupation of the buffer space of the receiving end device by all the queues of any priority and the size of the buffer space. The open back pressure waterline value of the queues is negatively correlated with the amount of buffer space occupied by all queues of any priority level on the receiving end device.

Optionally, the receiving end device may determine the waterline according to the ratio of the occupancy of all queues of any priority to the buffer space of the receiving end device and the size of the buffer space of the receiving end device (hereinafter referred to as the queue occupancy ratio) The coefficient is to adjust the opening back pressure waterline value of the queue of any priority according to the determined waterline coefficient. Wherein, the open backpressure waterline value of the queue of any priority is negatively related to the amount of buffer space occupied by all queues of the any priority on the receiving end device, that is, the buffer of the receiving end device The larger the space occupied by the queue, the smaller the open back pressure waterline and the stronger the back pressure.

Optionally, the receiving end device may determine the ratio of the amount of buffer space occupied by all queues of any priority to the size of the buffer space of the receiving end device (that is, the occupancy of queues of any priority Ratio), according to the queue occupancy ratio of any priority, and the corresponding relationship between the occupancy ratio interval and the waterline coefficient, determine the waterline coefficient corresponding to the queue occupancy ratio of any priority, and determine the waterline coefficient corresponding to the queue occupancy ratio of any priority. The waterline coefficient corresponding to the queue occupancy ratio adjusts the opening back pressure waterline value of the queue of any priority. Optionally, the receiving end may determine the product of the current open backpressure waterline value of the queue of any priority and the corresponding waterline coefficient as the adjusted open backpressure waterline of the queue of any priority value.

For example, the corresponding relationship between the occupancy ratio interval and the waterline coefficient can be shown in Table 1 below:

Table 1

占用比值区间Occupation ratio range	水线系数Waterline coefficient
(0.5，0.8](0.5, 0.8]	0.950.95
(0.8，1](0.8, 1]	0.90.9

For example, suppose that the ratio of the buffer space occupied by all queues with priority pri_k to the buffer space of the receiving device and the buffer space of the receiving device is 0.6, and 0.6 belongs to the ratio range (0.5, 0.8), and the receiving device According to Table 1, it is determined that the waterline coefficient corresponding to (0.5, 0.8) is 0.95, so the receiving end device determines that the adjusted open back pressure waterline value of the queue with priority pri_k is Vo_pri_k=0.95×Vo_pri_k, that is, The opening back pressure waterline value of the queue with priority pri_k is reduced by 5%.

As another example, suppose that the ratio of the occupancy of all queues with priority pri_k to the buffer space of the receiving end device to the size of the buffer space of the receiving end device is 0.85, and this 0.85 belongs to the ratio interval (0.8, 1], the receiving The end device determines that the waterline coefficient corresponding to (0.8, 1) is 0.9 according to Table 1, so the receiving end device determines that the adjusted open back pressure waterline value of the queue with priority pri_k is Vo_pri_k = 0.9×Vo_pri_k, and That is, the opening back pressure waterline value of the queue with priority pri_k drops by 10%.

Sub-step 5082b: According to the open back pressure waterline value of the queue of any priority, adjust the closed back pressure waterline value of the queue of any priority.

For the implementation process of this sub-step 5082b, reference may be made to the aforementioned sub-step 5082a, which will not be repeated in the embodiment of the present application.

The third implementation manner: the receiving end device adjusts the backpressure waterline value of the queues of each priority according to the average delay value of the longest queue.

Optionally, please refer to FIG. 10, which shows a flowchart of another method for a receiving end device to adjust the back pressure waterline value of any priority queue of the receiving end device according to an embodiment of the present application. A priority queue, the receiving end device can adjust the back pressure waterline value of the priority queue according to the method shown in FIG. 10. Referring to Figure 10, the method may include:

Sub-step 5081c, according to the average delay of the M longest queues of any priority and the average delay of all the longest queues of all priorities of the receiving end device, adjust the opening back pressure of the queue of any priority Line value.

Where, M is an integer greater than 0. For the explanation of the M longest queues of any priority, please refer to the explanation of the N longest queues of the first priority in step 503, which will not be repeated in this embodiment of the application. .

Optionally, for each longest queue among the M longest queues of any priority, the receiving end device may determine the delay of each longest queue, and determine the delay according to the delay of the M longest queues The average delay of the M longest queues is determined according to the delays of all the longest queues of all priorities, and the average delay of all the longest queues is determined according to the delays of the M longest queues of any priority. The average value and the average value of the delay of all the longest queues of all priorities of the receiving end device are adjusted to the open back pressure waterline value of the queue of any priority. It should be noted that the number of the longest queues of any two priority levels can be equal or unequal. When the number of the longest queues of any two priority levels is equal, all the longest queues of all priorities of the receiving end device The number can be M×p, p represents the number of priorities of the receiving end device, and M represents the number of the longest queue of each priority in the p priorities; when the longest of two or more priorities When the number of queues is not equal, the number of all the longest queues of all priorities of the receiving end device can be M1+M2+M3+...Mp, and M1～Mp in turn represent the number of the longest queues of priority 1～p .

Optionally, when the average delay of the M longest queues of any priority is greater than the average delay of all the longest queues of all priorities of the receiving end device, the receiving end device adjusts the queue of any priority The opening back pressure waterline value of is reduced by 5% to 10% on the basis of the current opening back pressure waterline value. For example, the receiving end device adjusts the opening back pressure waterline value of the queue of any priority to the target waterline Value range, the target waterline value range can be [0.95×Vo_pri, 0.8×Vo_pri], Vo_pri represents the current open backpressure waterline value of the queue of any priority; when the M of any priority is the longest When the average delay of the queue is not greater than the average delay of all the longest queues of all priorities of the receiving end device, the receiving end device adjusts the open backpressure waterline value of the queue of any priority to that any priority The initial opening backpressure waterline value of the queue of any priority, the initial opening backpressure waterline value of the queue of any priority level is the opening backpressure waterline value pre-configured by the system for the queue of any priority, for example, the any The initial opening backpressure waterline value of a priority queue may be the opening backpressure waterline value set by the receiving end device for the queue of any priority in step 501.

Sub-step 5082c: According to the open back pressure waterline value of the queue of any priority, adjust the closed back pressure waterline value of the queue of any priority.

For the implementation process of this sub-step 5082c, reference may be made to the aforementioned sub-step 5082a, which will not be repeated in this embodiment of the present application.

It should be noted that, in this embodiment of the present application, the receiving end device can dynamically adjust the minimum watermark threshold of the receiving end device, for example, the receiving end device periodically adjusts the minimum watermark threshold of the receiving end device. By adjusting the minimum watermark threshold in the embodiments of the present application, active congestion control can be avoided or reduced as much as possible, and high throughput and low latency can be ensured.

Optionally, the receiving end device adjusting the minimum watermark threshold of the receiving end device may include: the maximum occupancy of the buffer space of the receiving end device by all queues of the receiving end device within the target duration is greater than the minimum watermark threshold and less than The difference between the maximum watermark threshold and the first threshold, or the maximum occupancy of the buffer space of the receiving device by all queues of the receiving end device is greater than the difference between the maximum watermark threshold and the second threshold Value, the receiving end device adjusts the minimum watermark threshold. Optionally, the starting time of the target duration may be the time when the minimum watermark threshold of the receiving end device was adjusted last time. The target duration, the first threshold, and the second threshold can all be determined according to actual conditions. For example, the target duration can be greater than 100ms (milliseconds), and the first threshold can be 12MB (megabits). , The second threshold may be 6MB. For example, within the target duration, if Vo_min<B_total_high<Vo_max-12MB, the receiving end device adjusts the minimum watermark threshold to Vo_min=Vo_min+256KB (kilobits), if B_total_high>Vo_max-6MB, the receiving end device adjusts the minimum The watermark threshold is Vo_min=Vo_min-512KB, where B_total_high represents the maximum occupancy of the buffer space of the receiving end device by all queues of the receiving end device.

Step 509: The receiving end device adjusts the distribution coefficient of the sending credit value of the target queue according to the amount of the buffer space of the receiving end device occupied by the multiple queues of the receiving end device, and the distribution coefficient is compared with the multiple queues of the receiving end device. There is a negative correlation with the occupancy of the buffer space of the receiving end device.

After the receiving-end device notifies the sending-end device to enable congestion control for the target queue corresponding to the first queue in the sending-end device, it can continue to monitor multiple queues of the receiving-end device, and respond to the receiving device according to the multiple queues of the receiving-end device. The occupancy of the buffer space of the end device adjusts the distribution coefficient of the sending credit value of the target queue. Wherein, the distribution coefficient of the sending credit value of the target queue is negatively correlated with the amount of buffer space occupied by the multiple queues of the receiving end device on the receiving end device, that is, the multiple queues of the receiving end device have a negative correlation with the receiving end device. The larger the occupancy of the buffer space of the end device is, the smaller the distribution coefficient of the sending credit value of the target queue is, and the smaller the multiple queues of the receiving end device occupy the buffer space of the receiving end device, the smaller the size of the target queue is. The greater the distribution coefficient of the sending credit value.

Optionally, each priority queue in the receiving end device may have a discard backpressure waterline value. Generally, when multiple queues of the receiving end device occupy or exceed the discarded buffer space of the receiving end device When the pressure watermark value is reversed, the receiving end device can discard the packets in the priority queue to perform passive congestion control. In this embodiment of the present application, the receiving end device may reflect the amount of buffer space occupied by the receiving end device by the multiple queues of the receiving end device, the opening back pressure waterline value of the first queue, and the discarding value of the first queue. Press the waterline value to adjust the distribution coefficient of the sending credit value of the target queue.

Optionally, the receiving end device may determine the first difference between the amount of the buffer space occupied by the receiving end device by the multiple queues of the receiving end device and the opening back pressure waterline value of the first queue, and the first The target difference between the discarded backpressure waterline value of the queue and the open backpressure waterline value of the first queue is adjusted according to the first difference and the target difference to adjust the distribution coefficient of the sending credit value of the target queue. Optionally, when the first difference value is within the first difference value interval, the receiving end device adjusts the distribution coefficient of the sending credit value of the target queue to the first distribution coefficient, and the first distribution coefficient is less than 1 and greater than 0; When the first difference value is in the second difference value interval, the receiving end device adjusts the distribution coefficient of the sending credit value of the target queue to the second distribution coefficient, and the lower limit of the second difference value interval may be equal to the first difference value interval. The upper limit of the difference interval, the second distribution coefficient is less than the first distribution coefficient and greater than 0; when the first difference is in the third difference interval, the receiving end device uses the distribution coefficient of the send credit value of the target queue Adjusted to the third distribution coefficient, the lower limit of the third difference interval may be equal to the upper limit of the second difference interval, the third distribution coefficient is less than the second distribution coefficient and greater than 0; when the first difference When in the fourth difference interval, the receiving end device adjusts the distribution coefficient of the sending credit value of the target queue to the fourth distribution coefficient, and the lower limit of the fourth difference interval may be equal to the upper limit of the third difference interval Value, the fourth distribution coefficient is less than the third distribution coefficient and greater than 0; when the first difference is in the fifth difference interval, the receiving end device adjusts the distribution coefficient of the send credit value of the target queue to 0, and the first difference The lower limit of the five difference interval may be equal to the upper limit of the fourth difference interval. Optionally, the first difference interval is (10%X～20%X), the second difference interval is (20%X～30%X], and the third difference interval is (30%X～40%X). ], the fourth difference interval is (40%X～50%X], the fifth difference interval is (50%X～X], the first distribution coefficient is 0.8, the second distribution coefficient is 0.6, the third distribution The coefficient is 0.4, the fourth distribution coefficient is 0.2, and X is the target difference.

Optionally, when the first difference is in the first difference interval and the length of the back pressure waterline value of the queue whose first priority was adjusted last time is greater than the first time, the receiving end device sends the target queue The distribution coefficient of the credit value is adjusted to the first distribution coefficient; when the first difference is in the second difference interval and the length of time from the previous adjustment of the back pressure waterline value of the first priority queue is greater than the second time length, receive The end device adjusts the distribution coefficient of the sending credit value of the target queue to the second distribution coefficient; when the first difference is in the third difference interval and is less than the previous adjustment of the backpressure waterline value of the first priority queue When the duration is greater than the third duration, the receiving end device adjusts the distribution coefficient of the sending credit value of the target queue to the third distribution coefficient; when the first difference is in the fourth difference interval and the first priority is adjusted before When the duration of the backpressure waterline value of the queue is greater than the fourth duration, the receiving end device adjusts the distribution coefficient of the sending credit value of the target queue to the fourth distribution coefficient; when the first difference is in the fifth difference interval and the distance When the length of the previous adjustment of the backpressure waterline value of the first priority queue is greater than the fifth time length, the receiving end device adjusts the distribution coefficient of the sending credit value of the target queue to 0. Optionally, the first duration, the second duration, the third duration, the fourth duration, and the fifth duration are equal. For example, the first duration, the second duration, the third duration, the fourth duration, and the fifth duration are all 10 ms.

Step 510: The receiving end device allocates a sending credit value to the target queue using the adjusted allocation coefficient.

After the receiving end device adjusts the distribution coefficient of the transmission credit value of the target queue, it may allocate the transmission credit value to the target queue using the adjusted distribution coefficient. Optionally, the receiving end device receives the message containing the queue information (such as queue length and priority, etc.) of the target queue sent by the sending end device, and allocates the target queue according to the queue information in the message with the adjusted distribution coefficient Send credit value. Wherein, the message containing the queue information of the target queue may be REQ.

Step 511: The receiving end device sends a message containing the sending credit value to the sending end device.

Step 512: The sending end device sends the packet in the target queue to the receiving end device according to the sending credit value.

For the implementation process of step 511 and step 512, reference may be made to the foregoing step 506 and step 507, and details are not described in the embodiment of the present application.

Step 513: When the first queue meets the preset closing index, the receiving end device notifies the sending end device to close the congestion control for the target queue of the sending end device.

In the embodiment of the present application, after the receiving end device notifies the sending end device to enable congestion control for the target queue corresponding to the first queue in the sending end device, it can continue to monitor multiple queues of the receiving end device and detect the first queue. Whether the queue meets the preset closing index, and when the first queue meets the preset closing index, the receiving end device notifies the sending end device to close the congestion control for the target queue of the sending end device. Wherein, the preset closing index includes: the amount of the buffer space occupied by the multiple queues of the receiving end device for the receiving end device is less than the closing back pressure waterline value of the first queue.

Optionally, the preset closing indicator may further include one or more of the following:

The length of time that the multiple queues of the receiving end device occupies the buffer space of the receiving end device is less than the closing back pressure waterline value of the first queue reaches or exceeds the preset duration threshold;

The length of the first queue is less than the preset length threshold.

Optionally, the preset closing index includes: the buffer space occupied by the multiple queues of the receiving end device for the receiving end device is less than the closing backpressure waterline value of the first queue, and the multiple queues of the receiving end device When the occupancy of the buffer space of the receiving end device is less than the closing backpressure waterline value of the first queue, the duration reaches or exceeds the preset duration threshold, and the multiple queues of the receiving end device have an impact on the buffer space of the receiving end device. The length of time when the occupancy is less than the closing back pressure waterline value of the first queue reaches or exceeds the preset duration threshold, and the length of the first queue is less than the preset length threshold.

Wherein, the preset length threshold may be the average value of the lengths of all queues of the receiving end device, and the preset duration threshold may be set according to actual conditions. For example, the preset duration threshold may be 100 ms, and N of the first priority For the explanation of the longest queue, reference may be made to step 503, which is not repeated in this embodiment of the application.

Optionally, the receiving end device notifying the sending end device to close the congestion control of the target queue in the sending end device may include: the receiving end device sends a closing control message to the sending end device, and the closing control message instructs the sending end device to The target queue turns off congestion control. Optionally, the shutdown control message may carry an identifier of the target queue and a disable flag bit, the identifier of the target queue indicates the target queue, and the disable flag indicates the sending end device to disable congestion control, so that the shutdown control message Instruct the sending device to turn off congestion control on the target queue. Optionally, the shutdown control message may also carry the identification of the sending end device, so that the sending end device determines according to the identification of the sending end device that the shutdown control message is sent to the sending end device (that is, the sending end device). The identifier indicates the recipient of the shutdown control message). Optionally, the receiving end device may send the closing control message to the sending end device in a broadcast form. For example, the receiving end device broadcasts the closing control message to all sending end devices of the receiving end device.

It should be noted that the receiving end device can notify the sending end device to close congestion control for all target queues in the sending end device, or it can notify the sending end device to close congestion control for one or several target queues in the sending end device. Control, for example, the receiving end device notifies the sending end device to turn off congestion control for queues with higher priority on the receiving end device to ensure that queues with higher priority can be scheduled preferentially, which is not limited in this embodiment of the application.

Step 514: The sending end device closes the congestion control for the target queue of the sending end device according to the notification of the receiving end device.

Optionally, according to step 513, the receiving end device notifies the sending end device to close the congestion control of the target queue when the first queue of the receiving end device meets the preset closing index, and the preset closing indicator includes: the number of the receiving end device. The occupancy of the buffer space of the receiving end device by each queue is less than the closed back pressure waterline value of the first queue.

The length of the first queue is less than the preset length threshold.

Optionally, the sending-end device closing congestion control to the target queue of the sending-end device according to the notification of the receiving-end device may include: the sending-end device receives a closing control message sent by the receiving-end device, and controlling the target queue according to the closing control message. The congestion control is closed, where the close control message instructs the sending end device to close the congestion control for the target queue. Optionally, the shutdown control message may carry an identifier of the target queue and a disable flag bit, the identifier of the target queue indicates the target queue, and the disable flag indicates the sending end device to disable congestion control, so that the shutdown control message Instruct the sending device to turn off congestion control on the target queue. Optionally, the shutdown control message may also carry an identifier of the sender device, and the identifier of the sender device indicates the recipient of the shutdown control message. The sending end device can parse the shutdown control message to obtain the identification of the sending end device, the de-enable flag bit, and the identification of the target queue. According to the identification of the sending end device, it is determined that the shutdown control message is sent to the sending end device. The disable flag determines that congestion control needs to be turned off, and the target queue is determined according to the identifier of the target queue, so that the congestion control is turned off for the target queue according to the shutdown control message.

For example, please refer to FIG. 11, which shows another schematic diagram of congestion control provided by an embodiment of the present application. Referring to FIG. 11, when multiple queues of the receiving end device occupy more than the buffer space of the receiving end device When the backpressure waterline corresponding to the priority pri_k is turned on (that is, the backpressure waterline for the queue with priority pri_k) Vo_pri_k, congestion control is turned on. When the multiple queues of the receiving end device occupy less than the closed back pressure waterline corresponding to the priority pri_k (that is, the closed back pressure water line of the queue with priority pri_k) Vc_pri_k, However, when the multiple queues of the receiving end device occupy the buffer space of the receiving end device less than the priority pri_k corresponding to the closing time of the backpressure waterline t1 is less than the preset duration threshold, even if the multiple queues of the receiving end device The occupancy of the buffer space of the receiving end device is less than the closing back pressure waterline Vc_pri_k corresponding to the priority pri_k, and the congestion control is not closed. When the multiple queues of the receiving end device occupies the buffer space of the receiving end device less than the closed backpressure waterline Vc_pri_k corresponding to the priority pri_k, and the multiple queues of the receiving end device have buffer space for the receiving end device When the occupancy of is less than the time t2 for closing the backpressure waterline corresponding to the priority pri_k reaches or exceeds the preset time threshold, the congestion control is turned off. In this way, it is possible to avoid turning on congestion control due to instantaneous congestion, and turning off congestion control due to instantaneous non-congestion, and improve the flexibility of congestion control.

It should be noted that the sequence of the steps of the congestion control method provided in the embodiments of this application can be adjusted appropriately, and the steps can also be increased or decreased accordingly according to the situation. Anyone familiar with the technical field within the technical scope disclosed in this application can easily Thinking of the method of change should be covered in the scope of protection of this application, so I will not repeat it.

To sum up, in the congestion control method provided by the embodiments of the present application, the receiving end device monitors multiple queues of the receiving end device, and when the first queue of the multiple queues of the receiving end device meets the preset opening index, it notifies the The sending end device enables congestion control on the target queue corresponding to the first queue in the sending end device, and the sending end device enables congestion control on the target queue according to the notification of the receiving end device. The congestion control method helps to solve the congestion problem of the current communication network, and helps to improve the resource utilization rate and transmission quality of the communication network.

The congestion control method provided by the embodiments of the present application combines active congestion control and passive congestion control. After using passive congestion control to sense congestion, by actively controlling the congestion, congestion control can be performed from the entire communication network, and can be targeted at a specific priority. In the process of congestion control, you can flexibly adjust parameters such as the back pressure waterline and the distribution coefficient of the sending credit value, and can make full use of the buffer space of the receiving end device to avoid instantaneous congestion of messages The back pressure is on the sending end device to ensure the throughput of the sending end device, thereby ensuring the throughput of the workstation connected to the sending end device, and improving the flexibility of congestion control.

The congestion control method provided by the embodiment of the present application does not need to consider the backpressure waterline value and the sending credit value before the sender device starts congestion control, so that the traffic of the sender device can be pushed as much as possible, and all the traffic before the congestion control is enabled The queues try their best to maintain throughput.

The following are device embodiments of this application, which can be used to implement the method embodiments of this application. For details that are not disclosed in the device embodiments of this application, please refer to the method embodiments of this application.

Please refer to FIG. 12, which shows a schematic diagram of the logical structure of a congestion control apparatus 1200 provided by an embodiment of the present application. The congestion control apparatus 1200 may be a functional component in a network device, and the network device may be the one in the foregoing embodiment. Receiving device. Referring to FIG. 12, the congestion control device 1200 includes: a monitoring module 1210 and an opening module 1220. The monitoring module 1210 can be used to execute step 502 in the embodiment shown in FIG. 5, and the opening module 1220 can be used to execute the step shown in FIG. Step 503 in the embodiment.

The monitoring module 1210 is used to monitor multiple queues of the receiving end device, and each queue of the receiving end device corresponds to one or more queues in the sending end device;

The opening module 1220 is configured to notify the sending end device to start congestion for the target queue corresponding to the first queue in the sending end device when the first queue among the plurality of queues of the receiving end device meets the preset opening index control;

Wherein, the preset opening index includes: the amount of the buffer space occupied by the multiple queues of the receiving end device for the receiving end device reaches or exceeds the open back pressure waterline value of the first queue.

The length of the first queue reaches or exceeds the preset length threshold.

Optionally, the opening module 1220 is configured to send an opening control message to the sending end device, where the opening control message instructs the sending end device to enable congestion control on the target queue.

Optionally, please refer to FIG. 13, which shows a schematic diagram of the logical structure of another congestion control apparatus 1200 provided in an embodiment of the present application. Referring to FIG. 13, based on FIG. 12, the congestion control apparatus 1200 further includes: The closing module 1230 can be used to execute step 513 in the embodiment shown in FIG. 5.

The closing module 1230 is configured to notify the sending end device to close congestion control for the target queue of the sending end device when the first queue meets the preset closing index;

Wherein, the preset closing index includes: the amount of the buffer space occupied by the multiple queues of the receiving end device for the receiving end device is less than the closing back pressure waterline value of the first queue.

The length of the first queue is less than the preset length threshold.

Optionally, please continue to refer to FIG. 13. The congestion control apparatus 1200 further includes: an allocation module 1240 and a sending module 1250. The allocation module 1240 can be used to perform steps 505 and 510 in the embodiment shown in FIG. The module 1250 can be used to execute step 506 and step 511 in the embodiment shown in FIG. 5.

The allocation module 1240 is used to allocate the sending credit value to the target queue;

The sending module 1250 is configured to send a message containing the sending credit value to the sending end device, so that the sending end device sends the message in the target queue to the receiving end device according to the sending credit value.

Optionally, please continue to refer to FIG. 13, the congestion control apparatus 1200 further includes: a coefficient adjustment module 1260, and the coefficient adjustment module 1260 may be used to perform step 509 in the embodiment shown in FIG. 5.

The coefficient adjustment module 1260 is configured to adjust the distribution coefficient of the sending credit value of the target queue according to the amount of the buffer space occupied by the receiving-end device by the multiple queues of the receiving-end device. The queue is negatively related to the occupancy of the buffer space of the receiving end device.

Optionally, each of the multiple queues of the receiving end device has a back pressure waterline value, and the back pressure waterline values of queues with different priorities are different. Please continue to refer to FIG. 13, the congestion control apparatus 1200 also includes : Waterline adjustment module 1270. The waterline adjustment module 1270 can be used to execute step 508 in the embodiment shown in FIG. 5.

The waterline adjustment module 1270 is configured to adjust each of the receiving-end device when the amount of the buffer space occupied by the receiving-end device reaches or exceeds the minimum waterline threshold of the receiving-end device. The back pressure waterline value of the priority queue.

Optionally, the back pressure waterline value includes an open back pressure waterline value and a closed back pressure waterline value,

The waterline adjustment module 1270 is used for:

For any priority queue:

According to the maximum watermark threshold of the receiving end device, the minimum watermark threshold of the receiving end device, the occupancy of the buffer space of the receiving end device by all priority queues lower than any priority, and the receiving end device Adjust the open back pressure waterline value of the queue of any priority for the occupation of the buffer space of the receiving end device by all the queues;

According to the open back pressure waterline value of the queue of any priority, the closed back pressure waterline value of the queue of any priority is adjusted.

Optionally, the occupancy of the buffer space of the receiving end device by the queue of any priority is negatively correlated with the occupancy of the buffer space of the receiving end device by all priority queues lower than the any priority. The open back pressure waterline value of the queue of any priority is negatively correlated with the amount of buffer space occupied by the queue of any priority on the receiving end device;

The opening backpressure waterline value of the queue of any priority is greater than or equal to the opening backpressure waterline value of all priority queues lower than the any priority.

Optionally, the waterline adjustment module 1270 is used to:

According to the maximum watermark threshold and the minimum watermark threshold, all priority queues below any priority occupy the buffer space of the receiving end device, and the buffer space of all queues of the receiving end device for the receiving end device The occupancy of, the waterline value formula is used to calculate the open back pressure waterline value of the queue of any priority;

Adjust the opening back pressure waterline value of the queue of any priority according to the calculation result;

Among them, the waterline value formula is:

The waterline adjustment module 1270 is used for:

For any priority queue:

According to the occupancy of the buffer space of the receiving device by all queues of any priority and the size of the buffer space, adjust the opening back pressure waterline value of the queue of any priority, and the value of the queue of any priority Turning on the back pressure waterline value is negatively related to the amount of buffer space occupied by all queues of any priority level on the receiving end device;

The waterline adjustment module 1270 is used for:

For any priority queue:

According to the average delay of the M longest queues of any priority and the average delay of all the longest queues of all priorities of the receiving end device, adjust the opening backpressure waterline value of the queue of any priority, M is an integer greater than 0;

Optionally, please continue to refer to FIG. 13, the congestion control apparatus 1200 further includes:

The threshold adjustment module 1280 is used for the maximum occupancy of the buffer space of the receiving end device by all queues of the receiving end device within the target duration is greater than the minimum watermark threshold and less than the difference between the maximum waterline threshold and the first threshold, Or, when the maximum occupancy is greater than the difference between the maximum watermark threshold and the second threshold, the minimum watermark threshold is adjusted.

To sum up, in the congestion control apparatus provided by the embodiment of the present application, the receiving end device monitors multiple queues of the receiving end device, and when the first queue of the multiple queues of the receiving end device meets the preset opening index, a notification is sent The end device enables congestion control on the target queue corresponding to the first queue in the sender device, and the sender device enables congestion control on the target queue according to the notification of the receiver device. The congestion control method helps to solve the congestion problem of the current communication network, and helps to improve the resource utilization rate and transmission quality of the communication network.

Please refer to FIG. 14, which shows a schematic diagram of the logical structure of yet another congestion control apparatus 1400 provided by an embodiment of the present application. The congestion control apparatus 1400 may be a functional component in a network device, and the network device may be the one in the foregoing embodiment. The sending end device. Referring to FIG. 14, the congestion control apparatus 1400 includes: a first receiving module 1410 and an opening module 1420. The first receiving module 1410 and the opening module 1420 can be used to execute step 504 in the embodiment shown in FIG. 5.

The first receiving module 1410 is configured to receive a notification sent by the receiving end device to enable congestion control on the target queue of the sending end device, and each queue of the receiving end device corresponds to one or more queues in the sending end device, The target queue corresponds to the first queue of the receiving end device;

The opening module 1420 is configured to enable congestion control for the target queue of the sending end device according to the notification of the receiving end device.

Optionally, the first receiving module 1410 is configured to receive a turn-on control message sent by a receiving end device, where the turn-on control message carries an identifier of the target queue and an enable flag bit;

The opening module 1420 is configured to determine the target queue according to the identifier of the target queue, and enable congestion control for the target queue according to the enable flag bit.

Optionally, please refer to FIG. 15, which shows a schematic diagram of the logical structure of another congestion control apparatus 1400 provided by an embodiment of the present application. Referring to FIG. 15, based on FIG. 14, the congestion control apparatus 1400 further includes: The second receiving module 1430 and the closing module 1440 can be used to perform step 514 in the embodiment shown in FIG. 5.

The second receiving module 1430 is configured to receive a notification sent by the receiving end device to close the congestion control of the target queue of the sending end device;

The closing module 1440 is configured to close congestion control for the target queue of the sending end device according to the notification of the receiving end device.

Optionally, the second receiving module 1430 is configured to receive a shutdown control message sent by the receiving end device, where the shutdown control message carries an identifier of the target queue and a disabling flag bit;

The closing module 1440 is configured to determine the target queue according to the identifier of the target queue, and turn off congestion control for the target queue according to the disable flag bit.

Optionally, please continue to refer to FIG. 15. The congestion control apparatus 1400 further includes: a third receiving module 1450 and a sending module 1460. The third receiving module 1450 and the sending module 1460 can be used to perform the embodiment shown in FIG. 5的Steps 507 and 512.

The third receiving module 1450 is configured to receive the message containing the sending credit value sent by the receiving end device;

The sending module 1460 is configured to send the message in the target queue to the receiving end device according to the sending credit value.

It should be noted that when the congestion control device provided in the above embodiment performs congestion control, only the division of the above-mentioned functional modules is used as an example for illustration. In actual applications, the above-mentioned function allocation can be completed by different functional modules according to needs. That is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the congestion control device provided in the foregoing embodiment and the congestion control method embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, and will not be repeated here.

Please refer to FIG. 16, which shows a schematic diagram of the hardware structure of a congestion control apparatus 1600 provided by an embodiment of the present application. The congestion control apparatus 1600 may be a network device, and the network device may be the sender device in the foregoing embodiment or At least one of the receiving end devices. 16, the congestion control apparatus 1600 includes a processor 1602, a memory 1604, a communication interface 1606, and a bus 1608. The processor 1602, the memory 1604, and the communication interface 1606 are communicatively connected to each other through the bus 1608. Those skilled in the art should understand that the connection manner between the processor 1602, the memory 1604, and the communication interface 1606 shown in FIG. 16 is only exemplary. In the implementation process, the processor 1602, the memory 1604, and the communication interface 1606 may also Use other connection methods other than the bus 1608 to communicate with each other.

Among them, the memory 1604 may be used to store instructions 16042 and data 16044. In the embodiment of the present application, the memory 1604 may be various types of storage media, such as random access memory (RAM), read-only memory (ROM), and non-volatile RAM (non-volatile RAM). -volatile RAM, NVRAM), programmable ROM (programmable ROM, PROM), erasable PROM (erasable PROM, EPROM), electrically erasable PROM (electrically erasable PROM, EEPROM), flash memory, optical memory and registers, etc. In addition, the memory 1604 may include a hard disk and/or a memory.

The processor 1602 may be a general-purpose processor, and the general-purpose processor may be a processor that executes specific steps and/or operations by reading and executing instructions (for example, instructions 16042) stored in a memory (for example, the memory 1604). The processor may use data (for example, data 16044) stored in the memory (for example, the memory 1604) in the process of performing the above steps and/or operations. The general-purpose processor may be, for example, but not limited to, a central processing unit (CPU). In addition, the processor 1602 may also be a dedicated processor. The dedicated processor may be a processor specially designed to perform specific steps and/or operations. The dedicated processor may be, for example, but not limited to, a digital signal processor ( Digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), etc. In addition, the processor 1602 may also be a combination of multiple processors, such as a multi-core processor. The processor 1602 may include one or more circuits to execute all or part of the steps of the congestion control method provided in the foregoing embodiments.

Wherein, the communication interface 1606 may include an input/output (I/O) interface, a physical interface, a logical interface, and other interfaces used to realize the interconnection of devices within the congestion control apparatus 1600, and to realize the congestion control apparatus 1600. Interface for interconnection with other equipment (such as network equipment or user equipment). The physical interface can be a gigabit ethernet (GE), which can be used to realize the interconnection between the congestion control apparatus 1600 and other equipment (such as network equipment or user equipment), and the logical interface is an internal interface of the congestion control apparatus 1600. It can be used to realize the interconnection of devices within the congestion control apparatus 1600. It is easy to understand that the communication interface 1606 may be used for the congestion control apparatus 1600 to communicate with other network equipment and/or user equipment. For example, the communication interface 1606 is used for sending and receiving information between the congestion control apparatus 1600 and other network equipment.

The bus 1608 may be of any type, and is used to implement a communication bus interconnecting the processor 1602, the memory 1604, and the communication interface 1606, such as a system bus.

The above-mentioned devices may be respectively arranged on independent chips, or at least partly or fully arranged on the same chip. Whether each device is arranged independently on different chips or integrated on one or more chips often depends on the needs of product design. The embodiment of the present application does not limit the specific implementation form of the foregoing device.

The congestion control apparatus 1600 shown in FIG. 16 is only exemplary. In the implementation process, the congestion control apparatus 1600 may also include other components, which will not be listed here. The congestion control apparatus 1600 shown in FIG. 16 can control network congestion by executing all or part of the steps of the congestion control method provided in the above embodiment.

The embodiment of the present application provides a congestion control system. The congestion control system may include the sending end device and the receiving end device in the foregoing embodiment. In a possible implementation manner, the sending end device may include FIG. 14 or FIG. For the congestion control apparatus 1400 shown in FIG. 15, the receiving end device may include the congestion control apparatus 1200 shown in FIG. 12 or FIG. 13. In another possible implementation manner, at least one of the sending end device and the receiving end device may include the congestion control apparatus 1600 shown in FIG. 16.

For example, please refer to FIG. 17, which shows a schematic diagram of a congestion control system 1700 provided by an embodiment of the present application. Referring to FIG. 17, the congestion control system 1700 includes: a sending end device 1710 and a receiving end device 1720.

The receiving end device 1720 is used to monitor multiple queues of the receiving end device 1720, and each queue of the receiving end device 1720 corresponds to one or more queues in the sending end device 1710; when the receiving end device 1720 is When the first queue among the plurality of queues meets the preset opening index, the sending end device 1710 is notified to enable congestion control for the target queue corresponding to the first queue in the sending end device 1710; wherein, the preset opening index includes: The occupancy of the buffer space of the receiving end device 1720 by the multiple queues of the receiving end device 1720 reaches or exceeds the opening back pressure waterline value of the first queue;

The sending end device 1710 is configured to enable congestion control for the target queue corresponding to the first queue in the sending end device 1710 according to the notification of the receiving end device 1720.

The first queue belongs to the N longest queues of the first priority among the multiple queues of the receiving end device 1720, the priority of the first queue is the first priority, and when the first priority is the multiple When the priority of the queue is the lowest priority, N is an integer greater than 0, and when the first priority is the non-lowest priority among the priorities of the multiple queues, N is an integer greater than or equal to 0;

The length of the first queue reaches or exceeds the preset length threshold.

Optionally, the receiving end device 1720 is configured to send an opening control message to the sending end device 1710, where the opening control message instructs the sending end device 1710 to enable congestion control for the target queue.

The sending end device 1710 is configured to: receive the opening control message sent by the receiving end device 1710; and enable congestion control for the target queue according to the opening control message.

Optionally, the receiving end device 1720 is further configured to: when the first queue meets a preset closing index, notify the sending end device 1710 to close congestion control for the target queue of the sending end device 1710; wherein, the preset The closing index includes: the amount of the buffer space occupied by the multiple queues of the receiving end device 1720 to the receiving end device 1720 is less than the closing backpressure waterline value of the first queue;

The sending end device 1710 is further configured to: turn off congestion control for the target queue of the sending end device 1710 according to the notification of the receiving end device 1720.

The length of time that the multiple queues of the receiving end device 1720 occupy the buffer space of the receiving end device 1720 is less than the closing back pressure waterline value of the first queue reaches or exceeds the preset duration threshold;

The first queue does not belong to the N longest queues of the first priority among the multiple queues of the receiving end device 1720, the priority of the first queue is the first priority, and when the first priority is the multiple When the lowest priority among the priorities of the two queues, N is an integer greater than 0, and when the first priority is the non-lowest priority among the priorities of the multiple queues, N is an integer greater than or equal to 0;

The length of the first queue is less than the preset length threshold.

Optionally, the receiving end device 1720 is further configured to: allocate a sending credit value to the target queue; sending a message containing the sending credit value to the sending end device 1710;

The sending end device 1710 is further configured to: receive a message containing a sending credit value sent by the receiving end device 1720; and send the message in the target queue to the receiving end device according to the sending credit value.

Optionally, the receiving end device 1720 is further configured to adjust the distribution coefficient of the sending credit value of the target queue according to the amount of the buffer space occupied by the receiving end device 1720 by the multiple queues of the receiving end device 1720, and the allocation is The coefficient is negatively related to the occupancy of the buffer space of the receiving end device 1720 by the multiple queues of the receiving end device 1720.

Optionally, each of the multiple queues of the receiving end device 1720 has a back pressure waterline value, and the back pressure waterline values of queues with different priorities are different, and the receiving end device 1720 is also used to: When the occupancy of the buffer space of the receiving end device 1720 by the multiple queues of the end device 1720 reaches or exceeds the minimum waterline threshold of the receiving end device 1720, the back pressure level of the queues of each priority of the receiving end device 1720 is adjusted. Line value.

The receiving end device 1720 is used to: for any priority queue:

According to the maximum watermark threshold of the receiving end device 1720 and the minimum watermark threshold of the receiving end device 1720, the occupancy of the buffer space of the receiving end device 1720 by all priority queues lower than any one of the priorities, and Adjust the amount of the buffer space occupied by all the queues of the receiving end device 1720 to the receiving end device 1720, and adjust the opening back pressure waterline value of the queue of any priority;

Optionally, the occupancy of the buffer space of the receiving end device 1720 by the queue of any priority and the occupancy of the buffer space of the receiving end device 1720 by the queues of all priorities lower than the any priority are negative. Correlation, the opening back pressure waterline value of the queue of any priority is negatively correlated with the amount of the buffer space occupied by the queue of any priority on the receiving end device 1720;

Optionally, the receiving end device 1720 is used to:

According to the maximum watermark threshold and the minimum watermark threshold, the occupancy of the buffer space of the receiving end device 1720 by all priority queues lower than any one of the priority levels, and all the queues of the receiving end device 1720 The amount of buffer space occupied by the receiving end device 1720 is calculated by using a waterline value formula to calculate the open backpressure waterline value of the queue of any priority;

Among them, the waterline value formula is:

Vo_pri_k represents the open backpressure waterline value of the queue with priority pri_k, B_pri_i represents the occupation of the buffer space of the receiving end device 1720 by all queues with priority pri_i, and B_total represents the value of all queues of the receiving end device 1720. The occupancy of the buffer space of the receiving end device 1720, V_max represents the maximum watermark threshold, and V_min represents the minimum watermark threshold.

The receiving end device 1720 is used to: for any priority queue:

According to the occupancy of the buffer space of the receiving end device 1720 by all queues of any priority and the size of the buffer space, adjust the opening back pressure waterline value of the queue of any priority. The opening back pressure waterline value of the queue is negatively correlated with the amount of the buffer space occupied by the receiving end device 1720 by all queues of any one of the priority levels;

The receiving end device 1720 is used to: for any priority queue:

According to the average delay of the M longest queues of any priority and the average delay of all the longest queues of all the priorities of the receiving end device 1720, adjust the opening backpressure waterline of the queue of any priority Value, the M is an integer greater than 0;

Optionally, the receiving end device 1720 is further configured to: within the target duration, the maximum occupancy of all queues of the receiving end device 1720 to the buffer space of the receiving end device 1720 is greater than the minimum waterline threshold and less than the maximum waterline The difference between the threshold and the first threshold, or when the maximum occupancy is greater than the difference between the maximum watermark threshold and the second threshold, adjust the minimum watermark threshold.

In summary, in the congestion control system provided by the embodiment of the present application, the receiving end device monitors multiple queues of the receiving end device, and when the first queue of the multiple queues of the receiving end device meets the preset opening index, the notification is sent The end device enables congestion control on the target queue corresponding to the first queue in the sender device, and the sender device enables congestion control on the target queue according to the notification of the receiver device. The congestion control method helps to solve the congestion problem of the current communication network, and helps to improve the resource utilization rate and transmission quality of the communication network.

The embodiment of the present application provides a computer-readable storage medium in which a computer program is stored, and when the computer program is executed by a processor, it implements all or part of the congestion control method provided by the embodiment shown in FIG. 5 step.

The embodiment of the present application provides a computer program product containing instructions. When the computer program product runs on a computer, the computer executes all or part of the steps of the congestion control method provided by the embodiment shown in FIG. 5.

The embodiment of the present application provides a chip that includes a programmable logic circuit and/or program instructions, and when the chip is running, it is used to implement all or part of the steps of the congestion control method provided by the embodiment shown in FIG. 5.

In the embodiments of this application, the sending end device and the receiving end device can be on the same physical device or virtual device, that is, the device has sending and receiving functions, which can perform all operations or steps of the sending end device, and can also execute the receiving end device. All operations or steps.

It should be understood that the term "and/or" in this text is only an association relationship describing the associated objects, indicating that there can be three types of relationships, for example, A and/or B, which can mean: A alone exists, and both A and B exist. , There are three cases of B alone. In addition, the character "/" in this text generally indicates that the associated objects before and after are in an "or" relationship.

It should be understood that in the embodiments of the present application, the size of the sequence numbers of the foregoing processes does not mean the order of execution. The execution order of the processes should be determined by their functions and internal logic, and should not constitute the implementation process of the embodiments of the present application. Any restrictions.

The serial numbers of the foregoing embodiments of the present application are for description only, and do not represent the superiority of the embodiments.

A person of ordinary skill in the art can understand that all or part of the steps in the above embodiments can be implemented by hardware, or by a program to instruct relevant hardware. The program can be stored in a computer-readable storage medium. The storage medium mentioned can be a read-only memory, a magnetic disk or an optical disk, etc.

The above are only optional embodiments of this application and are not intended to limit this application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included in the protection of this application. Within range.

Claims

A congestion control method, characterized in that the method includes:

Monitoring multiple queues of the receiving end device, each queue of the receiving end device corresponding to one or more queues in the sending end device;

When the first queue of the multiple queues of the receiving end device meets a preset opening index, notifying the sending end device to enable congestion control on the target queue corresponding to the first queue in the sending end device;

Wherein, the preset opening index includes: the amount of the buffer space of the receiving end device occupied by the multiple queues of the receiving end device reaches or exceeds the open back pressure waterline value of the first queue.
The method of claim 1, wherein:

The preset opening indicator further includes one or more of the following:

The first queue belongs to the N longest queues of the first priority among the multiple queues of the receiving end device, and the priority of the first queue is the first priority. When the first priority is When it is the lowest priority among the priorities of the multiple queues, the N is an integer greater than 0, and when the first priority is the non-lowest priority among the priorities of the multiple queues, Said N is an integer greater than or equal to 0;

The length of the first queue reaches or exceeds a preset length threshold.
The method according to claim 1 or 2, characterized in that:

The informing the sending end device to enable congestion control on the target queue corresponding to the first queue in the sending end device includes:

Sending an open control message to the sending end device, where the open control message instructs the sending end device to open congestion control on the target queue.
The method according to any one of claims 1 to 3, characterized in that:

After notifying the sending end device to enable congestion control on the target queue corresponding to the first queue in the sending end device, the method further includes:

When the first queue meets a preset closing index, notify the sending end device to close congestion control on the target queue of the sending end device;

Wherein, the preset closing index includes: the amount of the buffer space of the receiving end device occupied by the multiple queues of the receiving end device is less than the closing back pressure waterline value of the first queue.
The method of claim 4, wherein:

The preset closing indicator also includes one or more of the following:

The duration of the multiple queues of the receiving end device occupying the buffer space of the receiving end device less than the closing backpressure waterline value of the first queue reaches or exceeds a preset duration threshold;

The first queue does not belong to the N longest queues of the first priority among the multiple queues of the receiving end device, and the priority of the first queue is the first priority. When the first priority is When the priority is the lowest priority among the priorities of the multiple queues, the N is an integer greater than 0, and when the first priority is the non-lowest priority among the priorities of the multiple queues, The N is an integer greater than or equal to 0;

The length of the first queue is less than a preset length threshold.
The method according to any one of claims 1 to 5, wherein:

After notifying the sending end device to enable congestion control on the target queue corresponding to the first queue in the sending end device, the method further includes:

Allocating a sending credit value to the target queue;

Sending a message including the sending credit value to the sending end device, so that the sending end device sends the message in the target queue to the receiving end device according to the sending credit value.
The method according to claim 6, wherein the method further comprises:

Adjust the distribution coefficient of the sending credit value of the target queue according to the amount of the multiple queues of the receiving end device occupying the buffer space of the receiving end device, and the distribution coefficient is the same as that of the multiple queues of the receiving end device. There is a negative correlation with the occupancy of the buffer space of the receiving end device.
The method according to any one of claims 1 to 7, wherein each of the multiple queues of the receiving end device has a back pressure waterline value, and queues of different priorities have different back pressure waterline values. , The method further includes:

When the occupancy of the buffer space of the receiving end device by the multiple queues of the receiving end device reaches or exceeds the minimum waterline threshold of the receiving end device, adjust the queue size of each priority of the receiving end device. Back pressure waterline value.
The method of claim 8, wherein:

The back pressure waterline value includes an open back pressure waterline value and a closed back pressure waterline value,

The adjusting the back pressure waterline value of each priority queue of the receiving end device includes:

For any priority queue:

According to the maximum watermark threshold of the receiving end device and the minimum watermark threshold of the receiving end device, all priority queues below any one of the priority levels occupy the buffer space of the receiving end device Adjusting the backpressure waterline value of the queue of any one of the priority levels, and the occupancy of all queues of the receiving end device to the buffer space of the receiving end device;

According to the open back pressure waterline value of the queue of any priority, the closed back pressure waterline value of the queue of any priority is adjusted.
The method of claim 9, wherein:

The occupation of the buffer space of the receiving end device by the queue of any priority is inversely related to the occupation of the buffer space of the receiving end device by the queues of all priorities lower than the any priority; The opening back pressure waterline value of the queue of any priority is negatively related to the amount of the buffer space occupied by the queue of any priority on the receiving end device;

The opening backpressure waterline value of the queue of any priority is greater than or equal to the opening backpressure waterline value of all priority queues lower than the any priority.
The method according to claim 9 or 10, wherein:

According to the maximum watermark threshold of the receiving end device and the minimum watermark threshold of the receiving end device, the buffer space of the receiving end device for all priority queues lower than any one of the priorities And the occupancy of all queues of the receiving end device to the buffer space of the receiving end device, adjusting the opening back pressure waterline value of the queue of any priority, including:

According to the maximum watermark threshold and the minimum watermark threshold, all priority queues below any one of the priority levels occupy the buffer space of the receiving end device, and all of the receiving end device The amount of the buffer space occupied by the receiving end device by the queue is calculated by using a waterline value formula to calculate the open backpressure waterline value of the queue of any priority;

Adjusting the opening back pressure waterline value of the queue of any priority according to the calculation result;

Wherein, the waterline value formula is:

The Vo_pri_k represents the opening back pressure waterline value of the queue with priority pri_k, the B_pri_i represents the occupation of the buffer space of the receiving end device by all queues with priority pri_i, and the B_total represents the receiving end The occupied amount of the buffer space of the receiving end device by all queues of the device, where the V_max represents the maximum watermark threshold, and the V_min represents the minimum watermark threshold.
The method of claim 8, wherein:

The back pressure waterline value includes an open back pressure waterline value and a closed back pressure waterline value,

The adjusting the back pressure waterline value of each priority queue of the receiving end device includes:

For any priority queue:

According to the occupancy of the buffer space of the receiving end device by all the queues of any priority and the size of the buffer space, adjust the opening back pressure waterline value of the queue of any priority, and The opening backpressure waterline value of a priority queue is inversely related to the amount of buffer space occupied by all queues of any one of the priorities of the receiving end device;

According to the open back pressure waterline value of the queue of any priority, the closed back pressure waterline value of the queue of any priority is adjusted.
The method of claim 8, wherein:

The back pressure waterline value includes an open back pressure waterline value and a closed back pressure waterline value,

The adjusting the back pressure waterline value of each priority queue of the receiving end device includes:

For any priority queue:

According to the average delay of the M longest queues of any priority and the average delay of all the longest queues of all the priorities of the receiving end device, adjust the opening backpressure of the queue of any priority Waterline value, where M is an integer greater than 0;

According to the open back pressure waterline value of the queue of any priority, the closed back pressure waterline value of the queue of any priority is adjusted.
The method according to any one of claims 8 to 13, wherein the method further comprises:

The maximum occupancy of all queues of the receiving end device to the buffer space of the receiving end device within the target duration is greater than the minimum watermark threshold and less than the difference between the maximum watermark threshold and the first threshold, Alternatively, when the maximum occupancy is greater than the difference between the maximum watermark threshold and the second threshold, the minimum watermark threshold is adjusted.
A congestion control method, characterized in that the method includes:

The congestion control is enabled on the target queue of the sending end device according to the notification of the receiving end device. Each queue of the receiving end device corresponds to one or more queues in the sending end device, and the target queue is corresponding to the receiving end. Corresponding to the first queue of the device;

Wherein, the receiving end device notifies the sending end device to enable congestion control on the target queue when the first queue meets a preset opening index, and the preset opening index includes: multiple numbers of the receiving end device The occupation of the buffer space of the receiving end device by the queue reaches or exceeds the opening back pressure waterline value of the first queue.
The method of claim 15, wherein:

The preset opening indicator further includes one or more of the following:

The first queue belongs to the N longest queues of the first priority among the multiple queues of the receiving end device, and the priority of the first queue is the first priority. When the first priority is When it is the lowest priority among the priorities of the multiple queues, the N is an integer greater than 0, and when the first priority is the non-lowest priority among the priorities of the multiple queues, Said N is an integer greater than or equal to 0;

The length of the first queue reaches or exceeds a preset length threshold.
The method according to claim 15 or 16, characterized in that:

The enabling congestion control on the target queue of the sending end device according to the notification of the receiving end device includes:

Receiving an opening control message sent by the receiving end device, where the opening control message instructs the sending end device to enable congestion control on the target queue;

Turn on congestion control on the target queue according to the turn on control message.
The method according to any one of claims 15 to 17, characterized in that,

After enabling congestion control on the target queue of the sending end device according to the notification of the receiving end device, the method further includes:

Turn off congestion control on the target queue of the sending end device according to the notification of the receiving end device;

Wherein, the receiving end device notifies the sending end device to close congestion control for the target queue when the first queue meets a preset closing index, and the preset closing indicator includes: multiple numbers of the receiving end device The occupation of the buffer space of the receiving end device by the queue is less than the closing back pressure waterline value of the first queue.
The method of claim 18, wherein:

The preset closing indicator also includes one or more of the following:

The duration of the multiple queues of the receiving end device occupying the buffer space of the receiving end device less than the closing backpressure waterline value of the first queue reaches or exceeds a preset duration threshold;

The first queue does not belong to the N longest queues of the first priority among the multiple queues of the receiving end device, and the priority of the first queue is the first priority. When the first priority is When the priority is the lowest priority among the priorities of the multiple queues, the N is an integer greater than 0, and when the first priority is the non-lowest priority among the priorities of the multiple queues, The N is an integer greater than or equal to 0;

The length of the first queue is less than a preset length threshold.
The method according to any one of claims 15 to 19, characterized in that,

After enabling congestion control on the target queue of the sending end device according to the notification of the receiving end device, the method further includes:

Receiving a message containing a sending credit value sent by the receiving end device;

Sending the message in the target queue to the receiving end device according to the sending credit value.
A congestion control device, characterized in that the device includes:

The monitoring module is used to monitor multiple queues of the receiving end device, and each queue of the receiving end device corresponds to one or more queues in the sending end device;

An opening module, configured to notify the sending end device of the target queue corresponding to the first queue in the sending end device when the first queue among the multiple queues of the receiving end device meets a preset opening index Turn on congestion control;

Wherein, the preset opening index includes: the amount of the buffer space of the receiving end device occupied by the multiple queues of the receiving end device reaches or exceeds the open back pressure waterline value of the first queue.
The device of claim 21, wherein:

The preset opening indicator further includes one or more of the following:

The first queue belongs to the N longest queues of the first priority among the multiple queues of the receiving end device, and the priority of the first queue is the first priority. When the first priority is When it is the lowest priority among the priorities of the multiple queues, the N is an integer greater than 0, and when the first priority is the non-lowest priority among the priorities of the multiple queues, Said N is an integer greater than or equal to 0;

The length of the first queue reaches or exceeds a preset length threshold.
The device according to claim 21 or 22, wherein:

The opening module is configured to send an opening control message to the sending end device, where the opening control message instructs the sending end device to enable congestion control on the target queue.
The device according to any one of claims 21 to 23, wherein the device further comprises:

The closing module is configured to notify the sending end device to close the congestion control of the target queue of the sending end device when the first queue meets a preset closing index;

Wherein, the preset closing index includes: the amount of the buffer space of the receiving end device occupied by the multiple queues of the receiving end device is less than the closing back pressure waterline value of the first queue.
The device of claim 24, wherein:

The preset closing indicator also includes one or more of the following:

The duration of the multiple queues of the receiving end device occupying the buffer space of the receiving end device less than the closing back pressure waterline value of the first queue reaches or exceeds a preset duration threshold;

The first queue does not belong to the N longest queues of the first priority among the multiple queues of the receiving end device, and the priority of the first queue is the first priority. When the first priority is When the priority is the lowest priority among the priorities of the multiple queues, the N is an integer greater than 0, and when the first priority is the non-lowest priority among the priorities of the multiple queues, The N is an integer greater than or equal to 0;

The length of the first queue is less than a preset length threshold.
The device according to any one of claims 21 to 25, wherein the device further comprises:

An allocation module, configured to allocate a sending credit value to the target queue;

The sending module is configured to send a message containing the sending credit value to the sending end device, so that the sending end device sends the message in the target queue to the receiving end device according to the sending credit value.
The device according to claim 26, wherein the device further comprises:

The coefficient adjustment module is configured to adjust the distribution coefficient of the send credit value of the target queue according to the amount of the buffer space occupied by the multiple queues of the receiving end device. The multiple queues of the end device are negatively related to the occupancy of the buffer space of the receiving end device.
The apparatus according to any one of claims 21 to 27, wherein each of the multiple queues of the receiving end device has a back pressure waterline value, and queues of different priorities have different back pressure waterline values. , The device further includes:

The waterline adjustment module is configured to adjust the receiving end device when the amount of the buffer space occupied by the receiving end device by the multiple queues of the receiving end device reaches or exceeds the minimum waterline threshold of the receiving end device The back pressure waterline value of each priority queue.
The device of claim 28, wherein:

The back pressure waterline value includes an open back pressure waterline value and a closed back pressure waterline value, and the waterline adjustment module is used for:

For any priority queue:

According to the maximum watermark threshold of the receiving end device and the minimum watermark threshold of the receiving end device, all priority queues below any one of the priority levels occupy the buffer space of the receiving end device Adjusting the backpressure waterline value of the queue of any one of the priority levels, and the occupancy of all queues of the receiving end device to the buffer space of the receiving end device;

According to the open back pressure waterline value of the queue of any priority, the closed back pressure waterline value of the queue of any priority is adjusted.
The device of claim 29, wherein:

The occupation of the buffer space of the receiving end device by the queue of any priority is inversely related to the occupation of the buffer space of the receiving end device by the queues of all priorities lower than the any priority; The opening back pressure waterline value of the queue of any priority is negatively related to the amount of the buffer space occupied by the queue of any priority on the receiving end device;

The opening backpressure waterline value of the queue of any priority is greater than or equal to the opening backpressure waterline value of all priority queues lower than the any priority.
The device according to claim 29 or 30, wherein the waterline adjustment module is configured to:

According to the maximum watermark threshold and the minimum watermark threshold, all priority queues below any one of the priority levels occupy the buffer space of the receiving end device, and all of the receiving end device The amount of buffer space occupied by the receiving end device by the queue is calculated by using a waterline value formula to calculate the opening backpressure waterline value of the queue of any priority;

Adjusting the opening back pressure waterline value of the queue of any priority according to the calculation result;

Wherein, the waterline value formula is:

The Vo_pri_k represents the opening back pressure waterline value of the queue with priority pri_k, the B_pri_i represents the occupation of the buffer space of the receiving end device by all queues with priority pri_i, and the B_total represents the receiving end The occupied amount of the buffer space of the receiving end device by all queues of the device, where the V_max represents the maximum watermark threshold, and the V_min represents the minimum watermark threshold.
The device of claim 28, wherein:

The back pressure waterline value includes an open back pressure waterline value and a closed back pressure waterline value, and the waterline adjustment module is used for:

For any priority queue:

According to the occupancy of the buffer space of the receiving end device by all the queues of any priority and the size of the buffer space, adjust the opening back pressure waterline value of the queue of any priority, and The opening backpressure waterline value of a priority queue is inversely related to the amount of buffer space occupied by all queues of any one of the priorities of the receiving end device;

According to the open back pressure waterline value of the queue of any priority, the closed back pressure waterline value of the queue of any priority is adjusted.
The device of claim 28, wherein:

The back pressure waterline value includes an open back pressure waterline value and a closed back pressure waterline value, and the waterline adjustment module is used for:

For any priority queue:

According to the average delay of the M longest queues of any priority and the average delay of all the longest queues of all the priorities of the receiving end device, adjust the opening backpressure of the queue of any priority Waterline value, where M is an integer greater than 0;

According to the open back pressure waterline value of the queue of any priority, the closed back pressure waterline value of the queue of any priority is adjusted.
The device according to any one of claims 28 to 33, wherein the device further comprises:

Threshold adjustment module, used for the maximum occupancy of the buffer space of the receiving end device by all queues of the receiving end device within the target duration is greater than the minimum waterline threshold and less than the maximum waterline threshold and the first gate Or, when the maximum occupancy is greater than the difference between the maximum watermark threshold and the second threshold, the minimum watermark threshold is adjusted.
A congestion control device, characterized in that the device includes:

The first receiving module is configured to receive a notification sent by the receiving end device to enable congestion control on the target queue of the sending end device, and each queue of the receiving end device corresponds to one or more queues in the sending end device, The target queue corresponds to the first queue of the receiving end device;

The enabling module is configured to enable congestion control on the target queue of the sending end device according to the notification of the receiving end device.
The device of claim 35, wherein:

The first receiving module is configured to receive a turn-on control message sent by the receiving end device, where the turn-on control message carries an identifier of the target queue and an enable flag bit;

The enabling module is configured to determine the target queue according to the identifier of the target queue, and enable congestion control on the target queue according to the enable flag bit.
The device according to claim 35 or 36, wherein the device further comprises:

The second receiving module is configured to receive a notification sent by the receiving end device that the target queue of the sending end device closes congestion control;

The closing module is used for closing congestion control on the target queue of the sending end device according to the notification of the receiving end device.
The device of claim 37, wherein:

The second receiving module is configured to receive a shutdown control message sent by the receiving end device, where the shutdown control message carries an identifier of the target queue and a disabling flag bit;

The closing module is configured to determine the target queue according to the identifier of the target queue, and close congestion control on the target queue according to the disable flag bit.
The device according to any one of claims 35 to 39, wherein the device further comprises:

The third receiving module is configured to receive the message containing the sending credit value sent by the receiving end device;

The sending module is configured to send the message in the target queue to the receiving end device according to the sending credit value.
A congestion control device, wherein the congestion control device includes a processor and a memory, the memory is stored with a program, and the processor is configured to execute the program stored in the memory to implement claims 1 to 14 Any of the congestion control methods described above.
A congestion control device, wherein the congestion control device includes a processor and a memory, the memory is stored with a program, and the processor is configured to execute the program stored in the memory to implement claims 15 to 20 Any of the congestion control methods described above.
A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, the congestion control method according to any one of claims 1 to 15 is realized, or , To implement the congestion control method of any one of claims 16 to 20.
A chip, characterized in that the chip includes a programmable logic circuit and/or program instructions, when the chip is running, it is used to implement the congestion control method according to any one of claims 1 to 15, or to implement the claims The congestion control method described in any one of 16 to 20.
A congestion control system, which is characterized in that it comprises: a sending end device and a receiving end device,

The receiving end device is used to monitor multiple queues of the receiving end device, and each queue of the receiving end device corresponds to one or more queues in the sending end device; When the first queue of the plurality of queues meets the preset opening index, the sending end device is notified to enable congestion control on the target queue corresponding to the first queue in the sending end device; wherein, the preset opening index Including: the amount of the buffer space of the receiving end device occupied by the multiple queues of the receiving end device reaches or exceeds the threshold value of the first queue to be turned on;

The sending end device is configured to enable congestion control on the target queue corresponding to the first queue in the sending end device according to the notification of the receiving end device.
The congestion control system according to claim 44, wherein the sending end device comprises the congestion control device according to any one of claims 35 to 39, and the receiving end device comprises any one of claims 21 to 34. The congestion control device; or, the sending end device includes the congestion control device according to claim 41, and the receiving end device includes the congestion control device according to claim 40.
A computer program product, characterized in that it comprises a computer program that, when executed by a processor, implements the congestion control method according to any one of claims 1 to 15, or implements any one of claims 16 to 20 Congestion control method.