WO2024061042A1

WO2024061042A1 - Data transmission method and data transmission system

Info

Publication number: WO2024061042A1
Application number: PCT/CN2023/118075
Authority: WO
Inventors: 叶秋红; 何子键; 林云
Original assignee: 华为技术有限公司
Priority date: 2022-09-20
Filing date: 2023-09-11
Publication date: 2024-03-28
Also published as: CN117793009A

Abstract

Provided in the embodiments of the present application are a data transmission method and system. The data transmission method provided in the present application comprises: receiving a first data stream, wherein the first data stream comprises a plurality of data packets; transmitting the first data stream to a destination node according to a preset first credit value, wherein the first credit value is used for indicating a preset size of data traffic transmitted at one time; receiving first information, wherein the first information is used for indicating that the cache capacity of a port which receives the first data stream exceeds a preset threshold value; and transmitting the first data stream to the destination node on the basis of the first information and a second credit value that is obtained from the destination node, wherein the second credit value is used for indicating the maximum data traffic which is transmitted at one time and is obtained from the destination node. The data transmission method can improve the data transmission efficiency of a network system.

Description

Data transmission method and data transmission system

This application claims priority to the Chinese patent application filed with the China Patent Office on September 20, 2022, with the application number 202211145199.8 and the application name "Data transmission method and data transmission system", the entire content of which is incorporated into this application by reference. .

Technical field

The embodiments of the present application relate to the field of network communication technology, and in particular, to a data transmission method and a data transmission system.

Background technique

In recent years, with the development of Internet services, distributed computing and other technologies, data center network (DCN), high performance computer cluster (HPC, High Performance Computing) network, network-on-chip (Noc, network-on-chip) ) and other network systems are widely used to realize data exchange between servers, between servers and terminal devices, or between various components working on the chip.

In current network systems, the source node usually pushes the data stream directly to the destination node to reduce the transmission delay of the data stream. However, when data traffic surges in the network system, this method usually causes transmission congestion at some nodes in the network system. When the congestion further spreads, it leads to head blocking or packet loss in the network system. In the existing technology, when congestion occurs, a local flow control method based on a priority flow control mechanism or a tail packet loss mechanism is usually adopted, or a source-end flow control method using an advance congestion notification mechanism is used. However, the flow control method provided by the existing technology does not distinguish between the data flow that causes congestion and the non-congested data flow, which cannot guarantee the low delay of the non-congested data flow; in addition, the local flow control method is also easy to cause Packet loss problem, the source-side flow control method cannot maintain high throughput when the network system is overloaded. Therefore, how to balance the data flow in the network system to ensure the data transmission efficiency of the network system has become a problem that needs to be solved.

Contents of the invention

The data transmission method and system provided by this application can improve the data transmission efficiency of the network system. In order to achieve the above purpose, this application adopts the following technical solutions:

In a first aspect, embodiments of the present application provide a data transmission method. The data transmission method includes: receiving a first data stream, where the first data stream includes a plurality of data packets; and sending data to a destination according to a preset first credit value. The node transmits the first data stream, and the first credit value is used to indicate a preset data traffic size transmitted at one time; and receives first information, the first information is used to indicate the number of the port that receives the first data stream. The cache capacity exceeds a preset threshold; based on the first information and a second credit value obtained from the destination node, the first data stream is transmitted to the destination node, where the second credit value is used to indicate The maximum data traffic obtained from the destination node for one transmission.

In the embodiment of this application, the source node used to send the first data stream may maintain two credit counters. Among the two credit counters, the value of one credit counter (denoted as credit1) (that is, the first credit value ) is preset by the system and is used to indicate the preset amount of data traffic transmitted at one time. The value of a credit counter (recorded as credit2) (that is, the second credit value obtained from the destination node) is based on the value obtained from the destination node. Set by the credit obtained by the node, it is used to indicate the maximum data flow obtained from the destination node for one transmission. The source node can send data streams based on credit1 and credit2. Among them, the data flow sent to the destination node based on the count of credit1 is an uncontrolled data flow, and the data flow sent to the destination node based on the count of credit2 is a controlled data flow.

In the embodiment of this application, transmitting the data stream to the destination node through the first credit value may include: the source node transmits request information to the destination node to obtain the credit value of the first data stream, but does not receive it within the preset time period. In response to feedback from the destination node, the source node directly transmits the first data stream to the destination node based on the first credit value preset in the network system.

In this embodiment of the present application, the first data stream sent by the source node does not carry an identifier used to distinguish whether the data traffic size is obtained from the destination node. In a possible implementation manner, the first data stream sent by the source node may respectively carry an identifier used to distinguish whether the data traffic size is obtained from the destination node. For example, a bit can be set in the packet header of the data stream being sent. "0" means sending with the first credit value, "1" means sending with the second credit value, that is, "0" means not getting it from the destination or node. Credit value, "1" Indicates that the credit value has been obtained from the destination node.

According to the data transmission method and network system provided by the embodiments of the present application, when the source node fails to obtain the first data stream indicated by the destination node due to the size of the data stream sent to the destination node, congestion occurs on some nodes (that is, the node receives the third data stream from the source node). (When the cache capacity of a port of a data flow exceeds a preset threshold), the node where congestion occurs feeds back an indication to the source node through the data link connected to the source node indicating that the first data flow output by the source node has caused congestion in the network system. information. Therefore, when the network system is congested, the source node can convert the first data flow transmitted based on the first credit value (that is, the uncontrolled data flow) into the first data flow transmitted based on the second credit value (that is, the controlled data flow). Data flow), the data flow that is assigned a credit value by the destination node will not be affected in any way, and can still continue to be transmitted based on the credit value assigned by the destination node, because what causes congestion in the network system is usually those who have not obtained the amount of data traffic from the destination node. Data flow, data flow with credit value distributed through the destination node will not cause congestion in the network system, unlike the existing technology that does not distinguish between data flows that cause congestion and non-congested data flows, thus affecting the normal transmission of non-congested data flows. Compared with the data transmission method provided by the embodiments of the present application, when congestion occurs in the network system, the data stream is transmitted only by transmitting the second credit value assigned by the destination node, which means that normal transmission of non-congested data in the network system can be ensured. Therefore, the congestion of the network system can be greatly reduced while ensuring a high throughput rate of the network system; in addition, compared with directly discarding newly incoming data packets in the prior art, the data transmission method provided by the embodiment of the present application can also Ensure the accuracy of data stream transmission.

In a possible implementation, after receiving the first information, the data transmission method further includes: stopping transmitting the first data stream based on the first credit value.

By stopping the transmission of data flows based on the first credit value, that is, stopping the transmission of data flows without a credit value assigned by the destination node, it is possible to transmit only controlled data flows when congestion occurs in the network system, thereby ensuring high throughput of the network system. In this case, the congestion of the network system is greatly reduced.

In a possible implementation, the first data flow also includes indication information requesting to obtain the second credit value; based on the first information and the second credit value obtained from the destination node, Before transmitting the first data stream to the destination node, the data transmission method further includes: receiving the second credit value from the destination node.

In a possible implementation, the first information is used to indicate that the buffer capacity of the port through which the destination node receives the first data flow exceeds a preset threshold of the destination node.

In a possible implementation, transmitting the first data stream to the destination node according to a preset first credit value includes: transmitting the first data stream to the destination node through a switching node; and The first information is used to indicate that the buffer capacity of the port through which the switching node receives the first data flow exceeds the preset threshold of the switching node.

In a possible implementation, the first information is carried by the switching node in the first data stream and transmitted to the destination node.

In this embodiment of the present application, after the switching node receives the first data flow, and detects that the cache capacity of the port receiving the first data flow exceeds the preset threshold, a part of the first data flow can be stored in the cache of the switching node. Then, the first information is added to the first data stream, and the part of the first data stream added with the first information is transmitted to the destination node.

By adding the first information to the first data stream for transmission, the number of times the data stream is sent can be reduced, which is beneficial to saving bandwidth of the network system.

In a possible implementation, the receiving the first information includes: receiving a second data stream, wherein the second data stream carries the first information, the second credit value and an indication and The identifier of the first data stream corresponding to the second credit value.

In the embodiment of this application, the source node receiving the second data stream may include two situations. In the first situation, the destination node adds the first information to the second data stream and sends it to the source node; in the second situation, the switching node adds the first information to the second data stream and sends it to the source node; in the second situation, the switching node The first information is added to the second data stream and sent to the source node.

By adding the first information to the second data stream for transmission, the number of times the data stream is sent can be reduced, which is beneficial to saving bandwidth of the network system.

In a possible implementation, after stopping transmitting the first data stream based on the first credit value, the data transmission method further includes: receiving second information, the second information being used to indicate the The cache capacity of the port is lower than a preset threshold; based on the second information and the first credit value, a third data flow is transmitted to the destination node, where the third data flow includes a plurality of data packets.

When the cache capacity of the port that receives the first data stream is lower than the preset threshold, it means that the congestion in the network system is relieved. At this time, the third data stream is transmitted to the destination node through the first credit value, which can improve the throughput rate and data of the network system. transmission efficiency.

In a possible implementation, after stopping transmitting the first data stream based on the first credit value, the data transmission The method further includes: transmitting a third data stream to the destination node based on the first credit value after a preset period, where the third data stream includes a plurality of data packets.

Generally, after the buffer capacity of the port that receives the first data flow is higher than the preset threshold for a certain period of time, the node where congestion occurs can process such that the buffer capacity of the port that receives the first data flow is lower than the preset threshold. If the source node does not receive information indicating that the buffer capacity of the port receiving the first data stream is lower than the preset threshold after a preset period of time, the transmission of the indication information may be lost, or there may be a delay in sending the information by the congested node. The source node uses the first credit value to transmit the third data stream after a preset period of time, which can improve the throughput rate and data transmission efficiency of the network system.

In a possible implementation manner, the second data stream is a confirmation character message.

In a second aspect, embodiments of the present application provide a data transmission system. The data transmission system includes: multiple chips, any one of the multiple chips includes a port, and any one of the chips communicates with other chips through the port. ; The source chip among the plurality of chips is used to: receive a first data stream, where the first data stream includes a plurality of data packets; and transmit the first data stream to the destination chip according to a preset first credit value , the first credit value is used to indicate a preset data traffic size transmitted at one time; receiving first information, the first information is used to indicate that the cache capacity of the port receiving the first data flow exceeds a preset threshold ; Based on the first information and a second credit value obtained from the destination chip, transmit the first data stream to the destination chip, wherein the second credit value is used to indicate the data received from the destination chip. The maximum data traffic that the end chip obtains at one time.

In a possible implementation, after receiving the first information, the source chip is further configured to: stop transmitting the first data stream based on the first credit value.

In a possible implementation, the first data stream further includes indication information requesting to obtain a second credit value; based on the first information and the second credit value obtained from the destination chip, Before the destination chip transmits the first data stream, the source chip is further configured to: receive the second credit value from the destination chip.

In a possible implementation, the first information is used to indicate that the cache capacity of the port through which the destination chip receives the first data flow exceeds a preset threshold of the destination chip.

In a possible implementation, the plurality of chips further include a switching chip. When the source chip transmits the first data stream to the destination chip according to a preset first credit value, the source chip specifically uses In: transmitting the first data flow to the destination chip through the switching chip; and the first information is used to indicate that the buffer capacity of the port of the switching chip receiving the first data flow exceeds the switching capacity. The preset threshold of the chip.

In a possible implementation, the switching chip is configured to: receive the first data stream; add the first information to the first data stream and transmit it to the destination chip.

In a possible implementation, the destination chip is specifically configured to: receive the first data stream; and generate a second data stream based on the first information and the indication information. The second data stream The first information, the second credit value, and the identifier used to indicate the first data stream corresponding to the second credit value are carried in the data stream; and the second data stream is sent to the source chip.

In a possible implementation, when the first information is used to indicate that the buffer capacity of the port through which the switching chip receives the first data flow exceeds the preset threshold of the switching chip, the purpose The end chip is specifically configured to: generate a second data stream based on the indication information, the second data stream carries the second credit value, and transmit the second data stream to the source chip through the switching chip. Data flow; the switching chip is specifically configured to: add the first information to the second data flow, and transmit the second data flow to the source chip.

In a possible implementation, after stopping transmitting the first data stream based on the first credit value, the source chip is further configured to: receive second information, where the second information is used to indicate the The cache capacity of the port is lower than a preset threshold; based on the second information and the first credit value, a third data stream is transmitted to the destination chip, where the third data stream includes a plurality of data packets.

In a possible implementation, after stopping transmitting the first data stream based on the first credit value, the source chip is further configured to: after a preset period of time, based on the first credit value, The destination chip transmits a third data stream, and the third data stream includes a plurality of data packets.

In a third aspect, embodiments of the present application provide a device, which may include: a first receiving unit, configured to receive a first data stream, where the first data stream includes a plurality of data packets; and a first sending unit, configured to Transmit the first data stream to the destination node according to a preset first credit value, where the first credit value is used to indicate the preset amount of data traffic transmitted at one time; the second receiving unit is used to receive the first information, The first information is used to indicate that the buffer capacity of the port receiving the first data flow exceeds a preset threshold; the second The transmitting unit transmits the first data stream to the destination node based on the first information and a second credit value obtained from the destination node, wherein the second credit value is used to indicate the data flow from the destination node. The maximum amount of data traffic a node gets transmitted at one time.

In a possible implementation, the device further includes: a transmission stopping unit configured to stop transmitting the first data stream based on the first credit value.

In a possible implementation, the first data stream further includes indication information requesting to obtain the second credit value.

In a possible implementation, the first information is used to indicate that the buffer capacity of the port through which the destination node receives the first data flow exceeds a preset threshold of the destination node. In a possible implementation, the first information is used to indicate that the buffer capacity of the port through which the destination node receives the first data flow exceeds a preset threshold of the destination node.

In a possible implementation manner, the second receiving unit is specifically configured to: receive a second data stream, where the second data stream carries the first information, the second credit value, and an identifier for indicating the first data stream corresponding to the second credit value.

In a possible implementation, after stopping transmitting the first data stream based on the first credit value, the device further includes: a third receiving unit configured to receive second information, the second information used to indicate that the buffer capacity of the port is lower than a preset threshold; a third sending unit used to transmit a third data stream to the destination node based on the second information and the first credit value, the third Three data streams include multiple data packets.

In a possible implementation, after stopping transmitting the first data stream based on the first credit value, the device further includes: a third sending unit configured to transmit the first data stream based on the first credit value after a preset period of time. The credit value is used to transmit a third data stream to the destination node, where the third data stream includes a plurality of data packets.

In a fourth aspect, embodiments of the present application provide a chip, which includes a processor, a memory, a cache, and a port; wherein the port is controlled by the processor and is used to communicate with other chips outside the chip. , to receive data streams from other chips or transmit data streams to other chips, and store the received data streams into the cache; the processor is used to execute program instructions in the memory to implement the first aspect Described data transmission method.

In a fifth aspect, embodiments of the present application provide a computer-readable storage medium. The computer-readable storage medium stores a computer program. When executed by the controller, the computer program is used to implement the data as described in the first aspect. Transmission method.

In a sixth aspect, embodiments of the present application provide a computer program product, which is used to implement the data transmission method described in the first aspect when the computer program product is executed by a controller.

It should be understood that the second to sixth aspects of the present application are consistent with the technical solution of the first aspect of the present application, and the beneficial effects achieved by each aspect and corresponding feasible implementations are similar, and will not be described again.

Description of the drawings

In order to explain the technical solutions of the embodiments of the present application more clearly, the drawings needed to be used in the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. , for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without exerting creative labor.

Figure 1 is a schematic structural diagram of the prior art used to alleviate node congestion provided by the embodiment of the present application;

FIG2 is a schematic diagram of an architecture of a network system provided in an embodiment of the present application;

Figure 3 is a schematic diagram of the node structure of the network system shown in Figure 2 provided by an embodiment of the present application;

Figures 4A to 4C are schematic diagrams of scenarios in which the source node sends data streams based on credit1 and credit2 provided by the embodiment of the present application;

Figure 5 is a schematic diagram of the interaction flow between nodes in the network system shown in Figure 2 provided by an embodiment of the present application;

Figures 6A to 6B are schematic diagrams of the interaction process shown in Figure 5 provided by the embodiment of the present application;

Figure 7 is another schematic diagram of the interaction process between nodes in the network system shown in Figure 2 provided by the embodiment of the present application;

FIG8 is a schematic diagram of a scenario of the interaction process shown in FIG7 provided in an embodiment of the present application;

Figure 9 is another schematic diagram of the interaction process between nodes in the network system shown in Figure 2 provided by the embodiment of the present application;

Figure 10 is a schematic diagram of the interaction process shown in Figure 8 provided by the embodiment of the present application;

Figure 11 is another schematic diagram of the interaction process between nodes in the network system shown in Figure 2 provided by the embodiment of the present application;

Figure 12 is a flow chart of the data transmission method provided by the embodiment of the present application;

Figure 13 is a schematic structural diagram of the device provided by the embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of this application.

"First", "second" and similar words mentioned herein do not indicate any order, quantity or importance, but are only used to distinguish different components. Likewise, similar words such as "a" or "one" do not indicate a quantitative limit, but rather indicate the presence of at least one.

In the embodiments of this application, words such as "exemplary" or "for example" are used to express examples, illustrations or illustrations. Any embodiment or design described as "exemplary" or "such as" in the embodiments of the present application is not to be construed as preferred or advantageous over other embodiments or designs. Rather, use of the words "exemplary" or "such as" is intended to present the concept in a concrete manner. In the description of the embodiments of this application, unless otherwise specified, the meaning of “plurality” refers to two or more.

In existing network systems, a data transmission method is usually adopted in which the source node directly pushes the data stream to the destination node to reduce the transmission delay of the data stream. When this data transmission method is used, when data traffic surges in the network system, this method usually causes transmission congestion at some nodes in the network system. When a network node is congested, local flow control or source flow control is usually used to alleviate the congestion in the existing technology. As shown in Figure 1. Figure 1 illustrates local flow control as a way to alleviate congestion. In Figure 1, a source node S1, a destination node D1 and switching nodes A1, B1, B2 and C1 are shown. When local flow control is used to alleviate congestion, assuming that switching node C1 is congested, and when the cache of switching node C1 is almost full, you can choose to directly discard newly incoming data packets; or, you can also use priority-based FPC (priority-based flow control) method, that is, the switching node C1 notifies the upstream node B1 to stop sending data flows. When source-side flow control is used to alleviate congestion, assuming that switching node C1 is congested, switching node C1 can use early congestion notification (ECN, Early Congestion Notification) to send information to source node S1 to reduce the load of source node S1. Data stream transfer rate. In summary, it can be seen that the above two methods of alleviating congestion do not distinguish between data flows that cause congestion and non-congested data flows, which cannot guarantee the normal transmission of non-congested data flows; in addition, when the local flow control method selects direct When discarding newly incoming data packets, it is easy to cause packet loss problems; the source-side flow control method cannot maintain high throughput when the network system is overloaded.

Based on the shortcomings of the above existing technologies, the industry further proposes to use push-pull hybrid scheduling to transmit data traffic. In this implementation, the source node S1 can obtain the credit value (credit) of a certain data flow from the destination node D1. The credit value is used to indicate the maximum data flow obtained from the destination node D1 for one transmission. When the source node S1 receives the credit, the source node S1 sends the data stream to the destination node D1 according to the credit, which can reduce network congestion and achieve high throughput of the data network. The above credit-based data flow can also be called controlled data flow. In addition, the source node S1 can also send an uncontrolled data stream to reduce the transmission delay of the data stream. The uncontrolled data stream is specifically a data stream that has not been allocated credit by the destination node. It is usually a data stream based on the system of the source node S1. The latest data stream sent to the destination node with a preset credit. Source node S1 can send controlled data streams and uncontrolled data streams at the same time. For example, while the source node S1 sends a controlled data flow to the destination node D1 through the credit allocated by the destination node D1, it also sends an uncontrolled data flow to the destination node D1. In this data traffic transmission method, in order to increase the transmission rate, the credit preset by the system is usually higher, which usually does not consider the data throughput in the network. That is, what causes congestion at various nodes in the network is usually uncontrolled data flow. When congestion occurs at switching node C1, switching node C1 does not detect whether the congestion is caused by controlled data flow or uncontrolled data flow. As long as the data flow causing congestion is caused by source node S1, switching node C1 will send data to the source node S1. The node S1 transmits information indicating that the data flow transmitted by the source node S1 is congested, so that the source node S1 stops transmitting the data flow to the network or transmits data to the data network in a manner that reduces network bandwidth. Since the controlled data flow does not cause network node congestion, this causes the controlled data flow sent by the source node S1 to be unable to be transmitted in the network, reducing the throughput of the network system.

The data transmission method and data transmission system provided by the embodiments of the present application are based on the above-mentioned push-pull hybrid scheduling method for data traffic transmission. When the source node causes congestion in some nodes due to sending a data stream to the destination node that has not been assigned a credit value by the destination node (that is, the data stream received by the network node from the source node exceeds a preset threshold, or the cache growth rate of a port connected to the source node in the network node exceeds a preset threshold), the congested node feeds back to the source node through the data link connected to the source node an indication that the data stream output by the source node has caused congestion in the network system. Thus, when the network system is congested, the source node can only transmit to the destination node data streams that have been assigned a credit value by the destination node. Since the data streams that cause congestion in the network system are usually those that have not been assigned a credit value by the destination node, the data streams that have been assigned a credit value by the destination node will not cause congestion in the network system. This is different from the prior art that does not distinguish between congested data streams and non-congested data streams, and the source node stops transmitting data streams to the destination node, thereby affecting Compared with the normal transmission of non-congested data streams, the data transmission method provided in the embodiment of the present application, when the network system is congested, only transmits the data stream that is allocated credit value by the destination node, that is, it can ensure the normal transmission of non-congested data in the network system, so that the congestion of the network system can be greatly reduced while ensuring the high throughput of the network system; compared with the prior art in which the source node transmits data to the data network by reducing the bandwidth in the network when congestion occurs, the embodiment of the present application can continue to transmit data based on the controlled credit value obtained from the destination end, so that the bandwidth in the network does not need to be reduced, thereby ensuring the high throughput of the network system; in addition, compared with the prior art in which the newly incoming data packets are directly discarded, the data transmission method provided in the embodiment of the present application can also ensure the accuracy of data stream transmission. It should be noted that in the embodiment of the present application, the data stream that is not allocated credit by the destination node is the data stream whose data flow size is not indicated by the destination node.

The data transmission method provided by the embodiments of this application can be applied to various network systems. For example, data center network systems, high-performance computer cluster network systems, cloud network systems, or chip-packaged on-chip network systems. This embodiment of the present application takes a data center network system as an example to describe in detail the network system provided by the embodiment of the present application. Please refer to Figure 2, which is a schematic structural diagram of a network system 100 provided by an embodiment of the present application. In Figure 2, the network system 100 may be a leaf-spine (leaf-spine) network architecture. The network system 100 includes multiple leaf nodes and multiple spine nodes, and the multiple leaf nodes and multiple spine nodes are fully connected. Leaf nodes and spine nodes can also be called switching nodes or routing nodes. Figure 2 schematically shows four leaf nodes, namely node a1, node a2, node a3 and node a4. Figure 2 schematically shows two spine nodes, respectively node b1 and node b2. The downlink port of the leaf node is connected to the server that needs to exchange data traffic, and the uplink port is connected to the spine node. In the network system shown in Figure 2, data exchange between servers connected to different leaf nodes can be implemented through the spine nodes commonly connected to the different leaf nodes. For example, if the servers connected by node a1 and node a2 need to exchange data, node a1 can send the data stream of the connected server S11 to node b1, and node b1 can send it to node a2, so that node a2 can transmit the data stream to node a2. Server S21.

In this embodiment of the present application, the leaf node used by the network system 100 to send data flows may be called a source node, the leaf node used to receive data flows may be called a destination node, and the spine node used to transmit data flows may be called a switching node. . For example, when server S11 transmits data to server S21, the leaf node a1 connected to server S11 can be called the source node, and the leaf node a21 connected to server S21 can be called the destination node. It should be noted that the embodiments of this application are described using a data center network system as an example and are not used to limit the solution. For example, in other scenarios, the server can communicate with the terminal device through multiple switches. In this case, the server can be the source node and the terminal device can be the destination node; for another example, two terminal devices can communicate with each other through the server and The switch communicates, the terminal device used to send the data flow is the source node, and the device used to receive the data flow is the destination node.

In the network system 100 shown in Figure 2, the structure of each node can be as shown in Figure 3. Figure 3 is a schematic structural diagram of a node provided by an embodiment of the present application. The node can be a source node, a destination node or a switching node in the network system 100, such as the leaf node or spine node shown in Figure 2. In practical applications, the switching nodes in the network system 100 may be switches, routers, or other network devices. Figure 3 is only an example of each node. Alternatively, the node provided by the embodiment of the present application can also be any type of device, such as a chip or chipset, or a circuit board equipped with a chip or chipset, etc. This embodiment provides This is not limited. The chip or chipset or the circuit board equipped with the chip or chipset can operate under a suitable software driver. This embodiment uses a switch chip as an example to describe the structure of the node. Referring to Figure 3, the node includes a processor, memory, and ports. Alternatively, the processor, memory and ports may be integrated into one or more chips, which may be regarded as a chipset. The processor performs various functions of the node by running or executing software programs stored in the memory and calling instructions and data stored in the memory. The processor may include one or more modules, such as a central processing unit (CPU) and a network processor (NP). The network processor may be composed of an application-specific integrated circuit (application-specific integrated circuit). ASIC) or field-Programmable Gate array (FPGA) chip implementation. Memory can be used to store software programs, instructions, and data, and can be implemented by any type of volatile or non-volatile memory or a combination thereof, including, for example, static random access memory (SRAM), dynamic random access memory (SDRAM) ), one or more of double-rate synchronous dynamic random access memory (DDR), erasable programmable read-only memory (EPROM), and read-only memory (ROM). A node may include multiple ports, n is schematically shown in the figure. Among the plurality of ports, some ports are configured as input ports of the node to receive data from other nodes, and other ports are configured as output ports of the node to send data to other nodes or servers. It should be noted that the embodiment of the present application does not specifically limit the number of ports included in each node. The number of ports included in each node is set according to the needs of the scenario.

In addition, in this embodiment of the present application, the above-mentioned memory may also include a cache, which is used to store data streams transmitted by other nodes. In the first possible implementation, the cache on each node in the network system 100 is a single-point cache, that is, the cache in the node is divided into multiple cache spaces, and each cache space is dedicated to one of the ports. . For example, assuming that node a1 in the network system 100 shown in FIG. 2 includes 10 ports, the cache in node a1 can be divided into 10 cache spaces, and the cache spaces correspond to the ports one-to-one. For example, node a1 receives a data stream through port port1 and stores the received data stream in the cache space corresponding to port1. In the second possible implementation, the cache on each node in the network system 100 is a dynamic shared cache, that is, the cache in the node is divided into multiple cache spaces, and the data streams received by multiple ports on the node Both can be stored in the same cache space. For example, node a1 receives data streams through ports port1 and port2 respectively, and stores the received data in the same cache space. In this case, the capacity of each cache space and the mapping relationship between each cache space and ports can be dynamically adjusted based on the needs of the scenario. The embodiments of the present application will be described below by taking the above-mentioned single-point cache as an example.

Based on the structure of each node described in Figure 3, please continue to refer to Figure 2, which schematically shows that each node includes three ports. Node a1 includes port Pa11, port Pa12 and port Pa13, node a2 includes port Pa21, port Pa22 and port Pa23, node a3 includes port Pa31, port Pa32 and port Pa33, node a4 includes port Pa41, port Pa42 and port Pa43, node b1 It includes port Pb11, port Pb12 and port Pb13, and node b2 includes port Pb21, port Pb122 and port Pb23. In Figure 2, the port Pa11 of the node a1 is connected to the server S11 and the server S12, the port Pa12 of the node a1 is connected to the port Pb11 of the node b1, the port Pa13 of the node a1 is connected to the port Pb21 of the node b2; the port Pa21 of the node a2 Connected to server S21 and server S22, the port Pa22 of node a2 is connected to the port Pb11 of node b1, the port Pa23 of node a2 is connected to the port Pb22 of node b2; the port Pa31 of node a3 is connected to server S31 and server S32, node a3 Port Pa32 is connected to port Pb12 of node b1, port Pa33 of node a3 is connected to port Pb23 of node b2; port Pa41 of node a4 is connected to server S41 and server S42, port Pa42 of node a4 is connected to port Pb13 of node b1 Connection, the port Pa43 of node a4 is connected to the port Pb23 of node b2. It should be noted that the number of nodes, the number of ports included in each node, and the connection relationship between the ports shown in Figure 2 are schematic and are set based on the needs of the application scenario. The application examples are not specifically limited. For example, in other application scenarios, the port Pa12 of the node a1 can be connected to the port Pb11 of the node b1 and the port Pb21 of the node b2 at the same time.

Based on the network system 100 shown in Figure 2, in this embodiment of the present application, the source node used to send the data stream can maintain two credit counters. Among the two credit counters, one credit counter (denoted as credit1) The value (that is, the preset credit value) is preset by the system. It is used to indicate the preset amount of data traffic transmitted at one time. The value of a credit counter (recorded as credit2) (that is, the credit obtained from the destination node value) is set based on the credit obtained from the destination node, which is used to indicate the maximum data flow obtained from the destination node for one transmission. The source node can send data streams based on credit1 and credit2. Among them, the data flow sent to the destination node based on the count of credit1 is an uncontrolled data flow, and the data flow sent to the destination node based on the count of credit2 is a controlled data flow. Taking node a1 as the source node and node a3 as the destination node shown in Figure 2 as an example, through the scenarios shown in Figures 4A to 4C, the configuration of the value of credit1 and the value of credit2, and the source node based on these two The credit counter transmits the data stream and is described in more detail.

In the scenario shown in FIG4A , it is assumed that node a1 is currently in the initial state. In this initial state, the credit1 count in node a1 is full, assuming it is 10; the credit2 count in node a1 is 0. Node a1 receives data stream D11 from the server. Data stream D11 is a new data stream. Based on the value in credit1, node a1 transmits data stream D11 to destination node a3. The data flow of the transmitted data stream D11 corresponds to the value of credit1. Data stream D11 is sent based on credit1, that is, it is an uncontrolled data stream. In addition, the data stream D11 also carries indication information indicating the credit value of data stream D11.

After the scenario shown in Figure 4A, the count of credit1 becomes 0. The value in credit1 is filled based on the system's preset rate. After a certain period of time, it is assumed that the count of credit1 becomes 3, and node a1 obtains the credit value of data stream D11 from node a3. Assume that the credit value is also 10.

If data flow D11 still has data traffic to be transmitted in the scenario of Figure 4A, and the size of the data traffic to be transmitted is exactly equal to the data traffic indicated by the credit value obtained from node a3, then node a1 will obtain the credit value from node a3. The value is filled to credit2, at which point the count of credit2 becomes 10, as shown in the scene in Figure 4B. Then, node a1 transmits data flow D11 to destination node a3 based on the value in credit2. The data flow size of the transmitted data flow D11 corresponds to the data flow size indicated by the value of credit2.

It should be noted that in the scenarios shown in Figure 4A and Figure 4B, the value in credit1 is the same as the value in credit2. Node a1 transmits the bandwidth of data stream D11 based on the value in credit1, and node a1 is based on the value in credit2. The bandwidth of the transmission data stream D11 is the same.

If data flow D11 still has data traffic to be transmitted in the scenario of Figure 4A, and the data traffic to be transmitted is less than the data traffic indicated by the credit value obtained from node a3, assume that the credit corresponding to the data traffic of data flow D11 to be transmitted is is 2, then node a1 fills 2 credit values from the credit value obtained from node a3 to credit2, and fills the other 8 credit values from the credit value obtained from node a3 into credit1, as shown in the scene in Figure 4C . In Figure 4C, the count of credit1 becomes 10 and the count of credit2 becomes 2. Then, node a1 transmits data flow D11 to destination node a3 based on the value in credit2. The data flow size of the transmitted data flow D11 corresponds to the data flow size indicated by the value of credit2.

If node a1 receives data flow D13 from the server, data flow D13 is a new data flow, and this data flow D13 is an uncontrolled data flow. In the scenario of Figure 4B, node a1 is based on the count 3 of credit1, that is, using the data traffic size corresponding to the count 3 of credit1, it transmits data flow D13 to node a3; in the scenario of Figure 4C, node a1 is based on the count of credit1 10, that is, the data flow size corresponding to the count 10 of credit1 is used to transmit data flow D13 to node a3.

Based on the principle of the network system 100 shown in Figure 2 and the source nodes shown in Figures 4A to 4C sending data streams, in the embodiment of the present application, each node can send or receive multiple data streams, such as eight, through one of the ports. data flow. The following takes the node a1 and the node a2 in Figure 2 as the source node, the node a3 and the node a4 as the destination node, the node b1 and the node b2 as the switching node as an example, combined with the interaction process 500 shown in Figure 5, and Figure 6A to Figure 6B The application scenario shown provides a more detailed description of the data transmission method provided by the embodiment of the present application. Please refer to Figure 5. Figure 5 is an interaction process 500 between node a1, node b1 and node a3 provided by the embodiment of the present application. The interaction process 500 includes:

Step 501: Node a1 transmits data stream D11 to node b1 based on the received data stream D11 and the preset credit value 1.

In this step, node a1 can receive the data stream from the server it is connected to, and then prioritize the received data stream. The priority here refers to the data stream sent to the network first. Assume that node a1 receives multiple pieces of data. After sorting based on priority, data stream D11 is sent first. Assume that data flow D11 is a new data flow, that is, it has not communicated with the destination node to which data flow D11 needs to be transmitted, so that data flow D11 has not obtained credit value from its corresponding destination node, that is, data flow D11 is currently Uncontrolled data flow. Based on the principle of reducing the transmission delay of the data flow, node a1 directly transmits data flow D11 to port Pb11 of node b1 through port Pa12 based on credit value 1, that is, the value of credit1 shown in Figure 4A, as shown in Figure 6A. It should be noted that node a1 can generate multiple data packets for transmission based on the frame format agreed by the data transmission protocol in the network system 100. In addition to fields carrying business data, the data packets also include It includes a field indicating the destination node a3, a field indicating the identity document (ID, Identity document) of the node a1, a field indicating whether the data flow D11 is a controlled data flow, a field indicating the ID of the data flow D11, and a field indicating the port of the node a1. number field. The data frame may also include more or fewer fields, which is not specifically limited in the embodiments of this application. In a possible implementation, while node a1 transmits data flow D11 to node a3, it may also transmit data flow D12 at the same time. This data flow D12 is a controlled data flow, and this controlled data flow D12 is pre-slave from the node. a3 obtains a data flow with a credit value, which is the maximum data flow of data flow D12 transmitted by node a1 at one time.

In step 502, node b1 saves part of the data stream D11 to the cache corresponding to port Pb11, and transmits the other part of the data stream D11 to node a3. In step 503, node b1 detects that the cache capacity of port Pb11 receiving data stream D11 exceeds the preset threshold, and transmits information I1 to node a1, where the information I1 is used to indicate that the cache capacity of port Pb11 receiving data stream D11 exceeds the preset threshold.

Based on the bandwidth on the transmission path, node b1 transmits part of the data flow D11 to node a3 through port Pb12, and saves another part of the data flow D11 in the cache corresponding to port Pb11. In addition, it can be seen from the scene shown in FIG. 6A that the port Pb11 of the node b1 not only receives the data flow D11 from the node a1, but also receives the data flow D21 from the node a2. The data stream D11 received from node a1 and the data stream D21 received from node a2 are both temporarily stored in the same cache corresponding to port Pb11. Assume that data flow D21 is a data flow assigned a credit value via the destination node. The credit value of data flow D21 is obtained from node a3 by node a2 communicating with node a3 in advance. In other words, the data flow D21 is a controlled data flow. Since the traffic size transmitted by data flow D21 at one time is configured based on the credit value assigned by the destination node, it usually does not cause congestion at node b1; while data flow D11 is an uncontrolled data flow, and the traffic size transmitted at one time is usually The network system is pre-configured, and the size of the data flow is usually not fixed. In order to reduce the transmission delay of the data flow, the flow of the data flow transmitted at one time is usually large, that is, uncontrolled data flow usually causes congestion on node b1. When node b1 detects that the cache capacity corresponding to port Pb11 exceeds the preset threshold, that is, node b1 is congested, node b1 can transmit information I1 to node a1 to indicate that the cache capacity occupied by data flow D11 exceeds the preset threshold. In specific implementation, the information I1 may be a Push Congestion Notification (PCN) message, and the format of the PCN message may be set based on the data transmission protocol in the network system 100 . In specific implementation, the PCN message may include a field indicating the ID of node a1 and an indication. Field identifying the data flow D11 that caused congestion. In addition, the PCN message may also include more or fewer fields, which is not specifically limited in this embodiment of the present application.

It should be noted that the cache capacity corresponding to the port Pb11 exceeds a preset threshold. The preset threshold is, for example, 60% of the cache capacity, 80% of the cache capacity, etc. Usually, limited by the data transmission bandwidth between node b1 and node a3, the received data stream D11 cannot be sent all at once. Node b1 sends part of the data stream and caches part of the data, and then based on the data transmission bandwidth, it can The amount of data to be accommodated, the cached data stream D11 is transmitted to node a3 in one or more times.

Step 504: Node a3 sends indication information indicating the credit value 2 of data flow D11 to node a1.

In the embodiment of this application, node a1 can transmit request information requesting to obtain credit value 2 to node a3 in various ways. In the first possible implementation, node a1 can add the request information requesting the credit value 2 to the above-mentioned data flow D11 and transmit it to node a3; in the second possible implementation, node a1 can transmit the data Before or after the flow D11, the request information requesting to obtain the credit value 2 is transmitted to the node a3 independently of the data flow D11. The embodiment of the present application does not specifically limit the method of requesting the request information to obtain the credit value 2.

Optionally, the credit value 2 can be obtained by node a1 sending a request (req, request) message to node a3 and obtaining an acknowledgment character (ACK, acknowledge character) message from node a3. That is to say, in the first possible implementation method mentioned above, node a1 can add the req message to the data flow D11. The req message carries the request information to obtain the credit value 2; in the second possible implementation method mentioned above, In a possible implementation, node a1 directly transmits a req message to node a3, and the req message carries request information for obtaining credit value 2.

Step 505: Node a1 transmits data flow D11 to node a3 based on credit value 2. In this step, node a1 can also continue to transmit data flow D12 to node a3 based on the credit value of data flow D12 obtained from node a3 in advance.

After node a1 receives the information I1 from node b1, it can first detect whether the credit value 2 of the data stream D11 obtained from node a3, that is, whether the count of credit2 corresponding to the data stream D11 as shown in Figure 4A is 0. . When it is detected that the count of credit2 shown in Figure 4A is not 0, that is, the credit value 2 of the data stream D11 has been obtained from node a3, at this time, it stops based on the preset credit value (that is, based on the credit1 shown in Figure 4A count) transmits the uncontrolled data flow D11, and transmits the controlled data flow to the node a3 based on the credit value 2 obtained from the node a3. In addition, the node a1 can also transmit the controlled data flow D12, as shown in Figure 6B. The data D12 is an old data flow, that is, it has communicated with the destination node to which the data flow D12 is to be transmitted, in order to obtain a credit value from the destination node corresponding to the data flow D12, that is, the data flow D12 is currently a controlled data flow. The credit value of data flow D12 is used to indicate the cache capacity allocated by node a3 to data flow D12. The credit value of data flow D12 is obtained by node a1 from node a3 in advance. It should be noted that the data flow D11 and data flow D12 as mentioned above are independent data flows. If both data flow D11 and data flow D12 need to become controlled data flows, node a1 needs to apply to node a3 for data flow D11. For the corresponding credit value, it is also necessary to apply to node a3 for the credit value corresponding to data flow D12. Therefore, when node a3 sends the credit value of each data flow to node a1, it also needs to carry a data flow identifier indicating the data flow corresponding to the credit value. In the scenario shown in Figure 6B, node a1 transmits data flow D11 to port Pb11 of node b1 through port Pa12 based on credit value 2. Node b1 forwards data flow D11 to port Pb11 of node b1 through port Pb12 to realize node a1 transmits data flow D11. to node a3. In addition, node a1 transmits data stream D12 to port Pb21 of node b2 through port Pa12, and node b2 forwards it to port Pa33 of node a3 through port Pb23, so that node a1 transmits data stream D12 to node a3. It can be seen from the interaction process 500 shown in Figure 5 that when node a1 sends data flow D11 with no credit value to node b1, causing node b1 to send congestion, node b1 can transmit to node a1 a message indicating the reception of data flow D11. The information I1 that the cache capacity of the port Pb11 exceeds the preset threshold is information indicating that congestion occurs in the data flow D11. Therefore, when the network system is congested, node a1 can stop transmitting the data stream D11 based on the credit value credit1 regardless of whether there is a remaining count of the credit value 1 (that is, whether the count in credit1 shown in Figure 4A is zero). The data flow D12 allocated with a credit value and the data flow D11 allocated with a credit value are transmitted to the network system 100. Since the data traffic that is not allocated a credit value usually causes congestion in the network system, the data traffic allocated with a credit value is not It will cause congestion in the network system. Compared with the existing technology that does not distinguish between data flows that cause congestion and non-congested data flows, which affects the normal transmission of non-congested data flows, the network system 100 provided by the embodiment of the present application, when the network When the system is congested, only the data flows assigned credit values are transmitted, which ensures the normal transmission of non-congested data flows in the network system, thus greatly reducing the congestion of the network system while ensuring a high throughput rate of the network system; in addition, , Compared with the existing technology that directly discards newly incoming data packets, the data transmission method provided by the embodiment of the present application can also ensure the accuracy of data stream transmission.

FIG5 shows the process of converting the first data stream from an uncontrolled data stream to a controlled data stream, so that node a1 can continue to transmit the controlled data stream D11 to node a3. It should be noted that the above step 504 can also occur before the above step 503. That is, node a1 may first receive the indication information indicating the credit value 2 of data stream D11 from node a3, and then receive the information I1 sent by node b1; or it can be said that node a3 first sends the indication information indicating the credit value 2 of data stream D11 to node a1, and node b1 then sends the information I1 to node a1. In addition, in a possible implementation of the embodiment of the present application, when node a1 first receives the information I1 sent by node b1, but does not receive the indication information indicating the credit value 2 of data stream D11 sent by node a3, in this case, as shown in FIG4A , the count of credit2 corresponding to data stream D11 is 0, and at this time node a1 may temporarily stop transmitting data stream D11 to node a3.

In addition, the interaction process 100 shown in Figure 5 schematically shows that node b1 directly transmits information I1 to node a1 when detecting congestion on port Pb11. In the second possible implementation, when node a1 transmits data flow D11 to node a3 through node b1, the above information I1 can be added by node b1 to the req message in data flow D11 and transmitted to node a3. After a3 generates an ACK message based on the req message, it can add the information I1 to the ACK message, and transmit the information I1 and the indication information indicating the credit value 2 of the data flow D11 to the node a1 together. More specifically, the above information I1 can be added to the req message in the data flow D11 by the node b1. In addition, in the three possible implementations, when node a3 transmits an ACK message to node a1 through node b1, the above information I1 can also be directly added to the ACK message, so node b1 will add information I1 and indication data The message indicating the credit value 2 of flow D11 is transmitted to node a1. The following still takes the node a1 and the node a2 in Figure 2 as the source node, the node a3 and the node a4 as the destination node, and the node b1 and the node b2 as the switching node as an example. Through the interaction process shown in Figure 7 and Figure 8, the above-mentioned The second possible implementation method and the third possible implementation method are described in more detail.

Please refer to Figure 7. Figure 7 is an interaction process 700 between node a1, node b1 and node a3 provided by the embodiment of the present application. Figure 8 is a schematic diagram of an application scenario of the interaction process shown in Figure 7. The interaction process 700 includes:

Step 701: Node a1 transmits data stream D11 to node b1 based on the received data stream D11 and the preset credit value 1. Data flow D11 includes multiple data packets that need to be sent, and also includes a req message, which carries request information for obtaining credit value 2. In this step, the req message may be set based on the data transmission protocol in the network system 100 . In specific implementation, the req message may include a field indicating the destination node a3, a field indicating the ID of the node a1, a field indicating the ID of the data flow D11, and a field indicating obtaining a credit value, etc. The req message may also include more or fewer fields, which is not specifically limited in the embodiments of this application. It should be noted that an empty field can also be set in the req message for switching node b1 to add information I1 to it.

Step 702: Node b1 saves part of the data flow D11 into the cache corresponding to port Pb11.

The specific implementation of steps 701 to 702 is the same as steps 501 to 502 shown in Figure 5. For details, refer to the relevant descriptions of steps 501 and 502, which will not be described again.

Step 703: Node b1 detects that the cache capacity of port Pb11 that receives data flow D11 exceeds the preset threshold, extracts the req message from data flow D11, and adds information I1 to the req message. Step 704: Transmit the data stream D11 with the information I1 added to node a3. The implementation of node b1 detecting that the buffer capacity of port Pb11 that receives data flow D11 exceeds the preset threshold is the same as that in step 503 shown in Figure 5 , node b1 detects that the buffer capacity of port Pb11 that receives data flow D11 exceeds the preset threshold. The implementation methods are the same. For details, refer to the relevant descriptions in step 503 and will not be described again. When node b1 detects that the cache capacity of port Pb11 exceeds the preset threshold, information I1 can be added to the req message. The position where the information I1 is added to the req message can be pre-agreed based on the data transmission protocol. For example, the information I1 may include a field indicating the ID of the node a1, a field indicating the ID of the data flow D11, and a field indicating PCN information. For example, it is agreed that the ID of node a1 is represented by sixteen bits, the ID of data stream D11 is represented by sixteen bits, and the PCN information is represented by one bit. When the field indicating the PCN information is 1, it indicates that the buffer capacity of the port Pb11 that receives the data flow D11 exceeds the preset threshold. Therefore, the req message carries information I1. Node b1 transmits data stream D11 carrying information I1 to node a1.

Step 705: Based on the req message, node a3 generates an ACK message carrying information I1 and indication information indicating the credit value 2 of data flow D11, and transmits the ACK message to node b1. In this step, after receiving the req message, node a3 parses out each field in the req message and obtains information I1 and information requesting the credit value 2 of data flow D11. Then, node a3 generates an ACK message, which may include a field indicating the ID of node a1, a field indicating the ID of node a3, a field indicating the ID of data flow D11, and a field indicating the credit value of data flow D11. As well as fields carrying information I1, etc. It can be understood that the ACK message may also include more or fewer fields, which is not specifically limited in this embodiment of the present application. The fields included in this information I1 are the same as the fields included in the information I1 described in step 704.

Step 706: Node b1 transmits the ACK message to node a1.

Step 707: Node a1 transmits data flow D11 to node a3 based on the ACK message and credit value 2. In this step, node a1 After receiving the ACK message, information indicating the credit value 2 of the data flow D11 and information indicating that the buffer capacity of the port Pb11 receiving the data flow D11 exceeds the preset threshold are obtained from the ACK message. Therefore, node a1 uses the credit value 2 assigned by node a3 to transmit the data flow D11 to node a3 through node b2.

As shown above, through Figures 7 and 8, information I1 can be added to the req message in the data stream D11. After node a3 generates an ACK message based on the req message, the information I1 is added to the ACK message and transmitted to node a1. Next, through Figures 9 and 10, it is introduced that information I1 is directly added to the ACK message, so that node b1 transmits the ACK message with information I1 added to node a1. Please refer to Figure 9, which is an interaction process 900 between nodes a1, node b1 and node a3 provided in an embodiment of the present application, and Figure 10 is a schematic diagram of an application scenario of the interaction process shown in Figure 9, and the interaction process 900 includes:

Step 901: Node a1 transmits data stream D11 to node b1 based on the received data stream D11 and the preset credit value 1. Data flow D11 includes multiple data packets that need to be sent, and also includes a req message, which carries request information for obtaining credit value 2.

Step 902: Node b1 saves part of the data flow D11 into the cache corresponding to port Pb11, and transmits another part of the data flow D11 to node a3.

The specific implementation of steps 901 to 902 is the same as steps 501 to 502 shown in Figure 5. For details, refer to the relevant descriptions of step 501 and step 502, and will not be repeated; the content of the req message is the same as that in step 701 shown in Figure 7. The contents of the req messages shown are the same. For details, refer to the relevant description of step 701 and will not be described again.

Step 903, node a3 generates an ACK message based on the req message in the data stream D11, and transmits the ACK message to node b1. The ACK message may include a field indicating the ID of node a1, a field indicating the ID of node a3, a field indicating the ID of data stream D11, and a field indicating the credit value of data stream D11. The generated ACK message does not carry information I1. Among them, some empty fields can also be set in the ACK message for node b1 to add information I1 to the ACK message.

Step 904: Node b1 detects that the buffer capacity of port pb11 that receives data flow D11 exceeds a preset threshold, adds information I1 to the ACK message, and transmits the ACK message to node a1. After this step, in addition to the fields described in step 903, the ACK message also includes a field carrying information I1.

Step 905: Node a1 transmits data flow D11 to node a3 based on the ACK message and credit value 2. The specific implementation of this step is the same as the specific implementation of step 707 shown in Figure 7. Refer to the description of step 707, which will not be described again.

It can be seen from the embodiment shown in Figure 9 that the difference from the embodiment shown in Figure 7 is that when node b1 forwards the ACK message, the information I1 is added to the ACK message. Therefore, the network system 100 provided by the embodiment of the present application, when the switching node in the network system 100 is congested, can notify the source node corresponding to the congested data flow through multiple methods or more opportunities, thereby further improving the efficiency of the network system 100. Efficiency of network system 100.

The above embodiments shown in Figures 5 to 10 illustrate that when a switching node is congested, information indicating congestion is transmitted to the source node corresponding to the congested data flow. In other possible implementations, the destination node also Congestion may occur. In this case, the destination node can also transmit congestion information to the source node corresponding to the congested data flow by directly transmitting congestion information or through an ACK message. Please continue to refer to Figure 11. Figure 11 is an interaction process 1100 between node a1, node b1 and node a3 provided by the embodiment of the present application. The interaction process 1100 includes:

Step 1101: Node a1 transmits data stream D11 to node b1 based on the received data stream D11 and the preset credit value 1.

Step 1102: Node b1 transmits data stream D11 to node a3.

Step 1103: Node a3 saves the data stream D11 into the cache. Step 1104: Node a3 detects that the buffer capacity of port Pa32 that receives data stream D11 exceeds a preset threshold, and transmits information I1 to node b1. In this step, node a3 detects the cache capacity corresponding to the port used to receive data flow D11 (for example, port Pa32 shown in Figure 2). When the cache capacity exceeds the preset threshold, it means that node a3 is congested and causes The congested data flow is data flow D11. Thus, node a3 can transmit information I1 to node b1. The information I1 may be a PCN message, and the PCN message may include a field indicating the ID of the node a1 and a field indicating the ID of the data flow D11 that causes congestion. In addition, the PCN message may also include more or fewer fields, which is not specifically limited in this embodiment of the present application.

Step 1105, node b1 transmits the information I1 to node a1.

Step 1106: Node a1 transmits data stream D11 to node a3 based on the information I1 and the credit value 2 obtained in advance from node a3.

Figure 11 exemplarily shows that when the buffer capacity of the port Pb32 of the destination node a3 that receives the data flow D11 exceeds the preset threshold, the Connect to node a1 to transmit information I1. In other possible implementations, node a3 can also add information I1 to the ACK message and transmit it to node a1 through the ACK message when the buffer capacity of port Pb32 that receives data flow D11 exceeds the preset threshold. In this implementation, node a3 needs to receive a req message from node a1 before detecting that the cache capacity of port Pb32 exceeds the preset threshold. The req message is used to request certain data from node a1 to be sent to node a3. The credit value of the data flow; then, node a3 generates an ACK message based on the req message, adds information I1 to the ACK message, and transmits it to node a1.

In the above embodiments shown in Figures 5 to 11, it is schematically shown that when the switching node b1 or the destination node a3 causes node congestion due to receiving the data flow D11 without assigned credit value, the data flow D11 is sent to the The node a1 transmits information indicating congestion, so that the node a1 stops sending the data flow D11 into the network system 100 or only sends the data flow D11 assigned a credit value. Based on the interaction process shown in Figure 5, Figure 7, Figure 9 and Figure 11 above, in a possible implementation manner of the embodiment of this application, on the basis of the above step 505, step 707, step 905 or step 1106, when the node After a1 stops sending data flows with unallocated information values for a preset time, if it has not received information indicating that the congestion is relieved, that is, it has not received information indicating that the buffer capacity of the port receiving data flow D11 is lower than the preset threshold, you can A new data stream, such as data stream D13, is transmitted to the network system 100 based on the preset credit value 1 (that is, the count of credit1 shown in FIG. 4A).

In a possible implementation, based on the above step 505, step 707, step 905 or step 1106, when node a1 receives the indication information I2 sent by node b1 or node a3, the indication information I2 is used to indicate The buffer capacity of the port receiving data flow D11 is lower than the preset threshold, and a new data flow, such as data flow D13, can be transmitted to the network system 100 based on the preset credit value 1 (that is, the count of credit1 shown in FIG. 4A). . Among them, node b1 or node a3 can send the indication information I2 separately; or, node b1 adds the indication information I2 to the req message, and the node a3 generates an ACK message with the information I2 added based on the req message, and transmits it to node a1 ; Alternatively, node b1 or node a3 directly adds the indication information I2 to the ACK message and transmits it to node a1. Among them, the format of each message, the format of the information I2 and the way in which the information I2 is added to each message are the same as the format of each message, the format of the information I1 and the addition of the information I1 to each of the above embodiments. The adding method in the message is similar. Please refer to the relevant description for details and will not be repeated again.

It should be noted that in the embodiment of the present application, nodes a1 and node a2 shown in Figure 2 are used as source nodes, nodes b1 and node b2 are switching nodes, and nodes a3 and node a4 are used as destination nodes, which are described through specific scenarios. of. It can be understood that in other scenarios, node a3 can also be a source node, and node a1 can also be a destination node. The embodiment of this application does not specifically limit the source node, switching node, and destination node. In addition, in each embodiment of the present application, a data flow D11 and a data flow D12 with a credit value allocated through the destination node are used as examples for description. It can be understood that in the network system 100, no credit is assigned by the destination node. The value data stream can include multiple data streams, and the data streams that are assigned credit values by the destination node can also include multiple data streams. The multiple data streams that are not assigned credit values by the destination node can come from the same source node, or they can come from different sources. node; the multiple data flows that are assigned credit values through the destination node can also come from the same source node, or they can come from different source nodes; the multiple data flows that are not assigned credit values by the destination node may partially cause the switching node or destination Node congestion may also completely cause congestion on the switching node or destination node, depending on the specific scenario. Thirdly, the embodiments of this application are described by taking the switching node to include only one layer as an example. In other application scenarios, the source node and the destination node may include multiple layers of switching nodes, and the source node may communicate with the destination through multiple switching nodes. Node communication, any one of the multiple switching nodes may be congested due to receiving data flows without assigned credit values.

Based on the same inventive concept, the embodiment of the present application provides a data transmission method 1200. The data transmission method 1200 is applied to the source node in any network system. For example, it can be applied to the node a1 and the node in the data center network shown in Figure 2. a2, node a3 or node a4, at this time any node among node a1, node a2, node a3 or node a4 is the source node. The data transmission method 1200 includes the following steps: Step 1201, receive a first data stream, the first data stream includes multiple data packets; Step 1202, transmit the first data stream to the destination node according to the preset first credit value , the first credit value is used to indicate the preset data traffic size transmitted at one time; Step 1203, receive first information, the first information is used to indicate that the buffer capacity of the port receiving the first data flow exceeds Preset threshold; step 1204, transmit the first data stream to the destination node based on the first information and the second credit value obtained from the destination node, where the second credit value is used to indicate The maximum data traffic obtained from the destination node for one transmission.

In a possible implementation, after receiving the first information, the data transmission method 1200 further includes: stopping transmission of the first data stream based on the first credit value.

In a possible implementation, the first data flow further includes indication information requesting to obtain the second credit value; based on the first information and the second credit value obtained from the destination node, the Before the destination node transmits the first data stream, the The data transmission method 1200 further includes: receiving the second credit value from the destination node.

In a possible implementation, after stopping transmitting the first data stream based on the first credit value, the data transmission method 1200 further includes: receiving second information, the second information being used to indicate the The cache capacity of the port is lower than a preset threshold; based on the second information and the first credit value, a third data stream is transmitted to the destination node, where the third data stream includes a plurality of data packets.

In a possible implementation, after stopping transmitting the first data stream based on the first credit value, the data transmission method 1200 further includes: after a preset period of time, based on the first credit value, The destination node transmits a third data stream, where the third data stream includes a plurality of data packets.

It can be understood that, in order to implement the above functions, the source node (such as node a1, node a2, node a3 or node a4 shown in Figure 2) includes hardware and/or software modules corresponding to each function. In conjunction with the steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is performed by hardware or computer software driving the hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions in conjunction with the embodiments for each specific application, but such implementations should not be considered to be beyond the scope of this application.

This embodiment can divide the source node into functional modules according to the above method examples. For example, different functional modules can be divided corresponding to each function, or two or more functions can be integrated into one processing module. The above integrated modules can be implemented in the form of hardware. It should be noted that the division of modules in this embodiment is schematic and is only a logical function division. In actual implementation, there may be other division methods.

In the case of dividing each functional module corresponding to each function, Figure 13 shows a possible schematic diagram of the device 1300 involved in the above embodiment. The previously mentioned device can be further expanded. For example, the device corresponding to Figure 13 Device 1300 may be a software device running on the source node, or device 1300 may be a combined software and hardware device embedded in the source node. As shown in Figure 13, the device 1300 may include: a first receiving unit 1301, used to receive a first data stream, where the first data stream includes a plurality of data packets; a first sending unit 1302, used to send data according to a preset The first credit value is used to transmit the first data stream to the destination node. The first credit value is used to indicate the preset data traffic size transmitted at one time; the second receiving unit 1303 is used to receive the first information. A piece of information is used to indicate that the buffer capacity of the port that receives the first data flow exceeds a preset threshold; the second sending unit 1304 is used to send a message to the port based on the first information and the second credit value obtained from the destination node. The destination node transmits the first data flow, wherein the second credit value is used to indicate a maximum data flow transmitted at one time obtained from the destination node.

In a possible implementation, the device 1300 further includes: a transmission stopping unit (not shown in the figure), configured to stop transmitting the first data stream based on the first credit value.

In a possible implementation manner, the first information is used to indicate that a cache capacity of a port of the destination node that receives the first data flow exceeds a preset threshold of the destination node.

In a possible implementation, the second receiving unit is specifically configured to: receive a second data stream, wherein the second data stream carries the first information, the second credit value and an indication An identification of the first data stream corresponding to the second credit value.

In a possible implementation, after stopping transmitting the first data stream based on the first credit value, the device 1300 further includes: a third receiving unit (not shown in the figure), configured to receive a third Two information, the second information is used to indicate the cache of the port The capacity is lower than the preset threshold; a third sending unit (not shown in the figure) is configured to transmit a third data stream to the destination node based on the second information and the first credit value, the third A data stream consists of multiple data packets.

In a possible implementation, after stopping transmitting the first data stream based on the first credit value, the device 1300 further includes: a third sending unit (not shown in the figure), configured to Assuming a period of time, based on the first credit value, a third data stream is transmitted to the destination node, where the third data stream includes a plurality of data packets.

Exemplarily, the above source node may also include at least one processor, memory and port. Among them, at least one processor can call all or part of the computer program stored in the memory to control and manage the actions of the source node. For example, it can be used to support the source node to execute the steps performed by each unit shown in Figure 13. The memory can be used to support node execution and storage of program codes and data, etc. The memory includes but is not limited to at least a part of the storage space, cache (Cache) or registers of the above-mentioned memory. At least one processor may implement or execute the various exemplary plurality of logic modules described in conjunction with the present disclosure, which may be a combination of one or more microprocessors that implement computing functions. In addition, at least one processor may also include other programmable logic devices, transistor logic devices, or discrete hardware components.

This embodiment also provides a computer-readable storage medium. Computer instructions are stored in the computer-readable storage medium. When the computer instructions are run on a computer, they cause the computer to execute the above related method steps to implement the data transmission in the above embodiment. method.

This embodiment also provides a computer program product. When the computer program product is run on a computer, it causes the computer to perform the above related steps to implement the data transmission method in the above embodiment.

Among them, the computer-readable storage medium or computer program product provided by this embodiment is used to execute the corresponding method provided above. Therefore, the beneficial effects it can achieve can be referred to the corresponding method provided above. The beneficial effects will not be repeated here.

Through the description of the above embodiments, those skilled in the art can understand that for the convenience and simplicity of description, only the division of the above functional modules is used as an example. In practical applications, the above functions can be allocated to different modules according to needs. The functional module is completed, 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, each functional unit in each embodiment of the present application can be integrated into one product, or each unit can exist physically alone, or two or more units can be integrated into one product. Corresponding to Figure 12, if the above modules are implemented in the form of software functional units and sold or used as independent products, they can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present application are essentially or contribute to the existing technology, or all or part of the technical solution can be embodied in the form of a software product, and the software product is stored in a storage medium , including several instructions to cause a device (which can be a microcontroller, a chip, etc.) or a processor to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned readable storage media include: U disk, mobile hard disk, read only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc. that can store program code. medium.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application, but not to limit it; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: The technical solutions described in the foregoing embodiments can still be modified, or some or all of the technical features can be equivalently replaced; and these modifications or substitutions do not deviate from the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present application. scope.

Claims

A data transmission method, characterized by including:

receiving a first data stream, the first data stream including a plurality of data packets;

Transmit the first data flow to the destination node according to a preset first credit value, where the first credit value is used to indicate a preset amount of data traffic transmitted at one time;

Receive first information, the first information being used to indicate that the buffer capacity of the port that receives the first data flow exceeds a preset threshold;

Transmit the first data stream to the destination node based on the first information and a second credit value obtained from the destination node, wherein the second credit value is used to indicate the credit value obtained from the destination node. The maximum amount of data transferred at one time.
The data transmission method according to claim 1, characterized in that after receiving the first information, the data transmission method further includes:

Stop transmitting the first data stream based on the first credit value.
The data transmission method according to claim 1, characterized in that, the first data flow further includes indication information requesting to obtain the second credit value; the information obtained based on the first information and from the destination node The second credit value, before transmitting the first data stream to the destination node, the data transmission method further includes:

The second credit value is received from the destination node.
The data transmission method according to any one of claims 1 to 3, characterized in that the first information is used to indicate that the buffer capacity of the port of the destination node that receives the first data flow exceeds that of the destination node. Preset threshold.
The data transmission method according to any one of claims 1 to 3, characterized in that transmitting the first data stream to the destination node according to the preset first credit value includes:

Transmitting the first data stream to the destination node through a switching node; and

The first information is used to indicate that the buffer capacity of the port through which the switching node receives the first data stream exceeds the preset threshold of the switching node.
The data transmission method according to claim 5, characterized in that:

The first information is carried by the switching node in the first data stream and transmitted to the destination node.
The data transmission method according to any one of claims 1-6, characterized in that receiving the first information includes:

A second data stream is received, wherein the second data stream carries the first information, the second credit value, and an identifier indicating the first data stream corresponding to the second credit value.
The data transmission method according to claim 2, characterized in that, after stopping transmitting the first data stream based on the first credit value, the data transmission method further includes:

Receive second information, the second information being used to indicate that the cache capacity of the port is lower than the preset threshold;

Based on the second information and the first credit value, a third data flow is transmitted to the destination node, where the third data flow includes a plurality of data packets.
The data transmission method according to claim 2, characterized in that, after stopping transmitting the first data stream based on the first credit value, the data transmission method further includes:

After a preset period of time, based on the first credit value, a third data stream is transmitted to the destination node, where the third data stream includes a plurality of data packets.
The data transmission method according to claim 7, characterized in that:

The second data stream is a confirmation character message.
A data transmission system, characterized in that the data transmission system includes: a plurality of chips, any one of the plurality of chips includes a port, and any one of the chips communicates with other chips through the port;

The source chip among the plurality of chips is used for:

receiving a first data stream, the first data stream including a plurality of data packets;

Transmit the first data stream to the destination chip according to a preset first credit value, where the first credit value is used to indicate a preset amount of data traffic transmitted at one time;

Receive first information, the first information being used to indicate that the buffer capacity of the port that receives the first data flow exceeds a preset threshold;

Based on the first information and the second credit value obtained from the destination chip, transmit the first data to the destination chip. data flow, wherein the second credit value is used to indicate the maximum data flow rate transmitted at one time obtained from the destination chip.
The data transmission system according to claim 11, characterized in that, after receiving the first information, the source chip is also used to:

Stop transmitting the first data stream based on the first credit value.
The data transmission system according to claim 11, wherein the first data stream further includes indication information requesting to obtain the second credit value; After obtaining the second credit value, before transmitting the first data stream to the destination chip, the source chip is also used to:

The second credit value is received from the destination chip.
The data transmission system according to any one of claims 11 to 13, characterized in that the first information is used to indicate that the buffer capacity of the port of the destination chip that receives the first data stream exceeds the destination port. The chip’s preset threshold.
The data transmission system according to any one of claims 11 to 13, wherein the plurality of chips further includes a switching chip, and the source chip transmits the first credit value to the destination chip according to a preset first credit value. During data streaming, the source chip is specifically used for:

Transmitting the first data stream to the destination chip through the switching chip; and

The first information is used to indicate that the cache capacity of the port of the switching chip receiving the first data flow exceeds the preset threshold of the switching chip.
The data transmission system according to claim 15, characterized in that the switching chip is used for:

receiving the first data stream;

Add the first information to the first data stream and transmit it to the destination chip.
The data transmission system according to any one of claims 11 to 16, characterized in that the destination chip is specifically used for:

receiving the first data stream;

Based on the first information and the indication information, generate a second data stream, where the second data stream carries the first information, the second credit value, and an identifier for indicating the first data stream corresponding to the second credit value;

Send the second data stream to the source chip.
The data transmission system according to claim 15, characterized in that when the first information is used to indicate that the buffer capacity of the port of the switching chip that receives the first data stream exceeds the preset value of the switching chip, When the threshold is reached, the destination chip is specifically used for:

Based on the indication information, generate a second data stream, the second data stream carries the second credit value, and transmit the second data stream to the source chip through the switching chip;

The switching chip is specifically used for:

Add the first information to the second data stream, and transmit the second data stream to the source chip.
The data transmission system according to claim 12, characterized in that, after stopping transmitting the first data stream based on the first credit value, the source chip is also used to:

Receive second information, the second information being used to indicate that the cache capacity of the port is lower than a preset threshold;

Based on the second information and the first credit value, a third data stream is transmitted to the destination chip, where the third data stream includes a plurality of data packets.
The data transmission system according to claim 12, characterized in that, after stopping transmitting the first data stream based on the first credit value, the source chip is also used to:

After a preset period of time, based on the first credit value, a third data stream is transmitted to the destination chip, where the third data stream includes a plurality of data packets.
The data transmission system according to claim 17 or 18, characterized in that,

The second data stream is a confirmation character message.