Switch, service board, network controller and network system
By establishing a communication link between the switch and the service board through the OTN protocol format and circuitry, the problem of complex data transmission in the switch is solved, and low-latency, high-efficiency AI computing data transmission is achieved.
Patent Information
- Application Number
- CN202410752198.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-11
- Publication Date
- 2025-12-12
AI Technical Summary
Existing switches are complex to process during data transmission, consume a lot of resources, have high power consumption and long latency, and cannot efficiently transmit AI computing data in artificial intelligence networks.
It adopts the Optical Transport Network (OTN) protocol format, establishes a communication link through the OTN circuit between the switch and the service board, and the switch and network controller obtain configuration information to determine the target port and send service information directly without parsing the received information.
It simplifies the switching process, reduces resource consumption and power consumption, improves data transmission efficiency, enables the establishment of low-latency and efficient communication links, and supports communication between multiple target devices.
Smart Images

Figure CN121125656A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] Embodiments of the present application relate to the technical field of communication, and particularly relate to a switch, a service board, a network controller and a network system. BACKGROUND
[0002] An artificial intelligence (AI) network or a high-performance computing cluster is mainly used for large computing tasks, such as machine learning large model training, inference and scientific problem computing. The AI network can include a switch and a plurality of computing boards connected with the switch. The computing board includes a chip for performing a computing function and a port, the port of the computing board is connected with the switch, and the computing board communicates with the switch through the port. The plurality of computing boards can transmit AI computing data through the switch. The AI network can adopt an Ethernet (ETH) technology or an Infiniband (IB) technology. However, the processing process of the switch forwarding data is relatively complex. SUMMARY
[0003] Embodiments of the present application provide a switch, a service board, a network controller and a network system, which solve the problem that the processing process of the switch forwarding data in the prior art is relatively complex.
[0004] To achieve the above object, the present application adopts the following technical solutions:
[0005] In a first aspect, a switch is provided, the switch comprising a first port and at least one second port. The first port of the switch is configured to connect a first target device, and the at least one second port of the switch is configured to correspondingly connect at least one second target device. The switch is configured to: receive service information sent by the first target device through the first port; obtain configuration information; the configuration information is used to indicate one or more second target ports of the switch for sending the service information, the service information is based on an optical transport network (OTN) protocol format, and the at least one second port includes the one or more second target ports; and in response to the configuration information, send the service information to the corresponding second target device through the one or more second target ports.
[0006] In the technical solution, the switch can determine to send the service information through one or more second target ports according to the obtained configuration information, so that the received service information does not need to be parsed. The processing of the switch is relatively simple, the resource cost is relatively small, the power consumption is relatively low, and the time delay is relatively low. In addition, the switch can receive the service information of one first target device and send the service information to one or more second target devices. That is, in one data transmission process, the switch can connect the communication links between two or more target devices. Therefore, the energy efficiency of the switch can be effectively improved. In addition, the switch uses the OTN technology to transmit data in the OTN protocol format, and the switch can establish the communication links between two or more target devices through the OTN circuit, so that the communication link establishment speed is relatively fast.
[0007] In a possible implementation of the first aspect, the switch comprises an OTN cross circuit and a plurality of OTN physical layer circuits, the plurality of OTN physical layer circuits comprise a first physical layer circuit and at least one second physical layer circuit, the first physical layer circuit is connected to the first port, and the at least one second physical layer circuit is connected to the at least one second port. The OTN cross circuit is configured to: in response to the configuration information, establish a transmission channel between the first physical layer circuit and one or more second target physical layer circuits, the at least one second physical layer circuit comprises the one or more second target physical layer circuits, and the one or more second target circuits are connected to the one or more second target ports. In the possible implementation, the transmission channel is established between the first physical layer circuit and the one or more second target physical layer circuits, so that the transmission channel is established between the first port and the one or more second target ports. In this way, the switch can send the service information to the second target devices corresponding to the one or more second target ports through the established transmission channel.
[0008] In a possible implementation of the first aspect, the bandwidth of the first port for transmitting the service information is greater than the bandwidth of the second target port for transmitting the service information. In the possible implementation, the bandwidth of the first port of the switch for transmitting the service information is greater than the bandwidth of the second target port for transmitting the service information, and the bandwidth granularity of the switch for transmitting data is relatively flexible. Therefore, more flexible and faster bandwidth switching capability can be provided.
[0009] In a possible implementation of the first aspect, the service information comprises one or more of artificial intelligence (AI) task calculation data, data service information, and communication interaction information. In the possible implementation, the type of the service information can be one or more of the AI task calculation data, the data service information, and the communication interaction information, and the application scenarios of the embodiments of the present application are relatively rich.
[0010] In a possible implementation of the first aspect, the switch is further configured to connect a network controller. The switch is specifically configured to: receive configuration information sent by the network controller. The network controller can be a standalone device. Alternatively, the network controller is arranged in a target switch connected to the switch. In the possible implementation, the network controller can be a standalone device, and the connection between the network controller and the switch is simple. Alternatively, the network controller is arranged in a target switch connected to the switch, and the network controller in the target switch can send the configuration information to the switch through a control channel. In this way, the number of devices is small, and the entire network architecture is simple.
[0011] In the second aspect, a service board is provided. The service board is configured to connect a switch. The service board is configured to: send service information to the switch, the service information being based on an optical transport network (OTN) protocol format; and send address information to the switch or a network controller, the address information including a communication address of one or more target service boards used to process the service information; the address information being used to instruct the switch to obtain configuration information according to the address information, or the address information being used to instruct the network controller to obtain the configuration information according to the address information and send the configuration information to the switch; and the configuration information being used to instruct the switch to send the service information through one or more ports.
[0012] In the technical solution, the service board sends the address information to connect the communication links between the service boards, so that the switch does not need to parse the received service information. The processing of the switch is simple, the resource cost is small, the power consumption is low, and the time delay is low. In addition, the communication links between two or more service boards can be connected in one data transmission process. Therefore, the energy efficiency of data transmission can be effectively improved. In addition, the service board and the switch use the OTN technology to transmit data based on the OTN protocol format, and the switch can establish the communication links between two or more service boards through the OTN circuit, and the communication links can be established quickly.
[0013] In a possible implementation of the second aspect, the service board includes an OTN frame adjuster and an OTN physical layer circuit. The OTN frame adjuster is configured to: convert input data into the OTN protocol format to obtain service information, and output the service information. The OTN physical layer circuit is configured to: input the service information, and send the service information to the switch. In the possible implementation, the OTN frame adjuster and the OTN physical layer circuit are arranged in the service board, so that the data is transmitted in the OTN protocol format, and the communication link between the service board and the corresponding switch is established.
[0014] In a possible implementation of the second aspect, the service information includes one or more of artificial intelligence (AI) task computing data, data service information, and communication interaction information. In the possible implementation, the type of the service information can be one or more of the AI task computing data, the data service information, and the communication interaction information, and the application scenarios of the embodiments of the present application are more abundant.
[0015] In a possible implementation of the second aspect, the service information is AI task computing data, and the service board is further configured to receive an AI computing task sent by the network controller and calculate the service information based on the AI computing task. In the possible implementation, the service board executes the AI computing task under the allocation of the network controller, so as to implement complex AI computing tasks.
[0016] In a possible implementation of the second aspect, the network controller can be a separate device. Alternatively, the network controller is arranged in a target switch connected with the switch. In the possible implementation, the network controller can be a separate device, and the connection mode between the network controller and the switch is simple. Alternatively, the network controller can be arranged in a target switch connected with the switch, and the network controller in the target switch can send the configuration information to the switch through a control channel. In this way, the number of devices is small, and the entire network architecture is simple.
[0017] In a third aspect, a network controller is provided, which is configured to connect at least one switch and a plurality of service boards. The switch includes a first port and at least one second port. The first port of the switch is configured to connect a first target device, and the at least one second port of the switch is configured to correspondingly connect at least one second target device. The plurality of service boards includes a first service board and one or more second service boards. The network controller is configured to: receive address information sent by the first service board, the address information being used to indicate a communication address of one or more second service boards used to process service information, and the service information being based on an optical transport network (OTN) protocol format; send configuration information to one or more switches in the at least one switch according to the address information, the configuration information being used to indicate one or more second target ports of the switch used to send the service information, the at least one second port including the one or more second target ports; and the configuration information being used to establish a target communication link between the switch, a first target device corresponding to the first port, and a second target device corresponding to the one or more second target ports, the target communication link being a sub-link of a communication link between the first service board and the one or more second service boards.
[0018] In the technical solution, the network controller controls the switches to connect the target communication links between the multiple target devices, so that the switches do not need to parse the received service information. The processing of the switches is simple, the resource cost is small, the power consumption is low, and the time delay is low. In addition, in one data transmission process, the communication links between two or more target devices can be connected. Therefore, the energy efficiency ratio of data transmission can be effectively improved. In addition, the service boards and the switches use OTN technology to transmit data based on the OTN protocol format, and the network controller can control the switches to establish communication links between two or more service boards through the OTN circuit, so that the communication link establishment speed is fast.
[0019] In a possible implementation of the third aspect, the network controller comprises an OTN route calculation circuit and an OTN distribution circuit. The OTN route calculation circuit is configured to receive the address information sent by the first service board, and obtain one or more switches on the communication links between the first service board and one or more second service boards according to the address information. The OTN distribution circuit is configured to send configuration information to the one or more switches. In the possible implementation, the OTN route calculation circuit and the OTN distribution circuit are deployed in the network controller, so that the communication address sent by the service board is received, and the communication links between multiple service boards are controlled to be established.
[0020] In a possible implementation of the third aspect, the service information comprises one or more of artificial intelligence (AI) task calculation data, data service information, and communication interaction information. In the possible implementation, the type of the service information can be one or more of the AI task calculation data, the data service information, and the communication interaction information, and the application scenarios of the embodiments of the present application are rich.
[0021] In a possible implementation of the third aspect, the service information is AI task calculation data, and the network controller is further configured to send corresponding AI calculation tasks to the first service board and the one or more second service boards respectively. In the possible implementation, the network controller further sends corresponding AI calculation tasks to each service board, so as to implement complex AI calculation tasks.
[0022] In a possible implementation of the third aspect, the network controller is a separate device. Alternatively, the network controller is arranged in the at least one switch. In the possible implementation, the network controller can be a separate device, and the connection between the network controller and the switch is simple. Alternatively, the network controller can be arranged in the at least one switch, and the network controller in the switch can send configuration information to the one or more switches through a control channel. In this way, the number of devices is small, and the entire network architecture is simple.
[0023] In a fourth aspect, a network system is provided, which comprises the OTN switch provided in the first aspect or any possible implementation manner of the first aspect, the service board provided in the second aspect or any possible implementation manner of the second aspect, and the network controller provided in the third aspect or any possible implementation manner of the third aspect.
[0024] In a possible implementation manner of the fourth aspect, the network system is an artificial intelligence (AI) cluster system, or a data center network system, or an OTN network system.
[0025] It can be understood that the network system provided above applies the above-mentioned switch, service board and network controller. Therefore, the beneficial effects achieved thereby can refer to the beneficial effects of the corresponding switch, service board and network controller provided above, which will not be described herein again. BRIEF DESCRIPTION OF DRAWINGS
[0026] Figure 1 A schematic diagram of a computing frame provided by an embodiment of the present application;
[0027] Figure 2 A schematic diagram of a network system provided by an embodiment of the present application;
[0028] Figure 3 A schematic diagram of a first network system provided by an embodiment of the present application;
[0029] Figure 4 A schematic diagram of a computing frame provided by an embodiment of the present application;
[0030] Figure 5 A schematic diagram of a second network system provided by an embodiment of the present application;
[0031] Figure 6 A schematic diagram of a third network system provided by an embodiment of the present application;
[0032] Figure 7 A schematic diagram of a data transmission process provided by an embodiment of the present application;
[0033] Figure 8 A schematic diagram of a service board provided by an embodiment of the present application;
[0034] Figure 9 A schematic diagram of a switch provided by an embodiment of the present application;
[0035] Figure 10 A schematic diagram of a network controller provided by an embodiment of the present application. DETAILED DESCRIPTION
[0036] It should be noted that the terms "first", "second", etc. used in the embodiments of the present application are only used for the purpose of distinguishing the same type of features, and should not be understood as indicating relative importance, quantity, order, etc.
[0037] The term "exemplary" or "for example" used in the embodiments of the present application is used to represent an example, an illustration or an exposition. Any embodiment or design scheme described as "exemplary" or "for example" in the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. In fact, the term "exemplary" or "for example" is intended to present the relevant concept in a specific manner.
[0038] The term "connection" used in the embodiments of the present application should be interpreted broadly, for example, it can refer to a direct physical connection or an indirect connection through electronic devices, such as a connection through resistors, inductors, capacitors or other electronic devices.
[0039] Before introducing the specific content of the embodiments of the present application, the application scenarios of the embodiments of the present application are introduced. The embodiments of the present application can be applied to network systems such as artificial intelligence (AI) networks, data center networks, optical transport networks (OTN) and the like. These network systems can include switches and multiple processing devices, and the multiple processing devices perform data transmission through the switches. For example, the processing devices of the data center network are usually servers based on central processing units (CPUs), and the processing devices of the AI network are usually computing boards based on general purpose graphic processing units (GPGPU). Taking the AI network as an example, a possible architecture of a computing frame including computing boards and switches is introduced. As shown in FIG. 1, the computing frame can include a top of the rack switch (TOR) 110 and multiple computing boards, and the multiple computing boards can include an AI computing board 120, an AI computing board 130 and other computing boards (not shown in the figure). Figure 1 Figure 1 The TOR 110 includes a TOR chip 111. For example, the TOR chip 111 can be an application specific integrated circuit (ASIC) dedicated to the TOR 110. Each computing board includes a plurality of AI chips. For example, the AI computing board 120 includes AI chips 121-124, and the AI computing board 130 includes AI chips 131-134. For example, the AI chips can be GPGPUs or AI acceleration chips as computing cores. The plurality of chips on each of the plurality of computing boards are connected in a mesh, a torus, or the like. Each computing board further includes a port (not shown in FIG. 1) connected to the TOR 110. The TOR 110 implements data exchange between different computing boards through the TOR chip 111. Figure 1
[0040] Embodiments of the present application provide a network system, as shown in FIG. 1, which includes a plurality of levels of switches and a plurality of processing devices. Figure 2 Figure 2 Only four levels of switches are shown, and more or fewer levels of switches can also be present in practice, which is not limited in the embodiments of the present application. The first level of switches includes a switch 11. The second level of switches includes a switch 21 and a switch 22. The third level of switches includes a switch 31, a switch 32, and a switch 33. The fourth level of switches includes a switch 41, a switch 42, a switch 43, and a switch 44. The number of switches in each level of switches can also be more or fewer, which is not limited in the embodiments of the present application. The connection relationship of the first level of switches, the second level of switches, the third level of switches, and the fourth level of switches is shown in FIG. 1. The switch 41 is connected to processing devices 211-214, the switch 42 is connected to processing devices 221-224, the switch 43 is connected to processing devices 231-234, and the switch 44 is connected to processing devices 241-244. Figure 2
[0041] For example, the processing device 211 needs to perform data transmission with the processing device 212. The processing device 211 sends the to-be-transmitted data to the switch 41, and the switch 41 forwards the to-be-transmitted data to the processing device 212. For another example, the processing device 211 needs to perform data transmission with the processing device 241. The processing device 211 sends the to-be-transmitted data to the switch 41, the switch 41 sends the to-be-transmitted data to the switch 31, and the switch 31 forwards the to-be-transmitted data to the processing device 241. The communication link between the processing device 211 and the processing device 241 can also be other links (for example, data transmission can also be performed through the second level of switches), which is not limited in the embodiments of the present application.
[0042] In one possible implementation, the network system 1000 may be a first network system. The first network system may be an AI network system that employs multiple computation frames to form a fat tree architecture.
[0043] In some examples, such as Figure 3 As shown. The first network system 1000A includes a core switch 11A as its first-level switch, spine switches 21A-22A as its second-level switch, leaf switches 31A-33A as its third-level switch, and TOR41A-44A as its fourth-level switch. TOR41A connects to computing boards 211A-214A, TOR42A connects to computing boards 221A-224A, TOR43A connects to computing boards 231A-234A, and TOR44A connects to computing boards 241A-244A. Exemplarily, the switches and computing boards in the first network system can employ Ethernet (ETH) technology or InfiniBand (IB) technology. These communication technologies are characterized by distributed forwarding. For example, as... Figure 4 As shown, with Figure 3 Taking computing board 211A as an example, computing board 211A includes computing circuit 201A, bus circuit 202A, ETH media access control (MAC) circuit 203A, ETH physical layer (PHY) circuit 204A, IB MAC circuit 205A, and IB PHY circuit 206A. Computing circuit 201A is used to execute actual AI computing tasks, obtain data packets, and generate communication requests with computing circuits on other computing boards based on computing needs. Computing circuit 201A is connected to ETH MAC circuit 203A and IB MAC circuit 205A via bus circuit 202A. ETH MAC circuit 203A or IB MAC circuit 205A receives data packets and communication requests from service circuits and converts the communication requests into communication addresses in the data packets. ETH PHY circuit 204A or IB PHY circuit 206A communicates with the corresponding TOR. For example, computing board 211A needs to transmit data with computing board 212A. Computing board 211A generates a data packet and sends it to TOR41A. This data packet includes the communication address of computing board 212A. TOR41A parses the data packet to obtain the communication address of computing board 212A, thus determining that the data packet needs to be sent to computing board 212A. In this implementation, each level of the switch needs to parse the received data packet, extract the communication address, and then forward the data packet according to the communication address. This process is relatively complex, resource-intensive, and consumes a lot of power.
[0044] In another possible implementation, the network system 1000 can be a second network system. The second network system can employ optical circuit switch (OCS) on the basis of the first network system 1000A.
[0045] In some examples, as shown in FIG. 1B, the first network system 1000A can be a network system of a data center. The first network system 1000A can include a first network system 1000A, a second network system 1000B, and a third network system 1000C. Figure 5 As shown in FIG. 1B, the first network system 1000A includes a first level switch including a core switch 11A and an OCS 12A, a second level switch including a spine switch 21A-22A and an OCS 23A, a third level switch including a leaf switch 31A-33A and an OCS 34A, and a fourth level switch including a TOR 41A-44A. The TOR 41A connects the computing boards 211A-214A, the TOR 42A connects the computing boards 221A-224A, the TOR 43A connects the computing boards 231A-234A, and the TOR 44A connects the computing boards 241A-244A. The OCS can reconfigure the sub-network to which the OCS belongs by using a micro electrical mechanical system (MEMS) or a wavelength selective switch (WSS) or the like. In this implementation, the OCS can connect the communication link between two computing boards within a preset time period, so that the OCS does not need to parse the received data packet, and the resource cost and power consumption are small. However, the OCS connects the transmission channel between two ports by a mechanical method, and the two ports are connected to two computing boards respectively. The speed of the OCS to establish the transmission channel is slow, which leads to slow reconfiguration speed of the OCS to the sub-network. In addition, the OCS can only connect the communication link between two computing boards in a data transmission process, and cannot connect the communication link between three or more computing boards. The bandwidth granularity of the OCS reconfiguration is too large (for example, fiber level or wavelength level).
[0046] In another possible implementation, the network system 1000 can be a third network system. The third network system can employ OTN technology.
[0047] In some examples, as shown in FIG. 1C, the third network system 1000C can be a network system of a data center. The third network system 1000C can include a first network system 1000A, a second network system 1000B, and a third network system 1000C. Figure 6 As shown in FIG. 1C, the first network system 1000A includes a first level switch including a core switch 11A and an OCS 12A, a second level switch including a spine switch 21A-22A and an OCS 23A, a third level switch including a leaf switch 31A-33A and an OCS 34A, and a fourth level switch including a TOR 41A-44A. The TOR 41A connects the computing boards 211A-214A, the TOR 42A connects the computing boards 221A-224A, the TOR 43A connects the computing boards 231A-234A, and the TOR 44A connects the computing boards 241A-244A. The OCS can reconfigure the sub-network to which the OCS belongs by using a micro electrical mechanical system (MEMS) or a wavelength selective switch (WSS) or the like. In this implementation, the OCS can connect the communication link between two computing boards within a preset time period, so that the OCS does not need to parse the received data packet, and the resource cost and power consumption are small. However, the OCS connects the transmission channel between two ports by a mechanical method, and the two ports are connected to two computing boards respectively. The speed of the OCS to establish the transmission channel is slow, which leads to slow reconfiguration speed of the OCS to the sub-network. In addition, the OCS can only connect the communication link between two computing boards in a data transmission process, and cannot connect the communication link between three or more computing boards. The bandwidth granularity of the OCS reconfiguration is too large (for example, fiber level or wavelength level). Figure 6 As shown in FIG. 1C, the first network system 1000A includes a first level switch including a core switch 11A and an OCS 12A, a second level switch including a spine switch 21A-22A and an OCS 23A, a third level switch including a leaf switch 31A-33A and an OCS 34A, and a fourth level switch including a TOR 41A-44A. The TOR 41A connects the computing boards 211A-214A, the TOR 42A connects the computing boards 221A-224A, the TOR 43A connects the computing boards 231A-234A, and the TOR 44A connects the computing boards 241A-244A. The OCS can reconfigure the sub-network to which the OCS belongs by using a micro electrical mechanical system (MEMS) or a wavelength selective switch (WSS) or the like. In this implementation, the OCS can connect the communication link between two computing boards within a preset time period, so that the OCS does not need to parse the received data packet, and the resource cost and power consumption are small. However, the OCS connects the transmission channel between two ports by a mechanical method, and the two ports are connected to two computing boards respectively. The speed of the OCS to establish the transmission channel is slow, which leads to slow reconfiguration speed of the OCS to the sub-network. In addition, the OCS can only connect the communication link between two computing boards in a data transmission process, and cannot connect the communication link between three or more computing boards. The bandwidth granularity of the OCS reconfiguration is too large (for example, fiber level or wavelength level). Figure 6The third network system 1000C can also be a fat tree architecture. The first level switches can act as core switches, the second level switches can act as spine switches, the third level switches can act as leaf switches, and the fourth level switches can act as TORs. The switch 41C is connected to the service boards 211C-214C, the switch 42C is connected to the service boards 221C-224C, the switch 43C is connected to the service boards 231C-234C, and the switch 44C is connected to the service boards 241C-244C. The third network system 1000C can further include a network controller 300C. The network controller 300C can be a standalone device. Alternatively, the network controller 300C can be arranged as a control board in a target switch in the plurality of switches, and the network controller 300C is connected to each of the plurality of switches. Figure 6 The network controller 300C is shown as a standalone device. The network controller 300C is connected to at least one of the switches. Figure 6 The network controller 300C is shown as a standalone device. The network controller 300C is connected to at least one of the switches. Figure 6 The network controller 300C is shown as a standalone device. The network controller 300C is connected to at least one of the switches. Figure 6 The network controller 300C is shown as a standalone device. The network controller 300C is connected to at least one of the switches. Figure 6 The network controller 300C is shown as a standalone device. The network controller 300C is connected to at least one of the switches. Figure 6 The network controller 300C is shown as a standalone device. The network controller 300C is connected to at least one of the switches. Figure 6 The switches shown in FIG. 10C include a first port and at least one second port. The first port of a switch is used to connect a first target device, and the at least one second port of the switch is used to connect at least one second target device. The first target device can be another switch or a service board, and the second target device can be another switch or a service board. For example, the first target device and the second target device corresponding to each of the first level switches, the second level switches, and the third level switches are all switches. The first target device corresponding to each of the fourth level switches is a service board, and the second target device corresponding to each of the fourth level switches is a switch; or the first target device corresponding to each of the fourth level switches is a service board, and the second target device corresponding to each of the fourth level switches is a service board; or the first target device corresponding to each of the fourth level switches is a switch, and the second target device corresponding to each of the fourth level switches is a service board.
[0048] Exemplarily, the plurality of service boards includes a first service board and a second service board. The first service board is configured to send service information to a switch connected with the first service board, the service information being based on an OTN protocol format. The first service board is further configured to send address information to the switch or a network controller, the address information including communication addresses of one or more second service boards processing the service information. For example, the network controller is arranged in the switch. The first service board is specifically configured to send the address information to the switch. The address information is used to instruct the switch to obtain configuration information according to the address information. The configuration information is used to instruct the switch to send the service information through one or more ports. For another example, the network controller is a separate device. The first service board is specifically configured to send the address information to the network controller 300C. The address information is used to instruct the network controller to obtain the configuration information according to the address information, and send the configuration information to the switch. For another example, the network controller is arranged in a target switch connected with the switch. The first service board is specifically configured to send the address information to the network controller 300C in the target switch. The address information is used to instruct the network controller to obtain the configuration information according to the address information, and send the configuration information to the switch. The network controller 300C is configured to receive the address information sent by the first service board. The network controller 300C is further configured to send the configuration information to one or more switches in at least one switch according to the address information, the configuration information being used to instruct the switch to send the service information through one or more second target ports, the at least one second port including the one or more second target ports. The configuration information is used to establish a target communication link between the switch, a first target device corresponding to the first port, and a second target device corresponding to the one or more second target ports, the target communication link being a sub-link of a communication link between the first service board and the one or more second service boards. The switch is configured to receive the service information sent by the first target device through the first port. The switch is further configured to obtain the configuration information. For example, the network controller 300C is arranged in the switch, and the switch obtains the configuration information from the network controller 300C. For another example, the network controller 300C can be a separate device or a control board arranged in the target switch. The switch is specifically configured to receive the configuration information sent by the network controller. The switch is further configured to send the service information to the corresponding second target device through the one or more second target ports in response to the configuration information.
[0049] As Figure 7As shown, the network controller 300C is an independent device, and the process is exemplified by taking the service board 211C sending service information to the service board 212C and the service board 221C as an example. The service board 211C is configured to send service information to the switch 41C, and the service information is based on the OTN protocol format. The service board 211C is also configured to send address information to the network controller 300C, and the address information is used to indicate the communication addresses of the service board 212C and the service board 221C. The network controller 300C is configured to receive the address information, and determine that the service board 212C and the service board 221C are the objects that the service board 211C needs to communicate with according to the address information. The network controller 300C is configured to send configuration information to the switch on the communication link between the service board 211C and the service board 212C, and also configured to send configuration information to the switch on the communication link between the service board 211C and the service board 221C. For example, the network controller 300C is also configured to send configuration information 1 to the switch 41C, and the configuration information 1 is used to indicate two second target ports of the switch 41C. The first port of the switch 41C is connected to the service board 211C, one second target port of the switch 41C is connected to the service board 212C, and the other second target port of the switch 41C is connected to the switch 31C. The configuration information 1 is used to establish a first target communication link between the service board 211C, the switch 41C and the service board 212C, and the first target communication link is the communication link between the service board 211C and the service board 212C. The configuration information 1 is also used to establish a second target communication link between the service board 211C, the switch 41C and the switch 31C, and the second target communication link is a sub-link of the communication link between the service board 211C and the service board 221C. The network controller 300C is also configured to send configuration information 2 to the switch 31C, and the configuration information 2 is used to indicate one second target port of the switch 31C. The first port of the switch 31C is connected to the switch 41C, and the second target port of the switch 31C is connected to the switch 42C. The configuration information 2 is used to establish a third target communication link between the switch 41C, the switch 31C and the switch 42C, and the third target communication link is a sub-link of the communication link between the service board 211C and the service board 221C. The network controller 300C is also configured to send configuration information 3 to the switch 42C, and the configuration information 3 is used to indicate one second target port of the switch 42C. The first port of the switch 42C can be connected to the switch 31C, and the second target port of the switch 42C can be connected to the service board 221C. The configuration information 3 is used to establish a fourth target communication link between the switch 31C, the switch 42C and the service board 221C, and the fourth target communication link is a sub-link of the communication link between the service board 211C and the service board 221C. In this way, the communication link between the service board 211C and the service board 212C is established, and the communication link between the service board 211C and the service board 221C is established.The switch 41C receives the service information sent by the service board 211C through the first port, sends the service information to the service board 212C through a second target port, and sends the service information to the switch 31C through another second target port. The switch 31C receives the service information sent by the switch 41C through the first port, and the switch 31C sends the service information to the switch 42C through the second target port. The switch 42C receives the service information sent by the switch 31C through the first port, and sends the service information to the service board 221C through the second target port.
[0050] In this embodiment, the OTN switch can determine to send the service information through one or more second target ports through the configuration information, so that the received service information does not need to be parsed. The processing process of the OTN switch is relatively simple, the resource cost is relatively small, the power consumption is relatively small, and the time delay is relatively low. Moreover, the OTN switch can receive the service information of one first target device and send the service information to one or more second target devices. That is, in one data transmission process, the OTN switch can connect the communication link between two or more target devices. Therefore, the energy efficiency ratio of the switch can be effectively improved. Moreover, the service board and the switch use the OTN technology to transmit the data based on the OTN protocol format, and the switch establishes the communication link between two or more service boards through the OTN circuit. Compared with the OCS which uses a mechanical method to establish the communication link, the OTN switch uses the circuit to establish the communication link, so that the speed of establishing the communication link of the OTN switch is relatively fast.
[0051] In a possible implementation, Figure 6 The bandwidth of the first port of the switch for transmitting the service information is greater than the bandwidth of the second target port for transmitting the service information. For example, the switch includes two second target ports. The bandwidth of the first port for transmitting the service information can be 100 MB / S, and the bandwidth of the two second target ports for transmitting the service information can each be 50 MB / S. Compared with the bandwidth for receiving the data and the bandwidth for sending the data which can only be a fixed relatively large granularity bandwidth (such as 100 MB / S), the OTN switch can transmit the data in a smaller granularity bandwidth (such as 50 MB / S). In this embodiment, the bandwidth of the first port of the switch for transmitting the service information is greater than the bandwidth of the second target port for transmitting the service information, and the granularity of the bandwidth of the switch for transmitting the data is relatively flexible. Therefore, the bandwidth switching capability can be provided to be more flexible and faster.
[0052] In a possible implementation, the service information includes one or more of AI task computing data, data service information, and communication interaction information. Exemplarily, the third network system 1000C can be an AI cluster system, and the service information can include AI task computing data. Alternatively, the third network system 1000C can be a data center network system, and the service information can include data service information. Alternatively, the third network system 1000C can be an OTN network system, and the service information can include communication interaction information. In this implementation, the type of service information can be one or more of AI task computing data, data service information, and communication interaction information, and the application scenarios of the embodiments of the present application are relatively rich.
[0053] In a possible implementation, the third network system 1000C is an AI cluster system, and the service information is AI task computing data. The process of allocating AI computing tasks is further included between the service board and the network controller 300C. In some examples, the network controller 300C is further configured to send corresponding AI computing tasks to the first service board and the second service board respectively. The first service board and the second service board are further configured to receive the AI computing tasks sent by the network controller 300C, and calculate the service information based on the AI computing tasks. In this implementation, the network controller 300C further sends corresponding AI computing tasks to each service board, thereby implementing complex AI computing tasks.
[0054] In a possible implementation, the OTN-related circuit can be introduced into the service board shown in Figure 6 , thereby implementing OTN technology.
[0055] In some examples, as shown in Figure 8 , the service board 211C in Figure 6 includes an OTN frame regulator 201C and an OTN PHY circuit 202C. The service board 211C can further include a service circuit 203C, a bus bridge circuit 204C, a MAC circuit 205C, and a demand control circuit 206C. As shown in Figure 9 , the service board 211C in Figure 6As an example of the switch 41C in the network device 40C, the switch 41C includes an OTN cross circuit 420C and a plurality of OTN PHY circuits. The plurality of OTN PHY circuits can include a first PHY circuit 410C and at least one second PHY circuit, for example, the plurality of OTN PHY circuits includes a second PHY circuit 421C, a second PHY circuit 422C, and a second PHY circuit 423C. The first PHY circuit 410C is connected to a first port of the switch 41C, thereby connected to the service board 211C. The second PHY circuit 421C is connected to a second port of the switch 41C, thereby connected to the service board 212C. The second PHY circuit 422C is connected to a second port of the switch 41C, thereby connected to the switch 31C. The second PHY circuit 423C is connected to a second port of the switch 41C, thereby connected to the service board 213C. The switch 41C can further include an OTN control circuit 410C. As shown in FIG. 4, the OTN control circuit 410C is connected to the first PHY circuit 410C and the second PHY circuit 421C, the second PHY circuit 422C, and the second PHY circuit 423C. Figure 10 Figure 6 The network controller 300C in the network device 40C includes an OTN route computation circuit 301C and an OTN distribution circuit 302C. The network controller 300C can further include a task deployment circuit 303C and a regulation circuit 304C.
[0056] Exemplarily, the task deployment circuit 303C in the network controller 300C is configured to distribute the large-scale services to the service boards in the third network system 1000C for distributed processing after parallel processing. The adjustment circuit 304C is configured to coordinate the synchronization and communication of the processing results of the service boards, for example, to control the processing sequence of the service boards. The service circuit 203C of the service board 211C is configured to perform actual services and obtain data packets according to service requirements, and generate communication requirements with the service circuits of other service boards. The service circuit 203C is connected to the MAC circuit 205C through the bus bridge circuit 204C. The MAC circuit 205C of the service board 211C receives the data packets and communication requirements from the service circuit 203C, and converts the communication requirements into communication addresses in the data packets. The requirement control circuit 206C of the service board 211C is configured to collect the communication addresses input from the MAC circuit 205C, and obtain the communication requirements of the service circuit 203C within a preset time. The requirement control circuit 206C sends address information to the OTN routing calculation circuit 301C of the network controller 300C, and the address information is used to indicate the communication addresses of the service board 212C and the service board 221C for processing service information within a preset time. Next, within the preset time, the OTN frame adjuster 201C of the service board 211C is configured to convert the data input from the MAC circuit 205C into OTN protocol format to obtain service information, and output the service information. The OTN PHY circuit 202C of the service board 211C is configured to input the service information, and send the service information to the first PHY circuit 410C of the switch 41C. The OTN routing calculation circuit 301C of the network controller 300C is configured to receive the address information sent by the requirement control circuit 206C of the service board 211C, and obtain the switch 41C on the communication link between the service board 211C and the service board 212C, and the switch 41C and the switch 31C on the communication link between the service board 211C and the service board 221C according to the address information. The OTN distribution circuit 302C of the network controller 300C is configured to send configuration information to the OTN control circuit 410C of the switch 41C and the OTN control circuit of the switch 31C. The control circuit 410C of the switch 41C receives the configuration information, controls the OTN cross circuit 420C to establish a transmission channel between the first PHY circuit 410C and the second PHY circuit 421C, and establishes a transmission channel between the first PHY circuit 410C and the second PHY circuit 422C. At this time, the second PHY circuit 421C and the second PHY circuit 422C are the second target PHY circuit.
[0057] In this embodiment, the OTN PHY circuit and the OTN cross circuit are deployed in each level of the switch, a transmission channel is established between the first PHY circuit and one or more second target PHY circuits, and a transmission channel is established between the first port and the one or more second target ports. The transmission channel between the first PHY circuit and the one or more second target PHY circuits is a sub-channel of the transmission channel between the first port and the one or more second target ports. In this way, the switch can send service information to the second target device corresponding to the one or more second target ports through the established transmission channel. The OTN frame regulator and the OTN PHY circuit are deployed in the service board, so that the data is transmitted in the OTN protocol format, and a communication link is established between the service board and the corresponding switch. The OTN routing calculation circuit and the OTN distribution circuit are deployed in the network controller 300C, so that the communication address sent by the service board is received, and the communication link between the plurality of service boards is controlled to be established.
[0058] It should be understood that the size of the sequence number of each process described above in various embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.
[0059] Those skilled in the art can realize that the modules and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered beyond the scope of the present application.
[0060] In several embodiments provided in the present application, it should be understood that the disclosed system and device can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the modules is only a logical function division, and actual implementation can have another division manner, for example, a plurality of modules or components can be combined or integrated into another device, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be indirect coupling or communication connection through some interface, device or module, and can be electrical, mechanical or other forms. The modules shown as separate components can be or can not be physically separated, and the components shown as modules can be or can not be physical modules, that is, they can be located in one device or distributed on a plurality of devices. According to actual needs, some or all of the modules can be selected to achieve the purpose of the embodiment.
[0061] In addition, each functional module in each embodiment of the present application can be integrated in one device, or each module can exist physically alone, or two or more modules can be integrated in one device.
[0062] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.
Claims
1. A switch, characterized in that, The switch includes a first port and at least one second port; the first port of the switch is used to connect to a first target device, and the at least one second port of the switch is used to connect to at least one second target device; the switch is used for: Receive service information sent by the first target device through the first port; Obtain configuration information, which is used to instruct the switch to send the service information to one or more second target ports, the service information being based on the Optical Transport Network (OTN) protocol format, and the at least one second port including the one or more second target ports; In response to the configuration information, the service information is sent to the corresponding second target device through one or more second target ports.
2. The switch according to claim 1, characterized in that, The switch includes an OTN cross-connect circuit and multiple OTN physical layer circuits. The multiple OTN physical layer circuits include a first physical layer circuit and at least one second physical layer circuit. The first physical layer circuit is connected to the first port, and the at least one second physical layer circuit is correspondingly connected to the at least one second port. The OTN cross-connect circuit is used for: In response to the configuration information, a transmission channel is established between the first physical layer circuit and one or more second target physical layer circuits; the at least one second physical layer circuit includes the one or more second target physical layer circuits, and the one or more second target circuits are correspondingly connected to the one or more second target ports.
3. The switch according to claim 1 or 2, characterized in that, The bandwidth of the first port for transmitting the service information is greater than the bandwidth of the second target port for transmitting the service information.
4. The switch according to any one of claims 1-3, characterized in that, The business information includes one or more of the following: AI task computation data, data business information, and communication interaction information.
5. The switch according to any one of claims 1-4, characterized in that, The switch is also used to connect to a network controller; The switch is specifically used to receive the configuration information sent by the network controller.
6. The switch according to claim 5, characterized in that, The network controller is installed in the target switch connected to the switch.
7. A service board, characterized in that, The service board is used to connect to the switch; the service board is used for: Send service information to the switch, the service information being based on the Optical Transport Network (OTN) protocol format; The address information is sent to the switch or network controller. The address information includes the communication address of one or more target service boards used to process the service information. The address information is used to instruct the switch to obtain configuration information based on the address information, or the address information is used to instruct the network controller to obtain configuration information based on the address information and send the configuration information to the switch. The configuration information is used to instruct the switch to send one or more ports for the service information.
8. The service board according to claim 7, characterized in that, The service board includes an OTN frame modulator and an OTN physical layer circuit. The OTN frame conditioner is used to: convert the input data into the OTN protocol format to obtain the service information, and output the service information; The OTN physical layer circuit is used to: input the service information and send the service information to the switch.
9. The service board according to claim 7 or 8, characterized in that, The business information includes one or more of the following: AI task computation data, data business information, and communication interaction information.
10. The service board according to any one of claims 7-9, characterized in that, The business information is AI task calculation data, and the business board is also used for: The system receives AI computing tasks sent by the network controller and calculates the service information based on the AI computing tasks.
11. The service board according to claim 10, characterized in that, The network controller is installed in the target switch connected to the switch.
12. A network controller, characterized in that, The network controller is used to connect at least one switch and multiple service boards; the switch includes a first port and at least one second port; the first port of the switch is used to connect to a first target device, and the at least one second port of the switch is used to connect to at least one second target device; the multiple service boards include a first service board and one or more second service boards; the network controller is used to: The system receives address information sent by the first service board, the address information being used to indicate the communication address of one or more second service boards used to process the service information, the service information being based on the Optical Transport Network (OTN) protocol format; Based on the address information, configuration information is sent to one or more of the at least one switch. The configuration information is used to instruct the switch to send the service information to one or more second target ports, and the at least one second port includes the one or more second target ports. The configuration information is used to establish a target communication link between the switch, the first target device corresponding to the first port, and the second target device corresponding to the one or more second target ports. The target communication link is a sub-link of the communication link between the first service board and the one or more second service boards.
13. The network controller according to claim 12, characterized in that, The network controller includes an OTN routing calculation circuit and an OTN distribution circuit; The OTN routing calculation circuit is used to: receive address information sent by the first service board, and obtain the one or more switches on the communication link between the first service board and the one or more second service boards based on the address information; The OTN distribution circuit is used to send the configuration information to the one or more switches.
14. The network controller according to claim 13, characterized in that, The business information includes one or more of the following: AI task computation data, data business information, and communication interaction information.
15. The network controller according to any one of claims 12-14, characterized in that, The business information is AI task computation data, and the network controller is also used for: Send the corresponding AI computing tasks to the first business board and the one or more second business boards respectively.
16. The network controller according to any one of claims 12-15, characterized in that, The network controller is located in at least one of the switches.
17. A network system, characterized in that, The network system includes a switch as described in any one of claims 1-6, a service board as described in any one of claims 7-11, and a network controller as described in any one of claims 12-16.
18. The network system according to claim 17, characterized in that, The network system is an artificial intelligence (AI) cluster system, a data center network system, or an OTN network system.