An electro-optical conversion device, computing node and cluster

Abstract

An electro-optical conversion device, computing node, and cluster are disclosed, relating to the field of optical communication technology, to reduce the number of cables between the external cabinet and the switching equipment. This application proposes an electro-optical conversion device for computing or communication nodes. It utilizes a processing module to multiplex multi-channel data input from multiple input ports, and after electro-optical conversion by one or more optical modules, couples the data to at least one optical port for transmission. This achieves both port electrical signal multiplexing and channel electrical signal multiplexing. Reducing the number of ports reduces the number of cable connections between the computing node and the optical switching equipment. Furthermore, it reduces the number of optical ports used by the optical switching equipment, thereby reducing costs.

Description

Technical Field

[0001] This application relates to the field of optical communication technology, and in particular to an electro-optical conversion device, a computing node, and a cluster. Background Technology

[0002] With the rapid development of artificial intelligence (AI) technology, especially the rapid rise of large-scale AI models, the demand for computing resources is growing exponentially. The operation and training of large-scale AI models require processing massive amounts of data, placing high demands on computing power. Traditional computing models are facing severe challenges, as the marginal benefits of individual processor performance are diminishing, approaching physical limits, and making it difficult to meet the continuously growing computing power demands of large-scale AI models.

[0003] To meet the ever-increasing computing power demands of large AI models, supernodes are currently implemented using GPU clusters. These GPU clusters are deployed in two or more racks. The individual GPUs are interconnected via electrical / optical switches. However, the racks are connected to the electrical / optical switches via external cables or fiber optic cables, resulting in massive cable data transfer and poor heat dissipation. Summary of the Invention

[0004] This application provides an electro-optical conversion device, a computing node, and a cluster to reduce the number of cables between the outside of the cabinet and the switching equipment.

[0005] In a first aspect, embodiments of this application provide an electro-optical conversion device applied to a computing node or a communication node. The electro-optical conversion device includes a processing module, M input ports, and K first output ports, where K is a positive integer, M is an integer greater than 1, and M>K. Each of the M input ports includes at least one input channel, and the number of input channels in the M input ports is P. Each of the K first output ports includes at least one first output channel, and the number of first output channels in the K first output ports is Q, where P and Q are positive integers, P>Q, and P is an integer multiple of Q. Each input channel is used to receive an electrical signal, which carries service data of the computing node or the communication node. The processing module is used to multiplex the electrical signals from the P input channels to the Q first output channels for output through the K first output ports.

[0006] This application proposes an electro-optical switching device for computing or communication nodes, based on the aforementioned solution. It utilizes a processing module to multiplex input multi-channel data, and after electro-optical conversion by one or more optical modules, couples the data to at least one optical port for transmission. This achieves both port electrical signal multiplexing and channel electrical signal multiplexing. Reducing the number of ports reduces the number of cable connections between the computing node and the optical switching device. Furthermore, it reduces the number of optical ports used by the optical switching device, thereby lowering costs.

[0007] For example, a communication node can be a node in an integrated scenario that includes multiple OTN chips, or a node in an integrated scenario that includes multiple Ethernet service processing chips.

[0008] In one possible implementation, the K first output ports are used to plug in and unplug the K first optical modules one by one.

[0009] Any one of the K first optical modules is used to perform photoelectric conversion on the electrical signal from at least one first output channel of the corresponding first output port and then couple it to an optical port for output.

[0010] In the above method, the hot-swappable function prevents the processing module 210 from stopping operation in the event of a signal interruption. This method enables fault isolation between the electrical domain of the computing node and the optical domain of the optical switching equipment. Furthermore, the hot-swappable method allows for timely replacement of optical modules in case of failure.

[0011] In one possible implementation, the electro-optical conversion device further includes K first optical modules corresponding one-to-one with the K first output ports; the K first optical modules are fixed inside the electro-optical conversion device through the K first output ports;

[0012] Any one of the K first optical modules is used to perform photoelectric conversion on the electrical signal from at least one first output channel of the corresponding first output port and then couple it to an optical port for output.

[0013] In one possible implementation, the rate of the optical signal output from the optical port of any first optical module is greater than the rate of the electrical signal output from the first output channel received by any first optical module.

[0014] The above solution improves the signal rate in the optical module, which can further enhance the signal transmission rate.

[0015] In one possible implementation, the rate of the electrical signal of any of the first output channels is greater than the rate of the electrical signal of any of the input channels.

[0016] In the above scheme, the signal rate of the multiplexed channel is increased while multiplexing, which can improve the signal transmission rate.

[0017] In one possible implementation, at least one of the M input ports is a hot-pluggable port.

[0018] Hot-swapping prevents the processing module 210 from stopping in the event of a signal interruption. This method enables fault isolation between the electrical domain of the computing node and the optical domain of the optical switching equipment.

[0019] In one possible implementation, K>1, the device further includes a multiplexing module;

[0020] The multiplexing module is used to multiplex the K optical signals output by the K first optical modules and then output them.

[0021] The multiplexing function can further reduce the number of optical ports, thereby reducing the number of cables connected to optical switching equipment.

[0022] In one possible implementation, the processing module is specifically used to: multiplex the electrical signals of P input channels into Q multiplexed electrical signals; and output the Q multiplexed electrical signals through the Q first output channels after FEC encoding.

[0023] By re-encoding the multiplexed signal using the above method, transmission performance can be further improved. Furthermore, it prevents electrical errors from propagating to the optical domain, achieving electro-optical error isolation.

[0024] In one possible implementation, the processing module is specifically configured to: perform forward error correction (FEC) decoding on the electrical signals of the P input channels; multiplex the FEC-decoded electrical signals of the P input channels into Q multiplexed electrical signals; and output the Q multiplexed electrical signals through the Q first output channels after FEC encoding.

[0025] By re-encoding the multiplexed signal using the above method, transmission performance can be further improved. Furthermore, it prevents electrical errors from propagating to the optical domain, achieving electro-optical error isolation.

[0026] In one possible implementation, each of the K first output ports further includes at least one second output channel;

[0027] The processing module is further configured to, when any of the Q first output channels fails, output the electrical signal of any first output channel through the target output channel corresponding to any of the first output channels in the second output channels included in the K first output ports, according to the configuration information.

[0028] By implementing the above solution and adding a backup protection channel, service transmission failures can be prevented if one channel fails.

[0029] In one possible implementation, K>1, the device further includes at least one second output port, each of the at least one second output port including a plurality of third output channels; the processing module is further configured to:

[0030] When the link corresponding to any of the K first output ports fails, the electrical signal to be output to any of the first output ports will be output through one of the at least one second output ports.

[0031] By using the above solution, and by adding a spare port to connect to the protection optical module, service transmission can be prevented from failing after a certain optical module fails.

[0032] Secondly, embodiments of this application provide an electro-optical conversion device applied to a computing node or a communication node. The electro-optical conversion device includes a processing module, M output ports, and K first input ports, where K is a positive integer, M is an integer greater than 1, and M>K. Each of the M output ports includes at least one output channel, and the number of output channels in the M output ports is P. Each of the K first input ports includes at least one first input channel, and the number of first input channels in the K first input ports is Q, where P and Q are positive integers, P>Q, and P is an integer multiple of Q. Each output channel is used to output an electrical signal to the computing node or the communication node, and the electrical signal is used to carry the service data of the computing node or the communication node.

[0033] The processing module is used to demultiplex the electrical signals input through the K first input ports and the Q first input channels to the P output channels.

[0034] In one possible implementation, the K first input ports are used to plug in and unplug the K third optical modules one by one.

[0035] Any one of the K third optical modules is used to perform photoelectric conversion processing on the optical signal input from the optical port into at least one electrical signal, and output the at least one electrical signal to at least one first input channel of the corresponding first input port.

[0036] In one possible implementation, the electro-optical conversion device further includes K third optical modules corresponding one-to-one with the K first input ports; the K third optical modules are fixed inside the electro-optical conversion device through the K first input ports;

[0037] Any one of the K third optical modules is used to perform photoelectric conversion processing on the optical signal input from the optical port into at least one electrical signal, and output the at least one electrical signal to at least one first input channel of the corresponding first input port.

[0038] In one possible implementation, the rate of the optical signal input to the optical port of any third optical module is greater than the rate of the electrical signal output from any third optical module to any first input channel.

[0039] In one possible implementation, K>1, the device further includes a wavelength division module;

[0040] The wavelength division module is used to divide one optical signal from the optical switching device into K optical signals, and output the K optical signals one by one to the K third optical modules.

[0041] In one possible implementation, the rate of the electrical signal of any of the first input channels is greater than the rate of the electrical signal of any of the output channels.

[0042] In one possible implementation, at least one of the M output ports is a hot-pluggable port.

[0043] In one possible implementation, the processing module is specifically used for:

[0044] The electrical signals of the Q first input channels are demultiplexed into P electrical signals;

[0045] The P-channel electrical signals are encoded using forward error correction (FEC) and then output through the P output channels.

[0046] In one possible implementation, the processing module is specifically used for:

[0047] The electrical signals input through the Q first input channels are subjected to FEC decoding to obtain Q-channel FEC-decoded electrical signals;

[0048] The electrical signal decoded by the Q-channel FEC is demultiplexed into a P-channel electrical signal;

[0049] The P-channel electrical signals are encoded and then output through the P output channels.

[0050] In one possible implementation, each of the K first input ports further includes at least one second input channel;

[0051] The processing module is further configured to, when any of the Q first input channels fails, switch the received electrical signal from any of the first input channels to the target input channel corresponding to any of the first input channels in the second output channels included in the K first input ports, according to the configuration information.

[0052] In one possible implementation, K>1, the device further includes at least one second input port, each of the at least one second input port including a plurality of third input channels;

[0053] The processing module is further configured to:

[0054] When a link failure occurs at any of the K first input ports, the electrical signal received from any of the first input ports will be switched to be received from one of the at least one second input ports.

[0055] Thirdly, this application provides a first computing node, including the electro-optical conversion device described in the first aspect or any implementation thereof, and N computing units; each of the N computing units includes L output ports, and the N*L output ports of the N computing units are connected one-to-one with the N*L input ports of the electro-optical conversion device, wherein N*L is less than or equal to M.

[0056] In one possible implementation, the N computing units are connected to the electro-optical conversion device via PCB traces or via cables.

[0057] Fourthly, this application provides a second computing node, including the electro-optical conversion device described in the second aspect or any implementation thereof, and N computing units; each of the N computing units includes L input ports, and the N*L input ports of the N computing units are connected one-to-one with the N*L output ports of the electro-optical conversion device, wherein N*L is less than or equal to M.

[0058] In one possible implementation, the N computing units are connected to the electro-optical conversion device via PCB traces or via cables.

[0059] Fifthly, embodiments of this application provide a computing cluster, including the first computing node described in the third aspect and an optical switching device;

[0060] The J input ports on the optical switching device are connected to the J optical ports on the first computing node, where J = 1 or J = K;

[0061] The optical switching device is used to switch the optical signal received from the first computing node through the first input port of the J input ports to the first output port of the optical switching device; the first output port is the target output port of the J output ports of the optical switching device corresponding to the first input port.

[0062] In one possible implementation, the second computing node described in the fourth aspect is also included; the J output ports on the optical switching device are connected to the J optical ports on the second computing node.

[0063] Sixthly, embodiments of this application provide an electro-optical conversion method, the method being applied to an electro-optical conversion device, the electro-optical conversion device being applied to a computing node, the electro-optical conversion device including a processing module, M input ports and K first output ports, where K is a positive integer, M is an integer greater than 1, and M>K; wherein, each of the M input ports includes at least one input channel, and the number of input channels included in the M input ports is P; each of the K first output ports includes at least one first output channel, and the number of first output channels included in the K first output ports is Q, where P and Q are positive integers, P>Q and P is an integer multiple of Q; each input channel is used to receive an electrical signal, the electrical signal being used to carry service data of the computing node or the communication node; the method includes: receiving electrical signals input through the P input channels through the M input ports; and multiplexing the electrical signals of the P input channels to the Q first output channels through the processing module for output through the K first output ports.

[0064] In a seventh aspect, embodiments of this application provide an electro-optical conversion method, the method being applied to an electro-optical conversion device, the electro-optical conversion device being applied to a computing node, the electro-optical conversion device including a processing module, M input ports and K first output ports, where K is a positive integer, M is an integer greater than 1, and M>K; wherein, each of the M input ports includes at least one input channel, and the number of input channels included in the M input ports is P; each of the K first output ports includes at least one first output channel, and the number of first output channels included in the K first output ports is Q, P and Q is a positive integer, P>Q and P is an integer multiple of Q; each of the input channels is used to receive electrical signals, which are used to carry service data of the computing node or the communication node; the processing module is used to multiplex the electrical signals of the P input channels to the Q first output channels for output through the K first output ports; the method includes: receiving electrical signals from a first computing unit in the computing node through the first input port of the processing module; and multiplexing the electrical signals of at least one input channel included in the first input port to a target output channel among the Q first output channels through the processing module.

[0065] Eighthly, embodiments of this application provide a chip for implementing the methods described in the sixth or seventh aspect.

[0066] Based on the implementations provided in the above aspects, this application can be further combined to provide more implementations. Attached Figure Description

[0067] Figure 1 This is a schematic diagram of a possible computing cluster structure;

[0068] Figure 2A This is a schematic diagram of the structure of an electro-optical conversion device 200 provided in an embodiment of this application;

[0069] Figure 2B This is a schematic diagram of the structure of an electro-optical conversion device 200 provided in an embodiment of this application;

[0070] Figure 2C This is a schematic diagram of the structure of an electro-optical conversion device 200 provided in an embodiment of this application;

[0071] Figure 2D This is a schematic diagram of the structure of an electro-optical conversion device 200 provided in an embodiment of this application;

[0072] Figure 3 This is a schematic diagram of the structure of an electro-optical conversion device 200 provided in an embodiment of this application;

[0073] Figure 4This is a schematic diagram of the structure of an electro-optical conversion device 200 provided in an embodiment of this application;

[0074] Figure 5 This is a schematic diagram of the structure of an electro-optical conversion device 200 provided in an embodiment of this application;

[0075] Figure 6 This is a schematic diagram of the structure of an electro-optical conversion device 200 provided in an embodiment of this application;

[0076] Figure 7 This is a schematic diagram of the structure of an electro-optical conversion device 200 provided in an embodiment of this application;

[0077] Figure 8 This is a schematic diagram of the structure of an electro-optical conversion device 200 provided in an embodiment of this application;

[0078] Figure 9 This is a schematic diagram of the structure of an electro-optical conversion device 200 provided in an embodiment of this application;

[0079] Figure 10 This is a schematic diagram of the structure of an electro-optical conversion device 200 provided in an embodiment of this application;

[0080] Figure 11A This is a schematic diagram of the structure of an electro-optical conversion device 300 provided in an embodiment of this application;

[0081] Figure 11B This is a schematic diagram of the structure of an electro-optical conversion device 300 provided in an embodiment of this application;

[0082] Figure 11C This is a schematic diagram of the structure of an electro-optical conversion device 300 provided in an embodiment of this application;

[0083] Figure 11D This is a schematic diagram of the structure of an electro-optical conversion device 300 provided in an embodiment of this application;

[0084] Figure 12 This is a schematic diagram of the structure of an electro-optical conversion device 300 provided in an embodiment of this application;

[0085] Figure 13 This is a schematic diagram of the structure of an electro-optical conversion device 300 provided in an embodiment of this application;

[0086] Figure 14 This is a schematic diagram of the structure of an electro-optical conversion device 300 provided in an embodiment of this application;

[0087] Figure 15 This is a schematic diagram of the structure of an electro-optical conversion device 300 provided in an embodiment of this application;

[0088] Figure 16 This is a schematic diagram of the structure of an electro-optical conversion device 300 provided in an embodiment of this application;

[0089] Figure 17 This is a schematic diagram of the structure of an electro-optical conversion device 300 provided in an embodiment of this application;

[0090] Figure 18 This is a schematic diagram of the structure of an electro-optical conversion device 300 provided in an embodiment of this application;

[0091] Figure 19 This is a schematic diagram of the structure of an electro-optical conversion device 300 provided in an embodiment of this application;

[0092] Figure 20 This is a schematic diagram of the structure of an electro-optical conversion device 400 provided in an embodiment of this application;

[0093] Figure 21 This is a schematic diagram of the structure of an electro-optical conversion device 400 provided in an embodiment of this application;

[0094] Figure 22 This is a schematic diagram of the structure of an electro-optical conversion device 400 provided in an embodiment of this application;

[0095] Figure 23 This application provides a schematic diagram of the structure of a computing node.

[0096] Figure 24 This is a schematic diagram of the structure of a communication node provided in an embodiment of this application;

[0097] Figure 25 A schematic diagram of a computing cluster architecture provided for an embodiment of this application;

[0098] Figure 26 A schematic diagram of a computing cluster architecture provided for an embodiment of this application;

[0099] Figure 27 This is a schematic flowchart of an electro-optic conversion method provided in an embodiment of this application;

[0100] Figure 28 This is a schematic diagram of another electro-optical conversion method provided in an embodiment of this application. Detailed Implementation

[0101] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0102] In the description of this application, unless otherwise stated, "multiple" refers to two or more. Additionally, " / " indicates that the related objects are in an "or" relationship; for example, A / B can represent A or B. "And / or" in this application merely describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone, where A and B can be singular or plural. Furthermore, to facilitate a clear description of the technical solutions of the embodiments of this application, the terms "first" and "second" are used in the embodiments to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that "first" and "second" are not necessarily different. It should also be noted that, unless specifically stated, the specific description of some technical features in one embodiment can also be used to explain the corresponding technical features mentioned in other embodiments.

[0103] Currently, to meet the ever-increasing computing power demands of large AI models, supernodes are implemented by deploying computing clusters. (See also...) Figure 1 The diagram shows a possible computing cluster structure. Figure 1 Taking a 128-port supernode as an example, the computing cluster consists of two racks, each deploying 64 GPUs. The 128 GPUs between the two racks are interconnected via optical or electrical switches. For example, a 2D interconnect method can be used. Each rack includes eight trays, each deploying eight GPUs. Each GPU has 22 ports. Seven ports on each GPU connect to the other seven GPUs in the same tray via optical or electrical switches, and the remaining 15 ports connect to 15 GPUs with the same number in 15 other trays via optical or electrical switches.

[0104] An optical switch is a device that directly switches optical signals without requiring electro-optical-to-photoelectric conversion. It can transmit and switch signals within the optical domain, offering advantages such as high speed, large capacity, and low loss. An electrical switch, on the other hand, is a network device used for forwarding electrical (optical) signals. It provides a dedicated electrical signal path for any two network nodes connected to the switch, ensuring efficient information transmission.

[0105] A single cabinet has 22 * ​​64 = 1408 electrical (or optical) ports, resulting in a large number of cables connecting the switching equipment, complex fiber routing, and higher maintainability and reliability risks. Furthermore, the scale of supernodes is constrained by the number of ports on the switching equipment. Additionally, if electrical switches are used, different manufacturers configure different bus protocols, requiring strong coupling between the cabinet and the electrical switch. Moreover, electrical switches have a relatively high latency (200ns+), impacting computational efficiency.

[0106] Based on this, embodiments of this application provide a photoelectric conversion device that can be applied to computing nodes. Embodiments of this application can also be applied to large-scale network scenarios such as OTN networks or Ethernet networks. The photoelectric conversion device is applied to a communication node integrating multiple communication chips. The communication chips can be OTN chips or Ethernet service processing chips, etc. The photoelectric conversion device is first connected to the computing unit (or the communication chip in the communication node) in the computing node, and then connected to the optical switching equipment in the optical domain. The electro-optical conversion device performs 1:X convergence on multiple input channels, that is, X electrical channels are converged and accelerated to one channel, where X is an integer greater than 1. Then, photoelectric conversion is performed through an optical module, and one or more optical signals are coupled to the optical port output. Exemplarily, the computing unit can be a graphics processing unit (GPU), a neural network processing unit (NPU), or a central processing unit (CPU). The computing unit can be a computing chip used in cluster processing services in high-performance computing (HPC), artificial intelligence (AI), and other application scenarios.

[0107] The electro-optic conversion device in the embodiments of this application can be a chip, a single board, or a device. The following description will use the device as an example, and will refer to it as an electro-optic conversion device.

[0108] See Figure 2A and Figure 2BThe diagram shown is a structural schematic of the electro-optical conversion device 200 provided in an embodiment of this application. The electro-optical conversion device 200 includes a processing module 210, M input ports, and K first output ports, where K is a positive integer, M is an integer greater than 1, and M>K. Each of the M input ports includes at least one input channel, and the number of input channels in the M input ports is P. Each of the K first output ports includes at least one first output channel, and the number of first output channels in the K first output ports is Q. P and Q are positive integers. P>Q. In some possible implementations, P is an integer multiple of Q, for example, P / Q=X. Each input channel is used to receive electrical signals, which are used to carry service data of computing nodes or communication nodes. Different input ports may have the same or different number of input channels.

[0109] In one possible implementation, the K first output ports are used to plug in and unplug the K first optical modules one-to-one. For example, the K first optical modules can be hot-plugged into the K first output ports of the optical switching device 200.

[0110] In another possible implementation, K first optical modules are disposed inside the optical switching device 200. Each of the K first optical modules corresponds one-to-one with one of the K first output ports. The K first optical modules are fixed inside the electro-optical conversion device via the K first output ports. In this case, the K first output ports can also be understood as the K output ports of the processing module 210.

[0111] Figure 2A In this example, we take K=1 and K first optical modules as an example, which are set inside the electro-optical conversion device. Figure 2B In this example, we take K>1 and K first optical modules as being installed inside the electro-optical conversion device.

[0112] Figure 2C In the example, K=1 and K first optical modules are plugged into K first output ports of the electro-optical conversion device. Figure 2D In the example, K>1 and K first optical modules are inserted into K first output ports of the electro-optical conversion device.

[0113] To facilitate differentiation from subsequent optical modules, the K optical modules will be referred to as the first optical module.

[0114] See Figure 2AAs shown, the electro-optical conversion device 200 includes a processing module 210 and a first optical module 220. Q first output channels are connected to the first optical module 220. The processing module 210 performs a multiplexing operation on the electrical signals of the P input channels; specifically, it multiplexes the electrical signals of the P input channels to the Q first output channels. The multiplexing operation can employ interleaved multiplexing, byte interleaving, or bit interleaving.

[0115] The first optical module 220 performs photoelectric conversion processing on the electrical signals from the Q first output channels and then couples them to an optical port for output.

[0116] In one possible implementation, the first optical module 220 photoelectrically converts the electrical signals from each of the Q first output channels from the processing module into a single optical signal, resulting in Q optical signals, and then couples (or combines) these Q optical signals into a single optical port for output. In another possible implementation, the first optical module 220 multiplexes the electrical signals from the Q first output channels from the processing module into a single electrical signal, and then photoelectrically converts this single electrical signal into a single optical signal, which is then coupled into a single optical port for output.

[0117] See Figure 2C As shown, the processing module 210 and the first optical module 220 are connected via a plug-in connection. In the electro-optical conversion device 200, the processing module 210 multiplexes the electrical signals from the P input channels to the first output port. Thus, the first optical module, which is plugged into the first output port, receives the electrical signals from the Q first output channels of the first output port, performs photoelectric conversion on the electrical signals from the Q first output channels, and couples them to an optical port for output.

[0118] See Figure 2B As shown, the electro-optical conversion device 200 includes a processing module 210 and K first optical modules 220. To distinguish the K first optical modules 220, Figure 2B In this example, K first optical modules 220 are designated as first optical module 220-1 to first optical module 220-K. One of the Q first output channels is connected to one of the K first optical modules, and each of the K first optical modules is connected to at least one of the Q first output channels. It should be noted that each first output channel is connected to only one first optical module; multiple first optical modules are not connected. The processing module 210 multiplexes the electrical signals from the P input channels to the Q first output channels. Each first optical module 220 performs photoelectric conversion on the electrical signal from at least one first output channel of the processing module and couples it to an optical port for output.

[0119] In one possible implementation, the first optical module 220 photoelectrically converts the electrical signal from each of the at least one first output channel of the processing module into an optical signal to obtain at least one optical signal, and couples the at least one optical signal to an optical port for output. In another possible implementation, the first optical module 220 multiplexes the electrical signal from at least one first output channel of the processing module into an electrical signal, and photoelectrically converts the electrical signal into an optical signal to couple it to an optical port for output.

[0120] See Figure 2D As shown, the processing module 210 is connected to the K first optical modules 220 via a plug-in connection. In the electro-optical conversion device 200, the processing module 210 multiplexes the electrical signals from the P input channels to the K first output ports. Thus, the first optical modules plugged into the first output ports receive the electrical signals from at least one first output channel of the first output port, perform photoelectric conversion processing on the electrical signals from at least one first output channel, and couple them to an optical port for output.

[0121] In one possible implementation, the processing module 210 also speeds up the electrical signals of the output channels during the multiplexing operation. After speeding up, the rate of the electrical signal output by any first output channel is higher than the rate of the electrical signal of any input channel.

[0122] In some implementation scenarios, after the first optical module 220 performs electro-optical conversion on the electrical signals of multiple first output channels, it can also speed up the optical signal processing during optical port coupling. After speed-up, the rate of the optical signal output from the optical port of the first optical module 220 is greater than the rate of the electrical signal input to any of the first output channels of the first optical module.

[0123] Through a single-stage speed-up via processing module 210 or the first optical module 220, or through a two-stage speed-up via processing module 210 and the first optical module 220, the data exchange rate between computing nodes can be increased, thereby further improving the computing efficiency of the computing cluster or supernode.

[0124] This application achieves convergence not only for channels but also for ports through the aforementioned method. Furthermore, the number of channels in the input port can be the same as or different from the number of channels in the output port.

[0125] It should be noted that the ports described above (including input ports and output ports) can also be called interfaces.

[0126] The aforementioned input ports can be used to connect to computing units. These input ports can support the Peripheral Component Interconnect Express (PCIE) standard, Ethernet (ETH) protocol, or other proprietary protocols.

[0127] As an example, see Figure 3 As shown, taking an input port with 8 ports as an example (ports 1 to 8), each port has 4 input channels (lanes), i.e., P = 32. Taking the first output port with 4 output channels as an example... Figure 3 Taking K=1 as an example, that is, Q=4.

[0128] As an example, see Figure 4 As shown, taking a number of input ports of 32 as an example, namely ports 1 to 32, with 4 input channels in each port, then P = 128. Taking a number of first output ports of 4, i.e., K = 4, each first output port includes 4 first output channels. Figure 4 In this example, four first optical modules 220 are first optical module 220-1 to first optical module 220-4.

[0129] In some possible implementations, the electro-optical conversion device 200 may also support forward error correction (FEC) isolation and regeneration between the electrical and optical domains.

[0130] In one approach, before performing the multiplexing operation, the processing module 210 can first perform FEC decoding on the electrical signals of the P input channels. Then, the FEC-decoded electrical signals of the P input channels are multiplexed into Q multiplexed electrical signals. Next, the Q multiplexed electrical signals are FEC-encoded and then output through the Q first output channels. Performing error correction processing on the signals before multiplexing, and then performing FEC regeneration after multiplexing, can improve data transmission capability. In another approach, the electrical signals of the Q first input channels are demultiplexed into P electrical signals; the P electrical signals are then subjected to forward error correction (FEC) encoding and output through the P output channels. Performing FEC encoding before sending to the optical domain can improve data transmission capability.

[0131] In some possible implementations, at least one of the M input ports is a hot-swappable port. The service implementation of the hot-swappable port is the hot-swappable function, which means that when a peripheral device is connected to the device, the device can sense the peripheral device and transmit signals with it without needing to power on again. The hot-swappable function prevents the processing module 210 from stopping operation in the event of a signal interruption. In this way, fault isolation between the electrical domain of the computing node and the optical domain of the optical switching equipment can be achieved.

[0132] Processing module 210 may include one or more processors. The processor may be a general-purpose processor or a dedicated processor, etc. One or more processors can multiplex the electrical signals of P input channels to Q first output channels. Optionally, in addition to implementing the scheme of the embodiments shown above, the processor may also implement other functions. Optionally, in one design, the processor can execute instructions that cause processing module 210 to perform the operation of multiplexing the electrical signals of P input channels to Q first output channels. The instructions may be stored entirely or partially within the processor, or entirely or partially in a memory coupled to the processor. In one possible design, processing module 210 may include one or more memories storing instructions that can be executed on the processor, causing processing module 210 to perform the operation of multiplexing the electrical signals of P input channels to Q first output channels. Optionally, the memory may also store data. Optionally, the processor may also store instructions and / or data. The processor and memory may be configured separately or integrated together.

[0133] It should be noted that the processor in the embodiments of this application can be an integrated circuit chip with signal processing capabilities. The aforementioned processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor.

[0134] It is understood that the memory in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM).

[0135] In some possible implementations, in order to further reduce the cable connections between the electro-optical conversion device 200 and the optical switching device, the optical signals output by the K first optical modules can be further combined.

[0136] See Figure 5 As shown, the electro-optical conversion device 200 may also include a multiplexing module 230. In this embodiment, the multiplexing module 230 may be a wavelength selective switch (WSS), an array waveguide grating (AWG), or other devices capable of multiplexing. The multiplexing module 230 multiplexes the K optical signals output from the K first optical modules and outputs the combined signals. In some implementation scenarios, the multiplexing module 230 may also be located outside the electro-optical conversion device 200.

[0137] In the above scheme, the multiplexing module 230 speeds up the optical signal during the multiplexing process. After speed-up, the rate of the optical signal after multiplexing by the multiplexing module 230 is greater than the rate of the optical signals output by the K first optical modules. Through the multiplexing module 230, the electro-optical conversion device 200 can support three-level speed-up: processing module + optical module + multiplexing module. This can further improve the data transmission rate.

[0138] In one possible implementation, the electro-optical conversion device 200 can also support channel-level protection. This can be achieved through protection switching or load sharing. For example, using protection switching, each of the K first output ports of the electro-optical conversion device 200 includes at least one second output channel. Then, when any first output channel fails, the processing module 210 switches the electrical signal of any first output channel to the second output channel. The processing module 210 can be configured with configuration information indicating the correspondence between the first and second output channels. Multiple first output channels may correspond to one output channel. For example, the configuration information may include the correspondence between the channel number of the first output channel and the channel number of the second output channel. In some implementations, multiple first output channels belonging to the same first output port correspond to second output channels belonging to the same first output port. In other implementations, each of the Q first output channels corresponds to one second output channel. The corresponding first and second output channels can belong to the same first output port or different first output ports. In some implementations, the number of second output channels can be multiple. It can protect multiple first output channels simultaneously. For example, if two first output channels fail, the output can be switched to two second output channels. Alternatively, by using load sharing, if one first output channel fails, the processing module 210 can multiplex the electrical signals of P input channels to the remaining Q-1 output channels.

[0139] See Figure 6 As shown, taking K=1 as an example. In Figure 2A Based on what is shown, Figure 6 The first output port of the electro-optical conversion device 200 shown also includes a second output channel. When one of the Q first output channels connected to the optical module 220 fails, the electrical signal of that first output channel can be output through the second output channel.

[0140] See Figure 7 As shown, taking K>1 as an example. In Figure 2A Based on what is shown, Figure 7The processing module 210 shown also includes a second output channel in each of its K first output ports. When a first output channel of any of the K first optical modules 220 fails, the electrical signal of that first output channel can be output through the second output channel. For example, if a first output channel of first optical module 220-1 fails, the electrical signal of that first output channel can be output through the second output channel. In some implementation scenarios, there can be multiple second output channels. Multiple first output channels can be protected simultaneously. For example, if two first output channels fail, the output can be switched to two second output channels.

[0141] In one possible implementation, the electro-optical conversion device 200 can also support module-level protection. Protection can be provided for K first optical modules 220 or optical ports. The electro-optical conversion device 200 may also include at least one third output port, which includes at least one third output channel. The third output port is used to connect a second optical module 221. The second optical module 221 can be connected to the electro-optical conversion device 200 via a plug-in method, or it can be disposed inside the electro-optical conversion device.

[0142] When the link of any of the K first output ports fails, the processing module 210 outputs the electrical signal from any of the first output ports through a third output port. The second optical module 221 performs photoelectric conversion processing on the electrical signals from the multiple first output channels from the third output port and then couples them to an optical port for output.

[0143] See Figure 8 As shown, taking K>1 as an example. In Figure 2B Based on what is shown, Figure 8 The electro-optical conversion device 200 shown may further include a second optical module 221. See also Figure 8 As shown, when the optical port of the first optical module 220-1 malfunctions (or a malfunction is detected), the electrical signal to be output to the first optical module 220-1 is output to the second optical module 221 through the third output port. The second optical module 221 performs photoelectric conversion on the electrical signal from at least one first output channel of the third output port of the processing module 210 and then couples it to an optical port for output. Figure 8 Taking the second optical module 221 as an example, which is installed in the electro-optical conversion device 200, in some implementation scenarios, the second optical module 221 can also be connected to the electro-optical conversion device by plugging and unplugging.

[0144] In one possible implementation, the electro-optical conversion device 200 can simultaneously support lane-level protection and module-level protection.

[0145] See Figure 9 As shown, taking K>1 as an example. In Figure 7 Based on what is shown, Figure 9 The electro-optical conversion device 200 shown may further include a second optical module 221. See also Figure 8 As shown, when the optical port of the first optical module 220-2 malfunctions (or a malfunction is detected), the electrical signal to be output to the first optical module 220-2 is output to the second optical module 221 through the third output port. The second optical module 221 performs photoelectric conversion processing on the electrical signal from at least one first output channel of the third output port of the processing module 210 and then couples it to an optical port for output. Simultaneously, if a first output channel connected to the first optical module 220-1 malfunctions, the electrical signal of that first output channel can be output through the second output channel connected to the first optical module 220-1. Figure 9 Taking the second optical module 221 as an example, which is installed in the electro-optical conversion device 200, in some implementation scenarios, the second optical module 221 can also be connected to the electro-optical conversion device by plugging and unplugging.

[0146] As an example, the electro-optical conversion device 200, while supporting module-level protection, also features optical port multiplexing functionality. Figure 9 Based on the electro-optical conversion device 200 shown Figure 10 The electro-optical conversion device 200 shown also includes a multiplexing module 230. The multiplexing module 230 is connected to the optical ports of K first optical modules 220 and also to the optical port of the second optical module 221, which is responsible for protection. Therefore, during optical module switching, the multiplexing module 230 can perform multiplexing processing on the optical signals of K-1 first optical modules 220 and the optical signals of the second optical module 221.

[0147] In some implementation scenarios, the multiplexing module 230 can also be located outside the electro-optical conversion device 200.

[0148] This application also provides an electro-optical conversion device for performing the reverse operation of the electro-optical conversion equipment 200 described above. This electro-optical conversion device can be a chip, a single board, or a device. The following description will use the device as an example, and will refer to it as an electro-optical conversion equipment.

[0149] See Figure 11A and Figure 11BThe diagram shown is a structural schematic of the electro-optical conversion device 300 provided in an embodiment of this application. The electro-optical conversion device 300 includes a processing module 310. The processing module 310 includes M output ports and K first input ports, where K is a positive integer, M is an integer greater than 1, and M>K; wherein each of the M output ports includes at least one output channel, and the number of output channels included in the M output ports is P; each of the K first input ports includes at least one first input channel, and the number of first input channels included in the K first input ports is Q, where P and Q are positive integers, and P>Q. Each output channel is used to output an electrical signal to the computing node or the communication node, and the electrical signal is used to carry the service data required by the computing node or the communication node. In some possible implementation scenarios, P is an integer multiple of Q, for example, P / Q = X.

[0150] In one possible implementation, the K first input ports are used to plug in and unplug the K first optical modules one-to-one. For example, the K first optical modules can be hot-plugged into the K first input ports of the optical switching device 200.

[0151] In another possible implementation, K first optical modules are disposed inside the optical switching device 200. Each of the K first optical modules corresponds one-to-one with one of the K first input ports. The K first optical modules are fixed inside the electro-optical conversion device via the K first input ports. In this case, the K first input ports can also be understood as the K input ports of the processing module 210.

[0152] Figure 11A In this example, we take K=1 and K third optical modules installed inside the electro-optical conversion device as an example. Figure 11B In this example, we take K>1 and K third optical modules as an example, which are set inside the electro-optical conversion device.

[0153] Figure 11C In this example, we take K=1 and K third optical modules plugged into the K first input ports of the electro-optical conversion device as an example. Figure 11D In the example, K>1 and K third optical modules are inserted into K first input ports of the electro-optical conversion device.

[0154] See Figure 11A As shown, the electro-optical conversion device 300 includes a processing module 310 and a third optical module 320. Q first input channels are connected to the third optical module.

[0155] The third optical module 320 performs photoelectric conversion on the optical signal input from the optical port into at least one electrical signal, and couples this at least one electrical signal to Q first input channels. The processing module 310 performs demultiplexing on the electrical signals from the Q first input channels. Specifically, it demultiplexes the electrical signals input from the Q first input channels to P output channels. The demultiplexing operation can use interleaved demultiplexing, byte deinterleaving, or bit deinterleaving. The Q first input channels belong to a first input port, which is connected to the third optical module 320.

[0156] See Figure 11C As shown, the processing module 310 and the third optical module 320 are connected by a plug-in method.

[0157] The third optical module 320 performs photoelectric conversion on the optical signal input from the optical port into at least one electrical signal, and couples this at least one electrical signal to the first input port. The electro-optical conversion device 300 performs demultiplexing operations on the electrical signals of the Q first input channels of the first input port. Specifically, it can demultiplex the electrical signals input from the Q first input channels to P output channels. The demultiplexing operation can employ interleaved demultiplexing, byte deinterleaving, or bit deinterleaving.

[0158] See Figure 11B As shown, the electro-optical switching equipment 300 includes a processing module 310 and K third optical modules 220. To distinguish the K third optical modules 320, Figure 11B In this example, K third optical modules 320 are designated as third optical module 320-1 to third optical module 320-K.

[0159] One of the Q first input channels is connected to one of the K third optical modules. Each of the K third optical modules is connected to at least one of the Q first input channels. It should be understood that different third optical modules are connected to different first input channels. The number of channels connected to different third optical modules may be the same or different.

[0160] Each third optical module 320 performs photoelectric conversion processing on the optical signal input from the optical port into at least one electrical signal, and couples the at least one electrical signal to at least one first input channel, or in other words, sends the at least one electrical signal one-to-one to at least one first input channel of the processing module. The processing module 310 performs demultiplexing operation on the electrical signals of the Q first input channels. In one possible implementation, when performing the demultiplexing operation, the processing module 310 also slows down the electrical signals of the output channels. After slowing down, the rate of the electrical signal input to any first input channel is higher than the rate of the electrical signal of any output channel.

[0161] See Figure 11C As shown, the processing module 310 is connected to the K third optical modules 320 via a plug-in connection. Each third optical module 320 performs photoelectric conversion processing on the optical signal input from the optical port into at least one electrical signal, and couples at least one electrical signal to the first input port corresponding to that third optical module in the electro-optical conversion device 300. Furthermore, the processing module 310 performs demultiplexing operations on the electrical signals of the Q first input channels.

[0162] In some implementation scenarios, the third optical module 320 can first reduce the speed of the optical signal at the optical port before performing electro-optical conversion. After speed reduction, the rate of the optical signal input to the optical port of the third optical module 320 is greater than the rate of the electrical signal output from the third optical module to any of the first input channels. It should be understood that this speed reduction operation is to match the aforementioned speed-up operation.

[0163] In some implementation scenarios, the number of output channels included in each output port can be the same or different. For example, if each output port includes m output channels, then P = M * m. In some implementation scenarios, M > K. The number of channels included in an output port can be the same or different from the number of channels included in the output port.

[0164] The aforementioned output ports can be used to connect to computing units. These output ports can support the Peripheral Component Interconnect Express (PCIE) standard, Ethernet (ETH) protocol, or other proprietary protocols.

[0165] As an example, see Figure 12 As shown, taking a total of 8 output ports as an example, namely ports 1 to 8, each port has 4 output channels (lanes), i.e., P = 32. Taking the first output port as an example with 4 output channels. Figure 12 Taking K=1 as an example, that is, Q=4.

[0166] As an example, see Figure 13 As shown, taking a total of 32 output ports as an example, namely ports 1 to 32, each port has 4 output channels, so P = 128. Taking the number of first output ports as 4, i.e., K = 4, each first output port includes 4 first output channels. Figure 13 In this example, four third optical modules 320 are designated as third optical module 320-1 to third optical module 320-4.

[0167] In some possible implementations, the electro-optical conversion device 300 can also support forward error correction (FEC) isolation and regeneration in the electrical and optical domains. Specifically, before performing the demultiplexing operation, the processing module 310 performs FEC decoding on the Q-channel electrical signals input through the Q first input channels to obtain Q-channel FEC-decoded electrical signals; demultiplexes the Q-channel FEC-decoded electrical signals into P-channel electrical signals; and encodes the P-channel electrical signals and outputs them through P output channels. Performing error correction processing on the signals before demultiplexing, and then performing FEC regeneration after demultiplexing, can improve data transmission capability.

[0168] In some possible implementations, at least one of the M output ports is a hot-swappable port. The hot-swappable function prevents the processing module 310 from stopping operation in the event of a signal interruption. This method enables fault isolation between the electrical domain of the computing node and the optical domain of the optical switching equipment.

[0169] The devices used in processing module 310 are similar to those used in processing module 210. For details, please refer to the aforementioned description of processing module 210, which will not be repeated here.

[0170] In some possible implementations, to further reduce the cable connections between the electro-optical conversion device 300 and the optical switching device, the number of optical ports connected to the optical switching device can also be reduced. Therefore, the optical signal input from the optical port can be first wavelength-divided into K optical signals.

[0171] See Figure 14 As shown, the electro-optical conversion device 300 may further include a wavelength division multiplexing (WDM) module 330. In this embodiment, the WDM module 330 may be a wavelength selective switch (WSS), an array waveguide grating (AWG), or other devices capable of wavelength division. The WDM module 330 divides one optical signal from the optical switching device into K optical signals, and outputs the K optical signals one-to-one to K third optical modules 320.

[0172] In the above scheme, the wavelength division module 330 reduces the speed of the optical signal during the wavelength division process. After the speed reduction, the speed of the optical signal before wavelength division by the wavelength division module 330 is greater than the speed of the optical signal output after wavelength division.

[0173] In one possible implementation, the electro-optical conversion device 300 can also support channel-level protection. This can be achieved through protection switching or load sharing. For example, using protection switching, each of the K first input ports of the electro-optical conversion device 200 includes at least one second input channel. Then, when any first input channel fails, the processing module 310 switches the reception of the electrical signal from the received input of that first input channel to the received signal from the second input channel. The processing module 210 can be configured with configuration information indicating the correspondence between the first and second input channels. Multiple first input channels may correspond to one input channel. For example, the configuration information may include the correspondence between the channel number of the first input channel and the channel number of the second input channel. In some implementations, multiple first input channels belonging to the same first input port correspond to second input channels belonging to the same first input port. In other implementations, each of the Q first input channels corresponds to one second input channel. The corresponding first and second input channels can belong to the same first input port or different first input ports. In some implementations, the number of second input channels can be multiple. It can protect multiple first input channels simultaneously. For example, if two first input channels fail, the input can be switched to two second input channels. Alternatively, by using load sharing, if a first input channel fails, the processing module 310 can demultiplex the electrical signals of Q-1 first and second input channels onto P output channels.

[0174] See Figure 15 As shown, taking K=1 as an example. In Figure 11A Based on what is shown, Figure 15 The first input port of the electro-optical conversion device 200 shown also includes a second input channel. If one of the Q first input channels connected to the third optical module 320 fails, the reception of electrical signals can be switched from receiving electrical signals through that first input channel to receiving electrical signals through the second input channel.

[0175] See Figure 16 As shown, taking K>1 as an example. In Figure 11B Based on what is shown, Figure 15The electro-optical conversion device 200 shown also includes a second input channel in each of its K first input ports. When a first input channel of any of the K first input ports fails, the electrical signal received from that first input channel can be switched to be received through the second input channel of that first input port. For example, if a first input channel connected to the third optical module 320-1 fails, the reception of the electrical signal from that first input channel can be switched to be received through the second input channel. In some implementation scenarios, there can be multiple second input channels. Multiple first input channels can be protected simultaneously. For example, if two first input channels fail, the reception can be switched to two second input channels.

[0176] In one possible implementation, the electro-optical conversion device 300 can also support module-level protection. Protection can be provided for K third optical modules 320 or optical ports. The electro-optical conversion device 300 may also include at least one third input port, which includes at least one third input channel. A third output port is used to connect a fourth optical module 321. The fourth optical module 321 can be connected to the electro-optical conversion device 300 via a plug-in method, or it can be disposed inside the electro-optical conversion device 300. Figure 16 Taking the fourth optical module 321 installed inside the electro-optical conversion device 300 as an example.

[0177] When the link at any of the K first input ports fails, the processing module 310 switches the reception of electrical signals from any first input port to a third input port. The fourth optical module 321 performs photoelectric conversion processing on the optical signal input from the optical port into at least one electrical signal, and couples the at least one electrical signal to the third input port.

[0178] See Figure 17 As shown, taking K>1 as an example. In Figure 11B Based on the given 0, Figure 17 The electro-optical conversion device 300 shown may further include a fourth optical module 321. See also Figure 17 As shown, when a fault occurs (or a fault is detected) in the optical port of the third optical module 320-1, the electrical signal received from the third optical module 320-1 will be switched to be received from the fourth optical module 321. The fourth optical module 321 will perform photoelectric conversion processing on the optical signal input from the optical port into at least one electrical signal, and couple the at least one electrical signal to the third input port of the processing module.

[0179] Figure 17 Taking the fourth optical module 321 as an example, which is installed in the electro-optical conversion device 300, in some implementation scenarios, the fourth optical module 321 can also be connected to the electro-optical conversion device by plugging and unplugging.

[0180] In one possible implementation, the electro-optical conversion device 300 can simultaneously support lane-level protection and module-level protection.

[0181] See Figure 18 As shown, taking K>1 as an example. In Figure 16 Based on the electro-optical conversion device 200 shown, Figure 18 The electro-optical conversion device 300 shown may further include a fourth optical module 321. See also Figure 18 As shown, when the optical port of the third optical module 320-2 malfunctions (or a malfunction is detected), the electrical signal received from the third optical module 320-2 is received through the fourth optical module 321. The fourth optical module 321 converts at least one optical signal input from the optical port into at least one electrical signal and couples the at least one electrical signal to the third input port of the processing module. Simultaneously, if a first input channel connected to the third optical module 320-1 malfunctions, the electrical signal from that first input channel can be input through a second input channel connected to the third optical module 320-1.

[0182] As an example, the electro-optical conversion device 300, while supporting module-level protection, also features optical port wavelength division multiplexing (WDM). Figure 18 On this basis, Figure 19 The electro-optical conversion device 300 shown also includes a wavelength division multiplexing (WDM) module 330. The WDM module 330 is connected to the optical ports of the K third optical modules 320, and also to the optical port of the fourth optical module 321 which is responsible for protection. Therefore, during optical module switching, the WDM module 330 can divide the optical signal input to the optical port and then couple it to the K-1 third optical modules 320 and the fourth optical module 321.

[0183] In this embodiment of the application, an electro-optical conversion device 400 is also provided, which can realize the functions of both electro-optical conversion devices 200 and 300. For example, the electro-optical conversion device includes a processing module 410 and an optical module 420. The processing module 410 has the functions of both processing module 210 and processing module 310. The optical module 420 can have the functions of both optical module 220 and optical module 320.

[0184] In some implementation scenarios, see Figure 20As shown, the electro-optical conversion device 400 includes a processing module 410 and K optical modules 420. The processing module 410 may have M input ports, M output ports, K first input ports, and K first output ports. Each of the M input ports includes at least one input channel, and the number of input channels in the M input ports is P. Each of the K first output ports includes at least one first output channel, and the number of first output channels in the K first output ports is Q. Each of the M output ports includes at least one output channel, and the number of output channels in the M output ports is P. Each of the K first input ports includes at least one first input channel, and the number of first input channels in the K first input ports is Q. The relevant descriptions of the P input channels, P output channels, Q first input channels, and Q first output channels are as described above and will not be repeated here. The K optical modules 420 possess the functions of both the third optical module 320 and the first optical module 220. The K optical modules 420 can be connected to the electro-optical conversion device 400 via plug-in connection, or they can be installed inside the electro-optical conversion device 400.

[0185] In some implementation scenarios, see Figure 21 As shown, the electro-optical conversion device 400 may also include a multiplexing / splitting module 430, which has the functions of the multiplexing module 230 and the splitting module 330, as described above, and will not be repeated here.

[0186] In some implementation scenarios, see Figure 22 As shown, the electro-optical conversion device 400 may also have channel-level protection and / or optical module-level protection functions. Specific details are as described above and will not be repeated here. Figure 22 The optical module used for protection is optical module 421.

[0187] The electro-optical conversion device provided in this application embodiment can be applied to computing nodes. See also... Figure 23 As shown, a computing node may include an electro-optical conversion device and N computing units. Figure 23 Taking N=64 as an example.

[0188] Taking a transmitting node as an example, each of the N computing units includes L output ports, and the N*L output ports of the N computing units are connected one-to-one with the N*L input ports of the electro-optical conversion device. The electro-optical conversion device can be an electro-optical conversion device 200. N*L <= M. The N computing units are connected to the electro-optical conversion device via PCB traces or cables.

[0189] Taking a receiving-side node as an example, each of the N computing units includes L input ports, and the N*L input ports of the N computing units are connected one-to-one with the N*L output ports of the electro-optical conversion device. The electro-optical conversion device can be an electro-optical conversion device 300. N*L <= M. The N computing units are connected to the electro-optical conversion device via PCB traces or cables.

[0190] Taking a computing node that functions as both a transmitting and receiving node as an example, each of the N computing units includes L input ports and L output ports. The N*L output ports of the N computing units are connected one-to-one with the N*L input ports of the electro-optical conversion device. The electro-optical conversion device can be electro-optical conversion device 400. Alternatively, it can integrate the structures of electro-optical conversion device 200 and electro-optical conversion device 300. The N computing units are connected to the electro-optical conversion device via PCB traces or cables.

[0191] In some implementation scenarios, the electro-optical conversion device provided in this application embodiment can be applied to a communication node. The communication node may include the electro-optical conversion device and N communication chips. See also... Figure 24 As shown, taking the OTN chip as an example, the communication node is used.

[0192] Taking a transmitting node as an example, each of the N communication chips includes L output ports, and the N*L output ports of the N communication chips are connected one-to-one with the N*L input ports of the electro-optical conversion device. The electro-optical conversion device can be an electro-optical conversion device 200. N*L <= M. The N communication chips are connected to the electro-optical conversion device via PCB traces or cables.

[0193] Taking a receiving node as an example, each of the N communication chips includes L input ports, and the N*L input ports of the N communication chips are connected one-to-one with the N*L output ports of the electro-optical conversion device. The electro-optical conversion device can be an electro-optical conversion device 300. N*L <= M. The N communication chips are connected to the electro-optical conversion device via PCB traces or cables.

[0194] Taking a communication node that functions as both a transmitting and receiving node as an example, each of the N communication chips includes L input ports and L output ports. The N*L output ports of the N communication chips are connected one-to-one with the N*L input ports of the electro-optical conversion device. The electro-optical conversion device can be electro-optical conversion device 400. Alternatively, it can integrate the structures of electro-optical conversion device 200 and electro-optical conversion device 300. The N communication chips are connected to the electro-optical conversion device via PCB traces or cables.

[0195] See Figure 25 The illustration shows a computing cluster provided in this application embodiment. The computing cluster includes a first computing node, a second computing node, and an optical switching device. The first computing node can be a transmitting node, and the second computing node can be a receiving node; or, the first computing node can be a receiving node, and the second computing node can be a transmitting node. Alternatively, the first computing node can function as both a transmitting and receiving node, and the second computing node can function as both a receiving and transmitting node.

[0196] Optical switching equipment can be either an optical crossconnect (OXC) using port-level optical switching technology or a wavelength-level optical switching technology. For example, wavelength-level OXCs can be WSSs, such as liquid crystal on silicon (LCoS), liquid crystals (LC), or micro-electro-mechanical systems (MEMS). Port-level OXCs (or port-level OCSs) can use MEMS, LC, piezoelectric ceramics, or waveguides.

[0197] In some embodiments, the network management device can generate a service configuration table. For example, the service configuration table records the number of the computing unit, the numbers of the input and output ports on the electro-optical conversion device to which the computing unit is connected, and the numbers of the input and output ports of the optical switching device to which the optical port of the electro-optical conversion device is connected. Furthermore, in this embodiment, when the electro-optical conversion device of the computing node performs multiplexing or demultiplexing operations, electrical signals from multiple input channels of the same computing unit are multiplexed and then output through a single output channel. The electro-optical conversion device can perform multiplexing or demultiplexing operations based on a routing table.

[0198] In some implementation scenarios, see Figure 26As shown, the computing cluster includes multiple computing nodes and multiple optical switching devices. The multiple computing nodes can adopt the computing node structure provided above, which will not be elaborated further.

[0199] The embodiments of this application can also be applied to communication clusters, such as communication clusters including one or more communication nodes and one or more switching devices. Multiple communication nodes can adopt the structure of the communication nodes provided above, and specific details will not be repeated here.

[0200] Based on the above embodiments, this application also provides an electro-optical conversion method. See [link to previous document]. Figure 27 The diagram shown is a schematic flowchart of an electro-optical conversion method provided in an embodiment of this application. This method is applied to the aforementioned electro-optical conversion device. The method includes steps S2701 and S2702.

[0201] S2701 receives electrical signals input from the P input channels through the M input ports.

[0202] S2702, the processing module multiplexes the electrical signals of the P input channels to the Q first output channels for output.

[0203] For details of other technical solutions and effects of the embodiments of this application, please refer to the descriptions in other embodiments of this application. Exemplarily, the method of the embodiments of this application can be executed using any of the electro-optical conversion devices described in the foregoing embodiments.

[0204] See Figure 28 The diagram shown is a schematic flowchart of another electro-optical conversion method provided in an embodiment of this application. This method is applied to the aforementioned electro-optical conversion device. The method includes steps S2801 and S2802.

[0205] S2801, receiving an electrical signal from the first computing unit in the computing node through the first input port of the processing module.

[0206] S2802, the processing module multiplexes the electrical signals of at least one input channel included in the first input port to the target output channel among the Q first output channels.

[0207] The method may further include: performing photoelectric conversion on the electrical signal from the target output channel through the target optical module connected to the target output channel among the K first optical modules, and then coupling it to an optical port for output.

[0208] For details of other technical solutions and effects of the embodiments of this application, please refer to the descriptions in other embodiments of this application. Exemplarily, the method of the embodiments of this application can be executed using any of the electro-optical conversion devices described in the foregoing embodiments.

[0209] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, etc.) containing computer-usable program code.

[0210] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0211] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. An electro-optical conversion device, characterized in that, This device, applied to computing nodes or communication nodes, includes a processing module, M input ports, and K first output ports, where K is a positive integer, M is an integer greater than 1, and M>K. Each of the M input ports includes at least one input channel, and the number of input channels in the M input ports is P. Each of the K first output ports includes at least one first output channel, and the number of first output channels in the K first output ports is Q, where P and Q are positive integers, P>Q, and P is an integer multiple of Q. Each input channel is used to receive electrical signals, which are used to carry service data of the computing node or the communication node. The processing module is used to multiplex the electrical signals of the P input channels to the Q first output channels for output through the K first output ports.

2. The device as described in claim 1, characterized in that, The K first output ports are used to plug in and unplug the K first optical modules one by one; Any one of the K first optical modules is used to perform photoelectric conversion on the electrical signal from at least one first output channel of the corresponding first output port and then couple it to an optical port for output.

3. The device as described in claim 1, characterized in that, The electro-optical conversion device further includes K first optical modules corresponding one-to-one with the K first output ports; the K first optical modules are fixed inside the electro-optical conversion device through the K first output ports; Any one of the K first optical modules is used to perform photoelectric conversion on the electrical signal from at least one first output channel of the corresponding first output port and then couple it to an optical port for output.

4. The device as described in claim 2 or 3, characterized in that, The rate of the optical signal output from the optical port of any first optical module is greater than the rate of the electrical signal output from the first output channel received by any first optical module.

5. The device as described in any one of claims 2-4, characterized in that, If K>1, the device also includes a wave multiplexing module; The multiplexing module is used to multiplex the K optical signals output by the K first optical modules and then output them.

6. The device according to any one of claims 1-4, characterized in that, The rate of the electrical signal in any of the first output channels is greater than the rate of the electrical signal in any of the input channels.

7. The device according to any one of claims 1-5, characterized in that, At least one of the M input ports is a hot-swappable port.

8. The device according to any one of claims 1-7, characterized in that, The processing module is specifically used for: Multiplex the electrical signals of P input channels into Q multiplexed electrical signals; The Q multiplexed electrical signals are FEC encoded and then output through the Q first output channels.

9. The device according to any one of claims 1-7, characterized in that, The processing module is specifically used for: Forward error correction (FEC) decoding is performed on the electrical signals of the P input channels; The electrical signals from the P input channels after FEC decoding are multiplexed into Q multiplexed electrical signals; The Q multiplexed electrical signals are FEC encoded and then output through the Q first output channels.

10. The device according to any one of claims 1-9, characterized in that, Each of the K first output ports also includes at least one second output channel; The processing module is further configured to, when any of the Q first output channels fails, output the electrical signal of any first output channel through the target output channel corresponding to any of the first output channels in the second output channels included in the K first output ports, according to the configuration information.

11. The device according to any one of claims 1-10, characterized in that, K>1, the device further includes at least one second output port, each of the at least one second output port including a plurality of third output channels; The processing module is further configured to: When the link corresponding to any of the K first output ports fails, the electrical signal to be output to any of the first output ports will be output through one of the at least one second output ports.

12. An electro-optical conversion device, characterized in that, This device, applied to computing nodes or communication nodes, includes a processing module, M output ports, and K first input ports, where K is a positive integer, M is an integer greater than 1, and M>K. Each of the M output ports includes at least one output channel, and the number of output channels in the M output ports is P. Each of the K first input ports includes at least one first input channel, and the number of first input channels in the K first input ports is Q, where P and Q are positive integers, P>Q, and P is an integer multiple of Q. Each output channel is used to output an electrical signal to the computing node or communication node, and the electrical signal is used to carry the service data of the computing node or communication node. The processing module is used to receive electrical signals through the K first input ports and the Q first input channels, and to demultiplex the electrical signals received through the Q first input channels to the P output channels for output.

13. The device as claimed in claim 12, characterized in that, The K first input ports are used to plug in and unplug the K third optical modules one by one; Any one of the K third optical modules is used to perform photoelectric conversion processing on the optical signal input from the optical port into at least one electrical signal, and output the at least one electrical signal to at least one first input channel of the corresponding first input port.

14. The device as claimed in claim 12, characterized in that, The electro-optical conversion device further includes K third optical modules corresponding one-to-one with the K first input ports; the K third optical modules are fixed inside the electro-optical conversion device through the K first input ports; Any one of the K third optical modules is used to perform photoelectric conversion processing on the optical signal input from the optical port into at least one electrical signal, and output the at least one electrical signal to at least one first input channel of the corresponding first input port.

15. The device as described in claim 13 or 14, characterized in that, The rate of the optical signal input to the optical port of any third optical module is greater than the rate of the electrical signal output from any third optical module to any first input channel.

16. The device according to any one of claims 13-15, characterized in that, K>1, and the device also includes a wavelength division module; The wavelength division module is used to divide one optical signal from the optical switching device into K optical signals, and output the K optical signals one by one to the K third optical modules.

17. The device according to any one of claims 12-16, characterized in that, The rate of the electrical signal in any of the first input channels is greater than the rate of the electrical signal in any of the output channels.

18. The device according to any one of claims 12-17, characterized in that, At least one of the M output ports is a hot-swappable port.

19. The device according to any one of claims 12-18, characterized in that, The processing module is specifically used for: The electrical signals of the Q first input channels are demultiplexed into P electrical signals; The P-channel electrical signals are encoded using forward error correction (FEC) and then output through the P output channels.

20. The device according to any one of claims 12-18, characterized in that, The processing module is specifically used for: The electrical signals input through the Q first input channels are subjected to FEC decoding to obtain Q-channel FEC-decoded electrical signals; The electrical signal decoded by the Q-channel FEC is demultiplexed into a P-channel electrical signal; The P-channel electrical signals are encoded and then output through the P output channels.

21. The device according to any one of claims 12-20, characterized in that, Each of the K first input ports also includes at least one second input channel; The processing module is further configured to, when any of the Q first input channels fails, switch the received electrical signal from any of the first input channels to the target input channel corresponding to any of the first input channels in the second output channels included in the K first input ports, according to the configuration information.

22. The device according to any one of claims 12-21, characterized in that, K>1, the device further includes at least one second input port, each of the at least one second input port including a plurality of third input channels; The processing module is further configured to: When a link failure occurs at any of the K first input ports, the electrical signal received from any of the first input ports will be switched to be received from one of the at least one second input ports.

23. A first computing node, characterized in that, Includes the electro-optical conversion device as described in any one of claims 1-11 and N computing units; Each of the N computing units includes L output ports, and the N*L output ports of the N computing units are connected one-to-one with the N*L input ports of the electro-optical conversion device, wherein N*L is less than or equal to M.

24. A second computing node, characterized in that, Includes the electro-optical conversion device as described in any one of claims 12-22 and N computing units; Each of the N computing units includes L input ports, and the N*L input ports of the N computing units are connected one-to-one with the N*L output ports of the electro-optical conversion device, wherein N*L is less than or equal to M.

25. The computing node as described in claim 23 or 24, characterized in that, The N computing units are connected to the electro-optical conversion device via printed circuit board (PCB) traces or via cables.

26. A computing cluster, characterized in that, Includes the first computing node as described in claim 23 and the optical switching device; The J input ports on the optical switching device are connected to the J optical ports on the first computing node, where J = 1 or J = K; The optical switching device is used to switch the optical signal received from the first computing node through the first input port of the J input ports to the first output port of the optical switching device; The first output port is the target output port of the J output ports on the optical switching device corresponding to the first input port.

27. The computing cluster as claimed in any one of claims 26, characterized in that, It also includes the second computing node as described in claim 24; The J output ports on the optical switching device are connected to the J optical ports on the second computing node.

28. An electro-optical conversion method, characterized in that, The method is applied to an electro-optical conversion device, which is applied to a computing node. The electro-optical conversion device includes a processing module, M input ports, and K first output ports, where K is a positive integer, M is an integer greater than 1, and M>K. Each of the M input ports includes at least one input channel, and the number of input channels in the M input ports is P. Each of the K first output ports includes at least one first output channel, and the number of first output channels in the K first output ports is Q, where P and Q are positive integers, P>Q, and P is an integer multiple of Q. Each input channel is used to receive electrical signals, which are used to carry service data of the computing node or the communication node. The method includes: The electrical signals input through the P input channels are received through the M input ports; The processing module multiplexes the electrical signals of the P input channels to the Q first output channels, so as to output them through the K first output ports.

29. An electro-optical conversion method, characterized in that, The method is applied to an electro-optical conversion device, which is applied to a computing node. The electro-optical conversion device includes a processing module, M input ports, and K first output ports, where K is a positive integer, M is an integer greater than 1, and M>K. Each of the M input ports includes at least one input channel, and the number of input channels in the M input ports is P. Each of the K first output ports includes at least one first output channel, and the number of first output channels in the K first output ports is Q, where P and Q are positive integers, P>Q, and P is an integer multiple of Q. Each input channel is used to receive electrical signals, which are used to carry service data of the computing node or the communication node. The processing module is used to multiplex the electrical signals from the P input channels to the Q first output channels for output through the K first output ports. The method includes: The processing module receives electrical signals from the first computing unit in the computing node through its first input port. The processing module multiplexes the electrical signals of at least one input channel included in the first input port to the target output channel among the Q first output channels.

30. A chip, characterized in that, The chip is used to implement the method as described in claim 28 or 29.

Images