Optical fiber link-based computing power topology generation method, application method and generation system

Abstract

The invention relates to the technical field of optical fiber link intelligent scheduling, in particular to a computing power topology generation method, application method and generation system based on an optical fiber link, and the generation method comprises the steps: obtaining the port connection information of an optical switch in an optical fiber network, and constructing an optical switch topology network according to the port connection information; obtaining computing power connection demand information; searching an optical switch link path in the optical switch topology network according to the computing power connection demand information and a preset computing power node-optical switch node mapping relation; obtaining optical switch routing configuration corresponding to the optical switch link path; and constructing the computing power topology network according to the computing power connection demand information and the corresponding optical switch routing configuration. Compared with the prior art, the complex optical fiber network is converted into the simple and abstract computing power topology network, and the optical switch connection path in the optical fiber network is utilized to represent the connection relationship in the computing power topology network, so that the optical fiber network is effectively utilized to construct the computing power topology graph.

Description

A method, application method, and generation system for generating computing topology based on fiber optic links. Technical Field

[0001] This invention relates to the field of intelligent scheduling technology for optical fiber links, and more specifically, to a method, application method, and generation system for generating computing topology based on optical fiber links. Background Technology

[0002] With the continuous development of artificial intelligence (AI) technology, the cost of training and local deployment of AI models is decreasing, making it less necessary for AI computing power to be highly centralized and large-scale. Meanwhile, AI training still requires ensuring data security. To address this, a storage-compute separation technology has been proposed, connecting multiple computing centers and local computing clusters via optical links. Leveraging the high-speed transmission capabilities of optical fibers, high-speed scheduling of computing power is achieved.

[0003] Existing computing power allocation and scheduling schemes are mainly implemented by constructing a topology network of computing power centers. However, the connection of fiber optic networks is complex, making it difficult to directly use fiber optic networks to construct the topology network of computing power centers, and thus difficult to directly use fiber optic networks to achieve computing power allocation and scheduling. Summary of the Invention

[0004] This invention provides a computing power topology generation method, application method, and generation system based on optical fiber links, which can effectively utilize optical fiber networks to construct computing power topology networks.

[0005] According to a first aspect of this application, a method for generating computing power topology based on optical fiber links is provided. The method includes: obtaining port connection information of optical switches in an optical fiber network, and constructing an optical switch topology network based on the port connection information; obtaining computing power connection requirement information; searching for optical switch link paths in the optical switch topology network based on the computing power connection requirement information and a preset mapping relationship between computing power nodes and optical switch nodes; obtaining the optical switch routing configuration corresponding to the optical switch link path; and constructing the computing power topology network based on the computing power connection requirement information and the corresponding optical switch routing configuration.

[0006] Optionally, the computing power connection requirement information includes a starting computing power node and a target computing power node; the step of searching for an optical switch link path in the optical switch topology network based on the computing power connection requirement information and a preset computing power node-optical switch node mapping relationship includes: obtaining, based on the computing power node-optical switch node mapping relationship, the starting optical switch node corresponding to the starting computing power node of the computing power connection requirement information in the optical switch topology network, and obtaining the target optical switch node corresponding to the target computing power node of the computing power connection requirement information in the optical switch topology network; and searching for an optical switch link path in the optical switch topology network that allows the starting optical switch node to reach the corresponding target optical switch node.

[0007] Optionally, the step of searching for optical switch link paths in the optical switch topology network that can reach the corresponding target optical switch node includes: using a preset path search algorithm to obtain candidate optical switch link paths that can reach the corresponding target optical switch node from the starting optical switch node; obtaining the path length and average bandwidth of the candidate optical switch link paths; obtaining the average transmission delay of the optical switch link paths based on the path length, average bandwidth, and a preset computing power transmission value; and selecting the optical switch link path from the candidate optical switch link paths based on the average transmission delay.

[0008] Optionally, obtaining the optical switch routing configuration corresponding to the optical switch link path includes: pre-encoding the optical switches traversed by the optical switch link path and the ports corresponding to the optical switches according to preset labels of the optical switches; obtaining the optical switches included in the optical switch link path and their corresponding ports; and obtaining the routing encoding vector of the optical switch link path as the optical switch routing configuration based on the encoding of the included optical switches and their corresponding ports. Optionally, the computing power connection requirement information is preset with wavelength labels; constructing the computing power topology network based on the computing power connection requirement information and the corresponding optical switch routing configuration includes dividing the computing power connection requirement information according to the wavelength labels to obtain several sets of computing power connection requirement information; the wavelength labels of the computing power connection requirement information in the same set of computing power connection requirement information are the same; constructing a computing power topology sub-network of the computing power connection requirement information according to the computing power connection requirement information in the set of computing power connection requirement information and the corresponding optical switch routing configuration; and using the set of computing power topology sub-networks as the computing power topology network.

[0009] Optionally, the step of constructing a computing power topology sub-network based on the computing power connection demand information in the computing power connection demand information set and the corresponding optical switch routing configuration includes: for the computing power connection demand information, using the corresponding optical switch routing configuration as an edge attribute, constructing a directed edge from the starting computing power node to the target computing power node; using the starting computing power node and the target computing power node in the computing power connection demand information as network nodes, and constructing the computing power topology sub-network based on the network nodes and the corresponding directed edge.

[0010] According to a second aspect of this application, a computing power topology application method based on optical fiber links is provided. The application method is based on a computing power topology network generated by the computing power topology generation method based on optical fiber links described in the first aspect above. The application method includes: obtaining computing power allocation demand information, and obtaining a starting allocation computing power node and a target allocation computing power node according to the computing power allocation demand information; obtaining corresponding optical switch routing configurations from the computing power topology network according to the starting allocation computing power node and the target allocation computing power node; and adjusting the port connection status of the corresponding optical switches in the optical switch topology network according to the optical switch routing configurations.

[0011] According to a third aspect of this application, a computing power topology generation system based on optical fiber links is provided. The generation system includes: an optical switch network construction module, used to acquire port connection information of optical switches in an optical fiber network and construct an optical switch topology network based on the port connection information; a demand acquisition module, used to acquire computing power connection demand information; a path search module, used to search for optical switch link paths in the optical switch topology network based on the computing power connection demand information and a preset computing power node-optical switch node mapping relationship; an optical switch configuration module, used to acquire the optical switch routing configuration corresponding to the optical switch link path; and a computing power topology network construction module, used to construct the computing power topology network based on the computing power connection demand information and the corresponding optical switch routing configuration.

[0012] According to a fourth aspect of this application, an electronic device is provided, comprising: a memory for storing one or more computer programs; and a processor, which, when the one or more computer programs are executed by the processor, implements the fiber optic link-based computing topology generation method described in the first aspect or the fiber optic link-based computing topology application method described in the second aspect.

[0013] According to a fifth aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions, the computer instructions being configured to cause a processor to execute and implement the computing power topology generation method based on fiber optic links as described in the first aspect or the computing power topology application method based on fiber optic links as described in the second aspect.

[0014] Based on any of the above aspects, the computing power topology generation method, application method, generation system, electronic device, and computer storage medium based on optical fiber links provided in this application, by mapping the computing power connection requirement information to an optical switch topology network, performing path search in the optical switch topology network based on the computing power connection requirement information, determining the optical switch routing configuration corresponding to the computing power connection requirement information, and finally constructing a computing power topology network using the computing power connection requirement information and the optical switch routing configuration, the computing power topology network reflects both the connection relationship between computing power nodes and the optical switch routing configuration on which the corresponding connection relationship is based, making the allocation of computing power directly associated with the configuration of optical switches, realizing the construction of a computing power topology network through an optical fiber network, and greatly reducing the difficulty of computing power allocation and scheduling. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 is a schematic diagram of the steps of the generation method provided in this embodiment.

[0017] Figure 2 is a schematic diagram of the steps for constructing the computing power topology network provided in this embodiment.

[0018] Figure 3 is a schematic diagram of the steps of the application method provided in this embodiment.

[0019] Figure 4 is a schematic diagram of the functional modules of the generation system provided in this embodiment.

[0020] Figure 5 is a schematic diagram of the device structure of the electronic device provided in this embodiment. Detailed Implementation

[0021] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this application. To better illustrate the following embodiments, some components in the drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions of the product; it is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0022] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0023] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0024] With the continuous development of artificial intelligence (AI) technology, the cost of training and local deployment of AI models is decreasing, making it less necessary for AI computing power to be highly centralized and large-scale. Meanwhile, AI training still requires ensuring data security. To address this, a storage-compute separation technology has been proposed, connecting multiple computing centers and local computing clusters via optical links. Leveraging the high-speed transmission capabilities of optical fibers, high-speed scheduling of computing power is achieved.

[0025] Existing computing power allocation and scheduling schemes are mainly implemented by constructing a topology network of computing power centers. However, the connection of fiber optic networks is complex, making it difficult to directly use fiber optic networks to construct the topology network of computing power centers, and thus difficult to directly use fiber optic networks to achieve computing power allocation and scheduling.

[0026] This embodiment provides a technical solution that can solve the above problems. The specific implementation of this application will be described in detail below with reference to the accompanying drawings.

[0027] As shown in Figure 1, this embodiment provides a computing topology generation method based on optical fiber links, which may include the following steps: S1: Obtain the port connection information of optical switches in the optical fiber network, and construct an optical switch topology network according to the port connection information; In this embodiment, the port connection information may include the port label of the corresponding port of the optical switch and the port label connected to the port through optical fiber. For example, for optical switch A, it includes input port 1, the port label of input port 1 is A1, and output port 1 and output port 2, the port label of input port 1 is A2, the port label of output port 2 is A3, input port 1 is connected to output port 1 of optical switch B, the port label of output port 1 of optical switch B is B2, output port 1 of optical switch A is connected to input port 1 of optical switch C, the port label of input port 1 of optical switch C is C1, output port 2 of optical switch A is connected to input port 1 of optical switch D, the port label of input port 1 of optical switch D is D1, then the port connection information of optical switch A can be represented as [A1-B2, A2-C1, A3-D1].

[0028] Understandably, the port tag is used to uniquely identify the corresponding port, and the port tag is set when constructing the fiber optic network and connecting the fiber optic cable to the port. In one embodiment, the port tag can be set as an RFID (Radio Frequency Identification) tag.

[0029] By acquiring the connection port information of the optical switch, the optical switch can be used as a network node, and a directed connection edge corresponding to the optical switch can be generated according to the connection port information. The optical switch topology network can be constructed based on the network node of the optical switch and the corresponding directed connection edge. It can be understood that the optical switch is an important device in the optical fiber network. The optical fiber network realizes the switching of optical paths through the optical switches to adjust the connection relationship between different optical fiber devices. The computing center is connected to the optical fiber network through the optical switch. Therefore, the connection relationship between the computing centers can be represented by the optical switch topology network.

[0030] S2: Obtain computing power connection requirement information; In this embodiment, the computing power connection requirement information is a set of virtual connection requests defined for constructing a computing power topology network, representing potential communication needs between computing power centers. It is understood that the computing power connection requirement information does not directly correspond to actual data transmission tasks, but is used to guide the generation of optical switch routing configurations and the construction of the computing power topology sub-network. The computing power connection requirement information consists of a starting computing power node and a target computing power node, indicating that computing power needs to be scheduled from the starting computing power node to the target computing power node.

[0031] S3: Based on the computing power connection requirement information and the preset computing power node-optical switch node mapping relationship, search for optical switch link paths in the optical switch topology network. In this embodiment, the computing power node-optical switch node mapping relationship can be obtained in advance based on the physical location of the computing power center and the physical location of the optical switch. Computing power centers and optical switches with the same physical location are associated as network nodes, and the associated computing power centers and optical switches are used to construct a computing power node-optical switch node mapping. It is understood that there is a specific connection relationship between computing power centers and optical switches with the same physical location. Computing power scheduled from the outside to the computing power center needs to pass through the corresponding optical switch. Therefore, the optical switch can be used to map the computing power scheduling of the corresponding computing power center. One computing power center can correspond to multiple optical switches.

[0032] As described above, the computing power connection requirement information includes a starting computing power node and a target computing power node. Therefore, in this embodiment, step S3 may include the following steps: according to the computing power node-optical switch node mapping relationship, obtain the starting optical switch node corresponding to the starting computing power node of the computing power connection requirement information in the optical switch topology network, and obtain the target optical switch node corresponding to the target computing power node of the computing power connection requirement information in the optical switch topology network; in the optical switch topology network, search for the optical switch link path from the starting optical switch node to the corresponding target optical switch node.

[0033] In this embodiment, searching for optical switch link paths in the optical switch topology network that can reach the corresponding target optical switch node from the starting optical switch node may include the following steps: using a preset path search algorithm to obtain candidate optical switch link paths that can reach the corresponding target optical switch node from the starting optical switch node; obtaining the path length and average bandwidth of the candidate optical switch link paths; obtaining the average transmission delay of the optical switch link paths based on the path length, average bandwidth, and a preset computing power transmission value; and selecting the optical switch link path from the candidate optical switch link paths based on the average transmission delay.

[0034] Among them, the candidate optical switch link path with the minimum average transmission delay can be used as the optical switch link path.

[0035] In this embodiment, the preset path search algorithm can be either breadth-first search or depth-first search. In some implementations, the path search algorithm can be set to the A* search algorithm.

[0036] Furthermore, in some embodiments, the average transmission delay, combined with the number of optical switch nodes traversed by the candidate optical switch link path, can be used as the basis for selecting the optical switch link path. Specifically, the candidate optical switch link path with the shortest average transmission delay and the fewest number of optical switch nodes can be selected as the optical switch link path.

[0037] S4: Obtain the optical switch routing configuration corresponding to the optical switch link path; In this embodiment, step S4 may include the following steps: pre-encode the optical switches traversed by the optical switch link path and the ports corresponding to the optical switches according to the preset tags of the optical switches; obtain the optical switches and their corresponding ports included in the optical switch link path; obtain the routing encoding vector of the optical switch link path as the optical switch routing configuration according to the encoding of the included optical switches and their corresponding ports.

[0038] In this embodiment, the preset label of the optical switch may include an optical switch label and port labels of each port in the optical switch. Then, the encoding of the optical switches traversed by the optical switch link path and the corresponding ports can be performed by connecting the optical switch label, the input port label and the output port label of each optical switch traversed by the optical switch link path in a preset order to obtain an optical switch sub-vector. Then, the optical switch sub-vectors of all optical switches in the optical switch link path are connected in an orderly manner according to the traversal order of the optical switch link path to obtain the routing encoding vector.

[0039] For example, if the optical switch link path passes through optical switch A and optical switch B, and passes through the input port 1 and output port 2 of optical switch A, and the input port 2 and output port 2 of optical switch B, wherein the optical switch label of optical switch A is A, the port label of its input port 1 is A1, and the port label of its output port 2 is A3, the optical switch label of optical switch B is B, the port label of its input port 2 is B2, and the port label of its output port 2 is B4, then the optical switch sub-vector of each optical switch is constructed in the order of optical switch, input port, and output port. The optical switch sub-vector of optical switch A is [A, A1, A3], and the optical switch sub-vector of optical switch B is [B, B2, B4]. Then the routing coding vector of the optical switch link path can be [A, A1, A3, B, B2, B4].

[0040] Understandably, optical switches adjust the optical path in a fiber optic network by switching the connection relationship between their input and output ports. Therefore, by encoding and vectorizing the optical switch link path, when computing power scheduling is required, the optical switches traversed by the optical path corresponding to the optical switch link path and the switching state of each optical switch can be quickly determined based on the encoded and vectorized optical switch routing configuration. This allows for direct and rapid adjustment of the switching state of the optical switches, ensuring efficient completion of computing power scheduling.

[0041] S5: Construct a computing power topology network based on the computing power connection requirement information and the corresponding optical switch routing configuration.

[0042] In this embodiment, the computing power connection request information is pre-labeled with a wavelength tag. This wavelength tag distinguishes different computing power connection request information and avoids conflicts between them. It is understood that the wavelength tag identifies the wavelength channel occupied by the optical switch link path corresponding to the computing power connection request information when transmitting optical signals. Although the computing power connection request information is a virtual connection request, pre-labeling it with a corresponding wavelength tag better simulates computing power transmission and provides a foundation for the subsequent practical use of the computing power topology network, enabling it to use optical signals of the corresponding wavelength based on the wavelength tag and avoiding optical signal conflicts during multi-task transmission.

[0043] Therefore, in this embodiment, as shown in FIG2, step S5 may include the following steps: S51: Divide the computing power connection demand information according to the wavelength label to obtain several sets of computing power connection demand information; the wavelength labels of the computing power connection demand information in the same set of computing power connection demand information are the same; S52: Construct a computing power topology sub-network of the computing power connection demand information according to the computing power connection demand information in the set of computing power connection demand information and the corresponding optical switch routing configuration; S53: Use the set of computing power topology sub-networks as the computing power topology network.

[0044] Understandably, in this embodiment, the computing power connection requirement information with the same wavelength label is divided into the same computing power connection requirement information set, so that corresponding computing power topology sub-networks can be independently constructed based on the wavelength label. Subsequently, in the actual application process, if computing power scheduling needs arise in the computing power topology sub-network, the optical signal corresponding to the wavelength label can be used to transmit the resources required for computing power calculation. Since connections under the same wavelength share the same optical channel resources, aggregating and processing the computing power connection requirement information with the same wavelength label can ensure the consistency of channel transmission within the computing power topology sub-network, thereby avoiding cross-wavelength resource conflicts. At the same time, computing power topology sub-networks corresponding to different wavelength labels are naturally isolated in the optical domain, which facilitates the realization of multi-task parallel scheduling and fault domain separation.

[0045] In one implementation, step S52 may include: for the computing power connection requirement information, using the corresponding optical switch routing configuration as an edge attribute, constructing a directed edge from the starting computing power node to the target computing power node; using the starting computing power node and the target computing power node in the computing power connection requirement information as network nodes, and constructing the computing power topology sub-network based on the network nodes and the corresponding directed edge.

[0046] Understandably, in this embodiment, the optical switch routing configuration expresses the switching state of the optical switch under the corresponding optical path. Therefore, by using the optical switch routing configuration as the edge attribute to construct the directed edge from the starting computing power node to the target computing power node of the computing power connection requirement information, the directed connection relationship between the starting computing power node, the target computing power node, and the directed edge of the computing power connection requirement information can be obtained. In this way, the directed connection relationship of all the computing power connection requirement information can be integrated to obtain the computing power topology sub-network.

[0047] It is understandable that the edges between network nodes in the computing power topology sub-network do not have actual physical meaning. They only indicate that two network nodes can be connected through an optical fiber network. How to achieve the connection between two network nodes can be determined by the optical switch routing configuration corresponding to the edge attribute of the connection edge between the two network nodes.

[0048] In this embodiment, the actual connection relationship between computing power centers in the optical fiber network is transformed into an abstract connection relationship in the computing power topology sub-network. With this setting, when computing power scheduling is required, this embodiment only needs to consider whether the starting computing power node and the target computing power node can be connected, and quickly and accurately adjust the switching state of the optical switches in the optical fiber network according to the optical switch routing configuration corresponding to the connected edge, without recalculating the path, which greatly improves the computing power allocation efficiency and topology executability.

[0049] This embodiment also provides a computing power topology application method based on optical fiber links. The application method is based on the computing power topology network generated by the above-described computing power topology generation method based on optical fiber links, as shown in Figure 3. The application method may include the following steps: A1: Obtain computing power allocation demand information, and obtain the initial allocation computing power node and the target allocation computing power node according to the computing power allocation demand information; it can be understood that the computing power allocation demand information is the actual computing power scheduling request, which includes the initial allocation computing power node and the target allocation computing power node, indicating that computing power resources need to be allocated from the initial allocation computing power node to the target allocation computing power node for processing.

[0050] A2: Obtain the corresponding optical switch routing configuration from the computing power topology network based on the initial allocated computing power node and the target allocated computing power node. In this embodiment, since the computing power topology network has pre-built the connectivity between each computing power node, the initial allocated computing power node and the target allocated computing power node can be queried in the computing power topology network to determine whether computing power resources can be transmitted from the initial allocated computing power node to the target allocated computing power node for computation. If transmission is possible, the transmission path can be determined in the computing power topology network, the corresponding edge can be extracted based on the path, and the edge attributes can be obtained from the extracted edge to obtain the optical switch routing configuration.

[0051] A3: Adjust the port connection status of the corresponding optical switch in the optical switch topology network according to the optical switch routing configuration.

[0052] As described above, in this embodiment, since the optical switch routing configuration includes the encoding vector of the switching state of the corresponding optical switch in the optical switch topology network, the switching state of the optical switch can be extracted based on the encoding vector, and then the optical switch can be directly adjusted based on the switching state to make the actual optical path between the initial allocation computing power node and the target allocation computing power node connected, thereby realizing the scheduling of computing power resources.

[0053] As shown in Figure 4, this application embodiment also provides a computing topology generation system based on optical fiber links. Optionally, the generation system may include: an optical switch network construction module 11, used to obtain port connection information of optical switches in the optical fiber network, and construct an optical switch topology network according to the port connection information; in this embodiment, the optical switch network construction module 11 can be used to execute step S1 shown in Figure 1, and a detailed description of the optical switch network construction module 11 can be found in the description of step S1.

[0054] The demand acquisition module 12 is used to acquire computing power connection demand information. In this embodiment, the demand acquisition module 12 can be used to execute step S2 shown in Figure 1. For a detailed description of the demand acquisition module 12, please refer to the description of step S2.

[0055] The path search module 13 is used to search for optical switch link paths in the optical switch topology network based on the computing power connection requirement information and the preset computing power node-optical switch node mapping relationship. In this embodiment, the path search module 13 can be used to execute step S3 shown in Figure 1. For a detailed description of the path search module 13, please refer to the description of step S3.

[0056] The optical switch configuration module 14 is used to obtain the optical switch routing configuration corresponding to the optical switch link path. In this embodiment, the optical switch configuration module 14 can be used to execute step S4 shown in FIG1. ​​For a detailed description of the optical switch configuration module 14, please refer to the description of step S4.

[0057] The computing power topology network construction module 15 is used to construct a computing power topology network according to the computing power connection requirement information and the corresponding optical switch routing configuration. In this embodiment, the computing power topology network construction module 15 can be used to execute step S5 shown in Figure 1. For a detailed description of the computing power topology network construction module 15, please refer to the description of step S5.

[0058] This application provides an electronic device, the structure of which is shown in Figure 5.

[0059] The electronic device includes a memory 21, a processor 22, a communication module 23, and an input / output interface 24, etc. Optionally, the memory 21, the processor 22, the communication module 23, and the input / output interface 24 can be connected and communicate with each other through a bus 25.

[0060] The memory 21 is used to store one or more computer programs and transmit the code of the computer programs to the processor 22; when the one or more computer programs are executed by the processor 22, the computing power topology generation method based on optical fiber links or the computing power topology application method based on optical fiber links in the embodiments of this application are implemented.

[0061] Optionally, the electronic device can be connected to a network via communication module 23 to communicate with other devices, such as terminals or servers, to achieve data interaction. The electronic device can be various forms of digital computers, exemplarily such as desktop computers, servers, workbenches, mainframes, or other types of computers. The electronic device can also be various forms of mobile terminals, exemplarily such as smartphones, tablets, wearable devices (such as helmets, glasses, watches, etc.), and other similar mobile terminals.

[0062] Optionally, the electronic device can connect to required input / output devices, such as a keyboard or display device, via the input / output interface 24. The electronic device itself may have a display device, and other display devices can also be connected externally via the input / output interface 24. Optionally, a storage device, such as a hard disk, can also be connected via the input / output interface 24 to store data from the electronic device, read data from the storage device, or store data from the storage device in the memory 21. It is understood that the input / output interface 24 can be a wired interface or a wireless interface. Depending on the actual application scenario, the device connected to the input / output interface 24 can be a component of the electronic device or an external device connected to the electronic device when needed.

[0063] Optionally, the memory 21 may be a volatile memory and / or a non-volatile memory. The volatile memory may be a random access memory, etc., and the non-volatile memory may be a read-only memory, a programmable read-only memory, an erasable programmable read-only memory, an electrically erasable programmable read-only memory, or a flash memory, etc.

[0064] Optionally, the computer program stored in the processor 22 can be divided into one or more modules, which are stored in the memory 21 and executed by the processor 22 to perform the method provided in this embodiment. The one or more modules can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the electronic device.

[0065] Optionally, the processor 22 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 22 include, but are not limited to, a central processing unit, a graphics processing unit, a digital signal processor, various special-purpose artificial intelligence computing chips, various processors running machine learning model algorithms, and can also be any suitable controller, microcontroller, processor, etc. The processor 22 executes the various methods and processes of this embodiment, exemplarily, such as the fiber optic link-based computing topology generation method or the fiber optic link-based computing topology application method of the embodiments of this application.

[0066] Optionally, the bus 25 may include a path for transmitting information. Depending on its function, the bus 25 may be divided into an address bus, a data bus, a control bus, etc.

[0067] In an optional implementation, this application embodiment also provides a computer storage medium storing a computer program thereon. When executed by a computer, the computer program enables the computer to perform the methods described in the above-described method embodiments. Part or all of the computer program can be loaded and / or installed on the memory 21 of an electronic device. When the computer program is executed by the processor 22, it can perform one or more steps of the fiber optic link-based computing topology generation method or the fiber optic link-based computing topology application method of this application embodiment.

[0068] Optionally, the computer-readable storage medium may be a random access memory, a read-only memory, a programmable read-only memory, an erasable programmable read-only memory, an electrically erasable programmable read-only memory, etc.

[0069] Obviously, the above embodiments of this application are merely examples for clearly illustrating the technical solution of this application, and are not intended to limit the specific implementation of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the claims of this application should be included within the protection scope of the claims of this application.

Claims

1. A method for generating computing topology based on fiber optic links, characterized in that, The generation method includes: obtaining port connection information of optical switches in an optical fiber network, and constructing an optical switch topology network based on the port connection information; obtaining computing power connection requirement information; searching for optical switch link paths in the optical switch topology network based on the computing power connection requirement information and a preset mapping relationship between computing power nodes and optical switch nodes; obtaining the optical switch routing configuration corresponding to the optical switch link path; and constructing a computing power topology network based on the computing power connection requirement information and the corresponding optical switch routing configuration.

2. The computing topology generation method based on fiber optic links according to claim 1, characterized in that, The computing power connection requirement information includes a starting computing power node and a target computing power node; the step of searching for an optical switch link path in the optical switch topology network based on the computing power connection requirement information and a preset computing power node-optical switch node mapping relationship includes: obtaining, based on the computing power node-optical switch node mapping relationship, the starting optical switch node corresponding to the starting computing power node of the computing power connection requirement information in the optical switch topology network, and obtaining the target optical switch node corresponding to the target computing power node of the computing power connection requirement information in the optical switch topology network; in the optical switch topology network, searching for an optical switch link path from the starting optical switch node to the corresponding target optical switch node.

3. The computing topology generation method based on fiber optic links according to claim 2, characterized in that, The step of searching for optical switch link paths in the optical switch topology network that can reach the corresponding target optical switch node includes: using a preset path search algorithm to obtain candidate optical switch link paths that can reach the corresponding target optical switch node from the starting optical switch node; obtaining the path length and average bandwidth of the candidate optical switch link paths; obtaining the average transmission delay of the optical switch link paths based on the path length, average bandwidth, and a preset computing power transmission value; and selecting the optical switch link path from the candidate optical switch link paths based on the average transmission delay.

4. The computing topology generation method based on fiber optic links according to claim 1, characterized in that, The step of obtaining the optical switch routing configuration corresponding to the optical switch link path includes: pre-encoding the optical switches traversed by the optical switch link path and the ports corresponding to the optical switches according to the preset tags of the optical switches; obtaining the optical switches and their corresponding ports included in the optical switch link path; and obtaining the routing encoding vector of the optical switch link path as the optical switch routing configuration based on the encoding of the included optical switches and their corresponding ports.

5. The computing topology generation method based on fiber optic links according to any one of claims 1-4, characterized in that, The computing power connection requirement information is pre-labeled with wavelength tags; the construction of the computing power topology network based on the computing power connection requirement information and the corresponding optical switch routing configuration includes dividing the computing power connection requirement information according to the wavelength tags to obtain several sets of computing power connection requirement information; the wavelength tags of the computing power connection requirement information in the same set of computing power connection requirement information are the same; constructing a computing power topology sub-network of the computing power connection requirement information based on the computing power connection requirement information in the set of computing power connection requirement information and the corresponding optical switch routing configuration; and using the set of computing power topology sub-networks as the computing power topology network.

6. The computing topology generation method based on fiber optic links according to claim 5, characterized in that, The step of constructing a computing power topology sub-network based on the computing power connection demand information in the computing power connection demand information set and the corresponding optical switch routing configuration includes: for the computing power connection demand information, using the corresponding optical switch routing configuration as an edge attribute, constructing a directed edge from the starting computing power node to the target computing power node; using the starting computing power node and the target computing power node in the computing power connection demand information as network nodes, and constructing the computing power topology sub-network based on the network nodes and the corresponding directed edge.

7. A computing topology application method based on fiber optic links, wherein the application method is based on a computing topology network generated by the computing topology generation method based on fiber optic links as described in claim 6, characterized in that, The application method includes: obtaining computing power allocation demand information, and obtaining the initial allocation computing power node and the target allocation computing power node based on the computing power allocation demand information; obtaining the corresponding optical switch routing configuration from the computing power topology network based on the initial allocation computing power node and the target allocation computing power node; and adjusting the port connection status of the corresponding optical switch in the optical switch topology network based on the optical switch routing configuration.

8. A computing topology generation system based on fiber optic links, characterized in that, The generation system includes: an optical switch network construction module, used to acquire port connection information of optical switches in an optical fiber network and construct an optical switch topology network based on the port connection information; a demand acquisition module, used to acquire computing power connection demand information; a path search module, used to search for optical switch link paths in the optical switch topology network based on the computing power connection demand information and a preset computing power node-optical switch node mapping relationship; an optical switch configuration module, used to acquire the optical switch routing configuration corresponding to the optical switch link path; and a computing power topology network construction module, used to construct a computing power topology network based on the computing power connection demand information and the corresponding optical switch routing configuration.

9. An electronic device, characterized in that, include: Memory, used to store one or more computer programs; A processor, when the one or more computer programs are executed by the processor, implements the computing power topology generation method based on fiber optic links as described in any one of claims 1-6 or the computing power topology application method based on fiber optic links as described in claim 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the computing power topology generation method based on fiber optic links as described in any one of claims 1-6 or the computing power topology application method based on fiber optic links as described in claim 7.

Images