All-optical data center network-oriented computing network fusion protection method and storage medium

Abstract

The invention relates to the technical field of resource allocation, and discloses an all-optical data center network-oriented computing network fusion protection method and a storage medium, and the method comprises the steps: carrying out the mapping of an all-optical data center network, obtaining a virtual server, a virtual wave optical switching device and a virtual link, and obtaining a corresponding computing power wave plane list; selecting a target working server and a current working path for the to-be-transmitted service based on the computing power wave plane list; constructing a shared protection server list and a shared protection path list based on a standby working server and a standby path of a successfully deployed service in a time period having time crossing with the actual transmission time period, and selecting a standby working server and a standby path for the to-be-transmitted service; according to the method, a sharing protection strategy is adopted in the path and the computing power resource level, the survivability of the server and the path is considered at the same time, and when the server or the path breaks down, the standby working server and the standby path can rapidly take over tasks, the downtime is shortened, and the service survivability is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of resource allocation, and in particular to a computing-network fusion protection method and storage medium for an all-optical data center network. Background Art

[0002] With the promotion and application of computing-intensive businesses such as the Metaverse, AR / VR, and machine learning, data centers need to ensure that computing resources are readily available for such businesses, which also puts higher demands on the computing resource scheduling capabilities of data centers. In terms of computing resource scheduling, traditional networks use a communication and computing separation framework design, which only serves as a pipeline for information transmission and cannot perceive the needs of upper-level business applications, making it difficult to coordinate and ensure computing resource scheduling needs.

[0003] At present, according to the application technology, data center networks can be divided into electrical switching, optoelectronic hybrid and all-optical data center networks. Comparing the three types of data center networks, all-optical data center networks avoid the electronic bottleneck in electrical switching technology. Not only can electrical switching and optoelectronic hybrid data center networks be upgraded, but they can also be built entirely of optical switches, with the advantages of low energy consumption, high bandwidth and low latency.

[0004] In an all-optical data center network, during service scheduling, once key components such as ports, optical switches, links, and computing servers fail, it will cause serious problems such as service interruption and even service data loss. At present, research on survivability is generally based on a single fault model. For link failures, a strategy of allocating two non-intersecting paths to each optical path is often adopted to ensure 100% survivability. Among these strategies, 1+1 protection and SBPP (Shared Backup Path Protection) protection mechanisms are particularly common, and they mainly focus on the impact of path breakage on service operation. 1+1 protection is a typical protection mechanism. In addition to the working link, it also sets up a backup optical fiber link; when data is transmitted, it is sent to the main link and the backup link synchronously; at the receiving end, the system detects the data of the main link and the backup link, selects the correct data packets and discards the damaged data packets to ensure the continuity of data transmission. And the protection link of each service cannot be shared. Shared backup path protection (SBPP) uses resource sharing by setting up shared backup links between multiple services. When a primary link fails, the backup link will be activated to take over the data transmission of the link. Because the backup link is shared and is only used when a failure occurs, the utilization rate of bandwidth resources is higher and the cost is lower.

[0005] However, in the study of the survivability of high-performance computing services, if a server node fails, such a simple path protection mechanism is inadequate. When a server node fails, the following series of problems may occur: ① Service interruption: The service on the server node will be interrupted immediately, and requests cannot be processed or computing tasks cannot be performed; ② Data loss: If there is unbacked up data on the failed server, data loss may occur; ongoing transactions or data processing may be interrupted, resulting in data inconsistency or loss; ③ Business continuity: affecting business continuity and reliability. Therefore, considering only path protection cannot fully guarantee the survivability of the service. It is necessary to consider the "path + computing node" protection in the data center to more comprehensively guarantee the survivability of AI services. Summary of the invention

[0006] Therefore, the technical problem to be solved by the present invention is to overcome the problem that only path protection is considered in the prior art, which easily leads to service termination, data loss and low business survivability when a server node fails.

[0007] In order to solve the above technical problems, the present invention provides a computing-network fusion protection method for an all-optical data center network, comprising: Build an all-optical data center network consisting of multiple servers and multiple wavelength division optical switching devices; Obtain the source server, candidate source server, computing resources required for the service, time required for service transmission, and service start time of the service to be transmitted; Based on the service start time of the service to be transmitted and the time required for service transmission, an actual transmission period of the service to be transmitted is obtained; Map servers and wavelength division optical switches in the all-optical data center network into virtual servers and virtual wavelength division optical switches respectively; map physical links with available wavelengths in the actual transmission period into virtual links; and construct a computing power wave plane list of the all-optical data center network in the actual transmission period based on virtual servers and their computing resources, virtual wavelength division optical switches and their wavelength connection tables, virtual links and their occupancy configuration information; Based on the computing power wave plane list, multiple candidate servers that meet the computing resources required by the business to be transmitted are obtained; for each candidate server, the shortest route between it and the source server of the business to be transmitted is obtained; based on the computing resources and shortest route of each candidate server, and the number of reconfigurations of the wavelength division switching device on the shortest route, the target working server is selected, and the target wavelength and current working path from the source server to the target working server are obtained; Building a shared protection server list and a shared protection path list based on the backup working servers and backup paths of the services that have been successfully deployed in a time period that overlaps with the actual transmission period; Determine whether there is a protection server in the shared protection server list, and obtain a protection path from the candidate source server of the service to be transmitted to the protection server in the shared protection path list, which does not intersect with the current working path: If it exists, the protection server is obtained as a backup working server for the service to be transmitted; and based on the protection path corresponding to the protection server, a backup path for the service to be transmitted is constructed; If it does not exist, calculate the shortest paths from the alternative source server of the service to be transmitted to each protection server in the shared protection server list respectively, and select the shortest path that does not intersect with the current working path from all the shortest paths as the protection path, and add it to the shared protection path list; construct a backup path for the service to be transmitted based on the protection path, and use the backup path and the protection server corresponding to the backup path as the backup path and backup working server for the service to be transmitted; The service to be transmitted is transmitted based on the target working server, the current working path, the backup working server and the backup path of the service to be transmitted.

[0008] Preferably, constructing a computing power wave plane list of the all-optical data center network within the actual transmission time period includes: Mapping servers in the all-optical data center network to virtual servers, and recording the computing resources of each virtual server in the actual transmission time period in a server resource list; Mapping the wavelength division optical switching devices in the all-optical data center network into virtual wavelength division optical switching devices, and creating a wavelength connection table corresponding to each virtual wavelength division optical switching device based on the wavelength connection information of each port in each virtual wavelength division optical switching device in the actual transmission time period; Mapping physical links with available wavelengths in the all-optical data center network during the actual transmission time period into virtual links, and obtaining occupancy configuration information of the virtual links during the actual transmission time period; the occupancy configuration information includes occupancy status of the receiving end server, occupancy status of the transmitting end server, and port configuration status of the wavelength division optical switching device; The list of virtual servers and server resources, the list of virtual wavelength division switching devices and their corresponding wavelength connection tables, and the virtual links and their occupancy configuration information of the all-optical data center network in the actual transmission time period are used as the computing power wave plane list information in the actual transmission time period to obtain the computing power wave plane list.

[0009] Preferably, after obtaining the current working path, the method further includes: Based on the configuration information and configuration capability of the wavelength division optical switch port in the current working path of the service to be transmitted, it is determined whether the wavelength division optical switch port in the current working path needs to be reconfigured: If reconfiguration is not required, the service start time of the service to be transmitted is used as the transmission start time, and the service to be transmitted is transmitted along the current working path; If reconfiguration is required, the service start time of the service to be transmitted plus the port reconfiguration time of the wavelength division optical switching device is used as the transmission start time, and the service to be transmitted is transmitted along the current working path.

[0010] Preferably, after obtaining the current working path, the method further includes: Get the idle start time of the source server and the idle start time of the target working server on the current working path, and take the later time as the earliest idle time; Compare the earliest idle time with the service start time of the service to be transmitted: If the earliest idle time is earlier than the service start time of the service to be transmitted, the service start time of the service to be transmitted is updated to the earliest idle time.

[0011] Preferably, based on the backup working servers and backup paths of the services that have been successfully deployed in a period that overlaps with the actual transmission period, a shared protection server list and a shared protection path list are constructed, including: Acquire all the backup working servers of the services that have been successfully deployed in a time period that overlaps with the actual transmission period of the services to be transmitted, and construct a shared protection server list; Obtain the backup paths corresponding to all backup working servers in the shared protection server list, obtain the optical channel construction order between all wavelength division optical switching devices on the backup path as the protection path, and construct a shared protection path list; Deleting the standby working servers whose computing resources do not meet the computing resources required by the services to be transmitted from the shared protection server list, and updating the shared protection server list; The protection path corresponding to the deleted standby working server is deleted from the shared protection path list, and the shared protection path list is updated.

[0012] Preferably, determining whether there is a protection server in the shared protection server list and obtaining a protection path from the candidate source server of the service to be transmitted to the protection server in the shared protection path list includes: The wavelength division optical switching device directly connected to the source candidate server is used as the source wavelength division optical switching device; The wavelength division optical switching device directly connected to the protection server in the shared protection server list is used as the destination wavelength division optical switching device; Determine whether there is a path from the source WDM optical switch to the destination WDM optical switch in the shared protection path list: If it exists, it means that there is a protection server in the shared protection server list, and a protection path from the candidate source server of the service to be transmitted to the protection server can be obtained in the shared protection path list, so as to build a backup path for the service to be transmitted; If it does not exist, it means that the protection servers in the shared protection server list cannot obtain a protection path from the candidate source server of the service to be transmitted to the protection server in the shared protection path list.

[0013] Preferably, based on the computing power wave plane list, multiple servers to be selected that meet the computing resources required for the business to be transmitted are obtained, including: obtaining the computing resources of the virtual server in each computing power wave plane, and obtaining multiple servers whose computing resources meet the computing resources required for the business to be transmitted as servers to be selected.

[0014] Preferably, when there are multiple services to be transmitted, and the current protection paths of the multiple services to be transmitted do not intersect with each other, the multiple services to be transmitted are allowed to share the same standby working server and the same standby path.

[0015] Preferably, the two paths do not intersect, including that any link and any wavelength division optical switching device in the two paths do not overlap.

[0016] This embodiment also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the computing-network fusion protection method for an all-optical data center network as described above are implemented.

[0017] The above technical solution of the present invention has the following beneficial effects compared with the prior art: The computing network fusion protection method for all-optical data center network described in the present invention maps the servers, wavelength division optical switching devices and physical links in the whole network data center network to obtain the corresponding computing power wave plane; based on the computing resources, wavelength connection table and occupancy configuration information recorded in the computing power wave plane, the target working server and the current working path are selected. In order to further ensure the survivability of the service, the present invention adopts a shared protection strategy at both the path level and the computing power resource level, while considering the survivability of the server and the path, and selects a backup working server and a backup path for the service to be transmitted through the constructed shared protection server list and shared protection path list; path protection ensures the stability and continuity of data transmission and improves the security of data transmission; when a server or path fails, the backup working server and the backup path can quickly take over the task, reduce downtime, and ensure business survivability; the present invention achieves a balance between cost-effectiveness and protection capabilities through efficient resource management and scheduling, optimizes resource utilization, and improves network performance.

[0018] Furthermore, the present invention provides a backup source server for the business to be transmitted, ensuring that data can be protected during storage and processing to avoid data loss and damage; The present invention also allows more than two services with non-intersecting links to share computing power protection resources on the same server, as well as share the same backup path, thereby optimizing resource utilization, improving overall system performance, and providing a flexible, efficient, and comprehensive survivability solution for services to be transmitted in all-optical data center networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein: Figure 1 It is a flowchart of the steps of the computing-network fusion protection method for an all-optical data center network provided by the present invention; Figure 2 It is a schematic diagram of the physical structure of the all-optical data center network provided by the present invention; Figure 3 It is a list of computing power wave planes of the all-optical data center network provided by the present invention; Figure 4 is a resource deployment flow chart provided by the present invention; Figure 5 It is a flow chart of the computing power wavelength plane algorithm provided by the present invention. DETAILED DESCRIPTION

[0020] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0021] Reference Figure 1 As shown, the flowchart of the computing-network fusion protection method for an all-optical data center network of the present invention includes the following specific steps: S101: Build an all-optical data center network consisting of multiple servers and multiple wavelength division optical switching devices; S102: Obtaining a source server, an alternative source server, computing resources required for the service to be transmitted, time required for service transmission, and service start time; S103: based on the service start time of the service to be transmitted and the time required for service transmission, obtaining an actual transmission period of the service to be transmitted; S104: Map servers and wavelength division optical switches in the all-optical data center network into virtual servers and virtual wavelength division optical switches respectively; map physical links with available wavelengths in the actual transmission period into virtual links; and construct a computing power wave plane list of the all-optical data center network in the actual transmission period based on the virtual servers and their computing resources, the virtual wavelength division optical switches and their wavelength connection tables, the virtual links and their occupation configuration information; S105: Based on the computing power wave plane list, multiple candidate servers that meet the computing resources required by the service to be transmitted are obtained; for each candidate server, the shortest route between it and the source server of the service to be transmitted is obtained; based on the computing resources and the shortest route of each candidate server, and the number of reconfigurations of the wavelength division optical switching device on the shortest route, a target working server is selected, and a target wavelength and a current working path from the source server to the target working server are obtained; The selection of the server to be selected includes: obtaining computing resources of the virtual server in each computing power wave plane, obtaining multiple servers whose computing resources meet the computing resources required by the business to be transmitted as the servers to be selected; S106: Building a shared protection server list and a shared protection path list based on the backup working servers and backup paths of the services that have been successfully deployed in a period that overlaps with the actual transmission period; S107: Determine whether there is a protection server in the shared protection server list, and obtain a protection path from the candidate source server of the service to be transmitted to the protection server in the shared protection path list, which does not intersect with the current working path: If it exists, the protection server is obtained as a backup working server for the service to be transmitted; and based on the protection path corresponding to the protection server, a backup path for the service to be transmitted is constructed; If it does not exist, calculate the shortest paths from the alternative source server of the service to be transmitted to each protection server in the shared protection server list respectively, and select the shortest path that does not intersect with the current working path from all the shortest paths as the protection path, and add it to the shared protection path list; construct a backup path for the service to be transmitted based on the protection path, and use the backup path and the protection server corresponding to the backup path as the backup path and backup working server for the service to be transmitted; S108: Transmit the service to be transmitted based on the target working server of the service to be transmitted, the current working path, the backup working server and the backup path.

[0022] Specifically, refer to Figure 2 The figure shows the physical structure of the all-optical data center network. Figure 3As shown, it is a list of computing power wave planes of the all-optical data center network; the list of computing power wave planes of the all-optical data center network in the actual transmission time period is constructed, including: S104-1: Map servers in the all-optical data center network to virtual servers, and record computing resources of each virtual server in an actual transmission time period in a server resource list; S104-2: Map the wavelength division optical switching devices in the all-optical data center network to virtual wavelength division optical switching devices, and create a wavelength connection table corresponding to each virtual wavelength division optical switching device based on the wavelength connection information of each port in each virtual wavelength division optical switching device in the actual transmission time period; S104-3: Mapping physical links with available wavelengths in the all-optical data center network during the actual transmission time period to virtual links, and obtaining occupancy configuration information of the virtual links during the actual transmission time period; the occupancy configuration information includes occupancy status of the receiving end server, occupancy status of the transmitting end server, and port configuration status of the wavelength division optical switching device; The transmitting end server refers to the source server and the alternative source server, and the receiving end server refers to the target working server and the alternative working server.

[0023] S104-4: The list of virtual servers and server resources, the list of virtual wavelength division switching devices and their corresponding wavelength connection tables, and the virtual links and their occupancy configuration information of the all-optical data center network in the actual transmission time period are used as the computing power wave plane list information in the actual transmission time period to obtain the computing power wave plane list.

[0024] This embodiment adds a mapping of the available status of the server and computing resources on the traditional wavelength plane, so as to achieve unified management and control of network resources and computing resources based on the computing power wave plane, improve the efficiency of path allocation, and ensure business survivability.

[0025] Specifically, in an embodiment of the present invention, after obtaining the current working path, it also includes: based on the configuration information and configuration capability of the port of the wavelength division optical switching device in the current working path of the service to be transmitted, judging whether the port of the wavelength division optical switching device in the current working path needs to be reconfigured: if reconfiguration is not required, the service start time of the service to be transmitted is used as the transmission start time, and the service to be transmitted is transmitted along the current working path; if reconfiguration is required, the service start time of the service to be transmitted plus the port reconfiguration time of the wavelength division optical switching device is used as the transmission start time, and the service to be transmitted is transmitted along the current working path.

[0026] Specifically, in an embodiment of the present invention, after obtaining the current working path, it also includes: obtaining the idle start time of the source server and the idle start time of the target working server on the current working path, and taking the later time as the earliest idle time; comparing the earliest idle time with the business start time of the service to be transmitted: if the earliest idle time is earlier than the business start time of the service to be transmitted, then updating the business start time of the service to be transmitted to the earliest idle time.

[0027] This embodiment adjusts the service start time of the service to be transmitted in combination with the actual network conditions of the all-optical data center network to achieve more efficient resource management and scheduling.

[0028] Specifically, in step S106, a shared protection server list and a shared protection path list are established, including: S106-1: Acquire all successfully deployed backup working servers of services in a time period that overlaps with the actual transmission period of the service to be transmitted, and construct a shared protection server list; S106-2: Obtain the backup paths corresponding to all backup working servers in the shared protection server list, obtain the optical channel construction sequence between all wavelength division optical switching devices on the backup path as the protection path, and construct a shared protection path list; S106-3: Deleting the backup working servers whose computing resources do not meet the computing resources required by the services to be transmitted from the shared protection server list, and updating the shared protection server list; S106-4: Delete the protection path corresponding to the deleted standby working server from the shared protection path list, and update the shared protection path list.

[0029] Specifically, in step S107, it is determined whether there is a protection server in the shared protection server list, and a protection path from the candidate source server of the service to be transmitted to the protection server can be obtained in the shared protection path list, including: The wavelength division optical switching device directly connected to the source candidate server is used as the source wavelength division optical switching device; The wavelength division optical switching device directly connected to the protection server in the shared protection server list is used as the destination wavelength division optical switching device; Determine whether there is a path from the source WDM optical switch to the destination WDM optical switch in the shared protection path list: If it exists, it means that there is a protection server in the shared protection server list, and a protection path from the candidate source server of the service to be transmitted to the protection server can be obtained in the shared protection path list, so as to build a backup path for the service to be transmitted; If it does not exist, it means that the protection servers in the shared protection server list cannot obtain a protection path from the candidate source server of the service to be transmitted to the protection server in the shared protection path list.

[0030] If there is a path from the source wavelength division optical switch to the destination wavelength division optical switch in the shared protection path list, determine whether the source wavelength division optical switch and the destination wavelength division optical switch need to perform port reconfiguration. If necessary, obtain the corresponding protection path after completing the port reconfiguration; if not, directly obtain the path at the current moment as the protection path.

[0031] When selecting a backup path, this embodiment selects a protection path from one wavelength division optical switching device to another wavelength division optical switching device, and directly connects the two wavelength division optical switching devices at both ends of the protection path to the alternative source server and the protection server respectively to obtain the corresponding backup path.

[0032] In this embodiment, when there are multiple services to be transmitted, and the current protection paths of the multiple services to be transmitted do not intersect with each other, multiple services to be transmitted are allowed to share the same backup working server and the same backup path. The two paths do not intersect, including that any link and any wavelength division optical switching device in the two paths do not overlap. The present invention allows more than two services with non-intersecting links to share computing power protection resources on the same server, as well as share the same backup path, thereby optimizing resource utilization, improving overall system performance, and providing a flexible, efficient, and comprehensive survivability solution for services to be transmitted in all-optical data center networks. While inheriting the traditional survivability algorithm, this embodiment additionally introduces considerations for the survivability of computing power resources. Drawing on the 1+1 protection and shared protection concepts in path protection, the embodiment of the present invention designs a corresponding shared backup computing power protection strategy (Shared Backup Computing Power Protection, SBCPP) for computing power resources. It is a shared protection strategy for computing power resources, similar to the SBPP (Shared Backup Path Protection) strategy for path protection, but the SBCPP strategy focuses on computing power, allowing more than two services to share computing power protection resources on the same server. During the service deployment process, this application comprehensively considers the dual guarantees of path survivability and computing power survivability. For path protection, this application continues to use the traditional SBPP strategy and combines it with the SBCPP strategy for computing power protection, that is, it combines SBPP path protection and SBCPP computing power protection to efficiently complete the allocation of working paths and protection paths for services.

[0033] Based on the above embodiments, in the embodiments of the present invention, the service sequence is transmitted by using the computing network fusion protection method for the all-optical data center network provided by the present invention. Figure 4 As shown in FIG. 1 , it is a resource deployment flowchart, and the specific steps include: S201: Generate a business sequence of intensive computing; Specifically, given a set of intensive computing requirements, each business requirement is represented as (s, d, s p ,d p ,T,r,r p ,t). Given that s, d, s p ,d p , T, t represent the source server, working server, backup source server, protection server, the time required for business transmission, and the time when the business starts scheduling and deployment, respectively. All data are randomly generated. r represents the working route of the business, r p Indicates the protection route of the service, which is usually empty. d indicates the working server of the service. p Indicates the protection server of the business. The default value is empty.

[0034] For each service demand, the destination server d needs to be obtained according to the algorithm, and then the path r between servers s and d and the wavelengths available in T consecutive time slots along the path are found.

[0035] And allocate protection servers for business p , then find the server s p and d p The path between p , the services are deployed in the order in which they are generated.

[0036] S202: Constructing a computing power wave plane, a shared protection server list, and a shared protection path list; According to the number of available wavelengths N, the computing power wave plane ω of the all-optical data center network in [t, t+T-1] is generated 1 ,ω 2 ,…,ω N According to the successfully deployed services with overlapping time, the link occupancy, receiver transmitter occupancy, and switch port configuration are updated on the computing power wave plane corresponding to the services. And the server computing power resource list is updated.

[0037] According to the successfully deployed services that intersect with the services to be transmitted within the time [t, t+T-1], check whether the protection server corresponding to the service already exists in the shared protection server list. If the protection server exists, check whether the shared protection computing power resources of the corresponding server in the shared server list are greater than the protection computing power resources occupied by the current service. If so, update the number of protection computing power resources, otherwise do not process. If the protection server does not exist, add the protection server corresponding to the service and the protection computing power resources occupied by the server to the shared protection server list. At the same time, check whether the protection optical channel corresponding to the service already exists in the shared protection path list. If not, record the protection optical channel corresponding to the service in the shared protection path list.

[0038] S203: Allocate a working server, a working path, a protection server and a protection path for the service; Use the computing power wave plane algorithm to complete the deployment of working servers and working paths: For a high-performance computing business, the system will default the service arrival time to the time when the business starts scheduling and deployment. First, the working server for business deployment is obtained according to the computing power wave plane algorithm. Then, the all-optical data center network in the next T consecutive time slots is established to use Dijkstra to find the shortest path and determine whether there are corresponding transceivers, wavelengths, and switch port configurations in the T consecutive time slots to deploy the business. If so, an optical channel is established for the business to transmit data, and the current business working path is successfully deployed; and the link occupancy, receiver transmitter occupancy, switch port configuration, and computing power resource usage of the working server are updated on the computing power wave plane corresponding to the working path.

[0039] This embodiment uses the SBPP path protection + SBCPP computing power protection strategy to complete the deployment of protection servers and protection paths: This strategy adopts a shared protection strategy at both the link and computing power resource levels, aiming to achieve a balance between cost-effectiveness and protection capabilities through efficient resource management and scheduling. In this strategy, when there are no intersecting links between the main paths of multiple services, multiple services are allowed to share the same shared protection path and computing resources on the same server.

[0040] First, check whether the current shared protection server list is empty. If it is empty, use the power wave plane algorithm to complete the selection of protection servers and protection paths, and update the protection servers and protection paths to the shared protection server list and shared protection path list. If it is not empty, check whether the shared protection path list has a connection. p and protect server d p If there is an optical channel, the protection path and protection server deployment are complete. pThere are no available protection optical channels. Dijkstra is used to find the shortest path and determine whether there are corresponding transceivers, wavelengths, and switch port configurations for deployment in T consecutive time slots. Finally, the found optical channels are recorded in the shared protection path list.

[0041] When the working path, working server, standby path and protection server are all deployed, the current service deployment ends. Otherwise, it will wait and repeat the above process until the service deployment is successful.

[0042] Specifically, the definition of the computing power wave plane is: The computing power wave plane can be understood as a new definition expanded from the concept of the traditional wave plane; the traditional wave plane only considers the mapping of network devices (switches) and links (optical fiber links), but does not perceive the computing power. The computing power wave plane adds the mapping of the server's available status and the computing resources on the server to the concept of the traditional wave plane, realizing the unified management and control of network resources and computing resources.

[0043] Reference Figure 5 As shown in the figure, it is a flowchart based on the computing power wavelength plane algorithm. The main steps are as follows: S301: Construct a computing power wave plane within a corresponding time period; For each wavelength, a computing power wave plane copied from the physical network topology is generated, in which each virtual server (FPGA) and virtual wavelength division optical switch device are mapped from the physical server node and the physical wavelength division optical switch device, and each virtual link corresponds to whether the corresponding wavelength in the physical link is available.

[0044] When the corresponding wavelength is available on the physical link within the service duration, the link is added to the corresponding computing power wave plane.

[0045] When its corresponding computing power is available to provide services on the physical server node within a given time, a virtual server is added and the computing resources of the server are recorded in the server resource list on the computing power wave plane.

[0046] In addition, a wavelength connection table is created for each wavelength division optical switch to store the wavelength connection information of each port therein; using the availability of each wavelength and the computing power on each node, the algorithm can generate a computing wave plane list ω of the all-optical DCN in the corresponding time period 1 ,ω 2 ,…,ω N ; And after each business is completed, update the network and server resource mapping list on the computing power wave plane; S302: Selecting a target server node to be selected for the server, and obtaining a server to be selected; Find the computing power resource status of the servers on the constructed computing power wave plane, and select k servers with sufficient computing power resources for the service, which are the candidate servers d 1 , d 2 , …, d k , where 0 < k < N, and N represents the total number of servers; S303: Search the computing power wave plane list and select a route; For each candidate server d 1 , d 2 , …, d k , for each computing power wave plane ω corresponding to each wavelength of the all-optical data center network within [t, t + T - 1] 1 , ω 2 , …, ω N Query one by one, and try to find the shortest route between the source server and the candidate server through Dijkstra's algorithm on each computing power wavelength plane, and the wavelength corresponding to the current computing power wave plane is the wavelength used by the route r; Scan to obtain the earliest idle times t x 、t r of the transmitter and receiver at both ends of the route, and update the service time; S304: Jointly optimize the computing power resources of the servers and the route selection through an algorithm, obtain the optimal server as the destination server D, and obtain the corresponding shortest route information R and wavelength ω*; The joint optimization includes: among all reachable alternative servers for which corresponding routes can be found, search in descending order according to the computing power resources from large to small, and select the corresponding route R and server with the fewest reconfiguration times of the wavelength division optical switch device switch as the optimal server D.

[0047] S305: Check the switch ports; Check the configuration information of the switch ports passed by the route R. According to the method of whether wavelengths are configured between ports and whether the switch has the reconfiguration ability, determine whether reconfiguration is required; If reconfiguration is required, extend the transmission time by ts, where ts is the switch port reconfiguration time, that is, let T = T + ts, and deploy the service; if no reconfiguration is required, directly deploy the service; S306: The transmission is completed, and the service is processed on the server side; During the service calculation time, the corresponding computing resources of the server are occupied, and the computing power resource list is updated.

[0048] Based on the above embodiments, in this embodiment, the all-optical data center network is composed of servers and switches. A total of 12 servers are set in the network, and the computing power of the servers is randomly generated within the range of [50TFLOPS / s, 500TFLOPS / s], where FLOPS / s represents the number of floating-point operations per second. The total number of service requests in the network is fixed, and the time required for service transmission is randomly generated within the range of [20, 50] time slots TS (TimeSlot), and the computing requirements of the service are randomly generated in the computing power requirement interval [2TGFLOPS, 20TFLOPS], where the unit FLOPS is a floating-point operation. The service arrival time is initialized to 0TS. Wavelength conversion is not allowed during service configuration, so each optical path is subject to the constraints of wavelength continuity. In addition, in order to prevent the risk of simultaneous failure of the working path and the protection path due to access switch failure, this embodiment takes a preventive measure: pre-backing up the service data to servers connected to two different input ports to ensure data accessibility and service continuity. The protection algorithm is used to allocate working servers and working paths, protection servers and protection paths to the services respectively, and the maximum number of wavelengths used to complete the task and the redundancy of spare resources are obtained.

[0049] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the computing-network fusion protection method for an all-optical data center network as described above are implemented.

[0050] The computing network fusion protection method for the all-optical data center network described in the present invention maps the servers, wavelength division optical switching devices and physical links in the data center network of the entire network to obtain the corresponding computing power wave plane; based on the computing resources, wavelength connection table and occupancy configuration information recorded in the computing power wave plane, the target working server and the current working path are selected. In order to further ensure the survivability of the service, the present invention adopts a shared protection strategy at both the path level and the computing power resource level, while considering the survivability of the server and the path, and selects a backup working server and a backup path for the service to be transmitted through the constructed shared protection server list and shared protection path list; path protection ensures the stability and continuity of data transmission and improves the security of data transmission; when a server or path fails, the backup working server and the backup path can quickly take over the task, reduce downtime, and ensure business survivability; the present invention achieves a balance between cost-effectiveness and protection capabilities through efficient resource management and scheduling. The present invention also provides a backup source server for the services to be transmitted, ensuring that data can be protected during storage and processing to avoid data loss and damage; the present invention allows more than two services with non-intersecting links to share computing power protection resources on the same server, as well as share the same backup path, thereby optimizing resource utilization, improving overall system performance, and providing a flexible, efficient and comprehensive survivability solution for the services to be transmitted in the all-optical data center network.

[0051] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0052] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0053] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0054] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0055] Obviously, the above embodiments are merely examples for the purpose of clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the present invention.

Claims

1. A computing-network fusion protection method for an all-optical data center network, characterized in that: include: Build an all-optical data center network consisting of multiple servers and multiple wavelength division optical switching devices; Obtain the source server, candidate source server, computing resources required for the service, time required for service transmission, and service start time of the service to be transmitted; Based on the service start time of the service to be transmitted and the time required for service transmission, an actual transmission period of the service to be transmitted is obtained; Map servers and wavelength division optical switches in the all-optical data center network into virtual servers and virtual wavelength division optical switches respectively; map physical links with available wavelengths in the actual transmission period into virtual links; and construct a computing power wave plane list of the all-optical data center network in the actual transmission period based on virtual servers and their computing resources, virtual wavelength division optical switches and their wavelength connection tables, virtual links and their occupancy configuration information; Based on the computing power wave plane list, multiple candidate servers that meet the computing resources required by the business to be transmitted are obtained; For each server to be selected, obtain the shortest route between it and the source server of the service to be transmitted; Based on the computing resources and shortest route of each candidate server and the number of reconfiguration of the wavelength division optical switching device on the shortest route, the target working server is selected, and the target wavelength and current working path from the source server to the target working server are obtained; Building a shared protection server list and a shared protection path list based on the backup working servers and backup paths of the services that have been successfully deployed in a time period that overlaps with the actual transmission period; Determine whether there is a protection server in the shared protection server list, and obtain a protection path from the candidate source server of the service to be transmitted to the protection server in the shared protection path list, which does not intersect with the current working path: If it exists, the protection server is obtained as a backup working server for the service to be transmitted; And based on the protection path corresponding to the protection server, a backup path for the service to be transmitted is constructed; If it does not exist, the shortest paths from the candidate source server of the service to be transmitted to each protection server in the shared protection server list are calculated respectively, and the shortest path that does not intersect with the current working path is selected from all the shortest paths as the protection path and added to the shared protection path list; Building a backup path for the service to be transmitted based on the protection path, and using the backup path and the protection server corresponding to the backup path as the backup path and the backup working server for the service to be transmitted; The service to be transmitted is transmitted based on the target working server, the current working path, the backup working server and the backup path of the service to be transmitted.

2. The computing-network fusion protection method for all-optical data center network according to claim 1 is characterized in that: Construct a list of computing power wave planes for the all-optical data center network during the actual transmission period, including: Mapping servers in the all-optical data center network to virtual servers, and recording the computing resources of each virtual server in the actual transmission time period in a server resource list; Mapping the wavelength division optical switching devices in the all-optical data center network into virtual wavelength division optical switching devices, and creating a wavelength connection table corresponding to each virtual wavelength division optical switching device based on the wavelength connection information of each port in each virtual wavelength division optical switching device in the actual transmission time period; Mapping physical links with available wavelengths in the all-optical data center network during the actual transmission time period into virtual links, and obtaining occupancy configuration information of the virtual links during the actual transmission time period; the occupancy configuration information includes occupancy status of the receiving end server, occupancy status of the transmitting end server, and port configuration status of the wavelength division optical switching device; The list of virtual servers and server resources, the list of virtual wavelength division switching devices and their corresponding wavelength connection tables, and the virtual links and their occupancy configuration information of the all-optical data center network in the actual transmission time period are used as the computing power wave plane list information in the actual transmission time period to obtain the computing power wave plane list.

3. The computing-network fusion protection method for all-optical data center network according to claim 1 is characterized in that: After getting the current working path, it also includes: Based on the configuration information and configuration capability of the wavelength division optical switch port in the current working path of the service to be transmitted, it is determined whether the wavelength division optical switch port in the current working path needs to be reconfigured: If reconfiguration is not required, the service start time of the service to be transmitted is used as the transmission start time, and the service to be transmitted is transmitted along the current working path; If reconfiguration is required, the service start time of the service to be transmitted plus the port reconfiguration time of the wavelength division optical switching device is used as the transmission start time, and the service to be transmitted is transmitted along the current working path.

4. The computing-network fusion protection method for all-optical data center network according to claim 1 is characterized in that: After getting the current working path, it also includes: Get the idle start time of the source server and the idle start time of the target working server on the current working path, and take the later time as the earliest idle time; Compare the earliest idle time with the service start time of the service to be transmitted: If the earliest idle time is earlier than the service start time of the service to be transmitted, the service start time of the service to be transmitted is updated to the earliest idle time.

5. The computing-network fusion protection method for all-optical data center network according to claim 1 is characterized in that: Based on the backup working servers and backup paths of the services that have been successfully deployed in the period that overlaps with the actual transmission period, a shared protection server list and a shared protection path list are constructed, including: Acquire all the backup working servers of the services that have been successfully deployed in a time period that overlaps with the actual transmission period of the services to be transmitted, and construct a shared protection server list; Obtain the backup paths corresponding to all backup working servers in the shared protection server list, obtain the optical channel construction order between all wavelength division optical switching devices on the backup path as the protection path, and construct a shared protection path list; Deleting the standby working servers whose computing resources do not meet the computing resources required by the services to be transmitted from the shared protection server list, and updating the shared protection server list; The protection path corresponding to the deleted standby working server is deleted from the shared protection path list, and the shared protection path list is updated.

6. The computing-network fusion protection method for all-optical data center network according to claim 1 is characterized in that: Determining whether there is a protection server in the shared protection server list, and obtaining a protection path from the candidate source server of the service to be transmitted to the protection server in the shared protection path list, including: The wavelength division optical switching device directly connected to the source candidate server is used as the source wavelength division optical switching device; The wavelength division optical switching device directly connected to the protection server in the shared protection server list is used as the destination wavelength division optical switching device; Determine whether there is a path from the source WDM optical switch to the destination WDM optical switch in the shared protection path list: If it exists, it means that there is a protection server in the shared protection server list, and a protection path from the candidate source server of the service to be transmitted to the protection server can be obtained in the shared protection path list, so as to build a backup path for the service to be transmitted; If it does not exist, it means that the protection servers in the shared protection server list cannot obtain a protection path from the candidate source server of the service to be transmitted to the protection server in the shared protection path list.

7. The computing-network fusion protection method for all-optical data center network according to claim 1 is characterized in that: Based on the computing power wave plane list, multiple servers to be selected that have computing resources required for the business to be transmitted are obtained, including: obtaining computing resources of virtual servers in each computing power wave plane, and obtaining multiple servers whose computing resources meet the computing resources required for the business to be transmitted as servers to be selected.

8. The computing-network fusion protection method for all-optical data center network according to claim 1 is characterized in that: When there are multiple services to be transmitted, and the current protection paths of the multiple services to be transmitted do not intersect with each other, the multiple services to be transmitted are allowed to share the same standby working server and the same standby path.

9. The computing-network fusion protection method for all-optical data center network according to claim 1, characterized in that: The two paths do not intersect, including that any link and any wavelength division optical switching device in the two paths do not overlap.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the computing-network fusion protection method for an all-optical data center network are implemented.

Images