Message transmission method and system, network equipment and computing equipment cluster

Abstract

The embodiment of the invention provides a message transmission method and system, network equipment and a computing equipment cluster, and relates to the technical field of cloud computing, and the message transmission method is applied to a revolution network equipment group in networking. The method comprises the following steps: a first upper network device receives a detour message which is from a source network device group and is forwarded to a target network device by a to-be-wound network device group, and forwards the detour message to a first lower network device according to a public equivalent path group; the first lower network device forwards the received bypassing message to a second upper network device according to an overall equivalence path group used for indicating an equivalence path formed by the interconnection ports in a second group, wherein the equivalence path between the second upper network device and the target network device group meets an inter-group revolution rule; and the second upper network equipment forwards the received bypassing message to the target network equipment group according to the direct connection equivalent path group. According to the embodiment of the invention, the load sharing demand of message transmission in networking can be more fully met.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of cloud computing technology, and in particular, to a method and system for message transmission, network devices, and a cluster of computing devices. Background Art

[0002] A network (such as a data center, Ethernet, etc.) can form a topology through multiple network device groups, and message transmission can be performed between host clusters in the network and / or between a host cluster in the network and devices outside the network. Among them, a network device group refers to a group of multiple network devices (such as switches or routers, etc.). Multiple network device groups can change the optical path through an optical switching device (such as an optical cross-connect or an optical circuit switch, etc.), so as to flexibly adjust the topological connection relationship between groups, taking into account lower transmission costs and higher transmission performance.

[0003] Taking the transmission of messages between different switch groups as an example. In the scenario where a source switch group sends a message to a destination switch group, the source switch group can forward the message to the destination switch group through a bypass switch group between it and the destination switch group to better handle emergencies. At this time, the out-ports connected to the destination switch group in the bypass switch group are likely to be distributed on different upper aggregation switches in the bypass switch group, and load balancing needs to be performed on these out-ports. For example, each upper aggregation switch in the bypass switch group determines an equivalent path table according to the routing protocol, and then forwards messages with a load balancing effect based on the equivalent path table.

[0004] However, in specific applications, the scale of the network is often huge. Correspondingly, the number of ports involved in the equivalent path table also increases. For example, the number of internal interconnection ports of a switch is generally at least in the dozens. The next hops of the out-ports corresponding to these internal interconnection ports in the equivalent path table are usually at least in the dozens. In this way, for the network, a large number of entries in the equivalent path table need to be consumed to achieve load balancing of the out-ports on different switches in the switch group. However, the scale of the entries in the equivalent path table is limited, which easily leads to limited number of entries allocated to different destination switch groups, causing the problem of uneven load sharing. Summary of the Invention

[0005] To solve the above technical problems, the present application provides a method and system for message transmission, network devices, and a cluster of computing devices. The message transmission method ensures that the types of each interconnection port in the equivalent path table, that is, the equivalent path group, for transmitting the bypass message are the same through the overall scheduling of the lower network device for the bypass message received by the upper network device, and realizes load sharing of the transmitted message at the cost of fewer entries in the equivalent path group, so as to more fully meet the load sharing requirements of message transmission in the network under the limited entry scale of the equivalent path table.

[0006] In a first aspect, an embodiment of the present application provides a packet transmission method, which is applied to a rotating network device group in a network. The rotating network device group includes a plurality of upper network devices that are docked with devices outside the rotating network device group, and a plurality of lower network devices that are docked with computing devices in the network. Each upper network device and each lower network device are communicatively connected through an intra-group interconnection port. The network further includes a source network device group and a destination network device group, and the destination network device group is communicatively connected to each upper network device through an inter-group interconnection port. The method includes: a first upper network device receives a detouring packet, where the first upper network device includes any one of the plurality of upper network devices, and the detouring packet includes a packet from the source network device group and to be forwarded by the rotating network device group to the destination network device; the first upper network device forwards the detouring packet to a first lower network device among the plurality of lower network devices according to a common equivalent path group, where the common equivalent path group is used to indicate an equivalent path formed by a first group of intra-group interconnection ports that meet the load balancing rule among the intra-group interconnection ports, and the first lower network device corresponds to the first group of intra-group interconnection ports; the first lower network device forwards the received detouring packet to a second upper network device among the plurality of upper network devices according to a coordinated equivalent path group, where the coordinated equivalent path group is used to indicate an equivalent path formed by a second group of intra-group interconnection ports among the intra-group interconnection ports, the second group of intra-group interconnection ports meet the load balancing rule and correspond to the second upper network device, and the equivalent path between the second upper network device and the destination network device group meets the inter-group detouring rule; the second upper network device forwards the received detouring packet to the destination network device group according to a direct connection equivalent path group, where the direct connection equivalent path group is used to indicate an equivalent path formed by a first group of inter-group interconnection ports that meet the inter-group detouring rule among the inter-group interconnection ports.

[0007] In the embodiment of the present application, the first upper network device in the rotation network device group that receives the detour message sent by the source network device group forwards the detour message to the first lower network device, so as to ensure that the types of the interconnection ports in the common equivalent path group indicating the transmission path of the detour message are all intra-group interconnection ports during this process. On this basis, the first upper network device forwards the detour message to the second upper network device in the rotation network device group that docks with the destination network device group, so as to ensure that the types of the interconnection ports in the overall equivalent path group indicating the transmission path of the detour message are all intra-group interconnection ports during this process. In this way, when the second upper network device forwards the detour message to the destination network device group, it can ensure that the types of the interconnection ports in the direct connection equivalent path group indicating the transmission path of the detour message are all extra-group interconnection ports. That is to say, through the forwarding or overall scheduling of the lower network device, the various equivalent path tables corresponding to the transmission path of the detour message can be adjusted to have the same type of interconnection ports involved, so as to avoid different types of interconnection ports aggregating in an equivalent path group, that is, an equivalent path table, and greatly reduce the consumption of table entries in an equivalent path group. Therefore, the solution of the present application can achieve load sharing of the transmitted messages at the cost of fewer table entries in the equivalent path group, so as to more fully meet the load sharing requirements of message transmission in the network under the limited table entry scale of the equivalent path table.

[0008] According to the first aspect, when the rotation network device group is divided into a federated routing autonomous system, each upper network device and each lower network device respectively correspond to a single routing autonomous system identifier under the federated routing autonomous system. Each of the equivalent paths indicated by the direct connection equivalent path group satisfies one or more of the following conditions: any one of the equivalent paths indicated by the direct connection equivalent path group corresponds to at least one routing autonomous system identifier, and any one of the at least one is different from the single routing autonomous system identifier under the federated routing autonomous system to which the local device belongs; the total number of routing autonomous system identifiers corresponding to the equivalent paths indicated by the direct connection equivalent path group is less than the quantity threshold.

[0009] In the embodiment of the present application, when the rotation network device group is divided into a federated routing autonomous system, each upper network device and each lower network device respectively correspond to a single routing autonomous system identifier under the federated routing autonomous system: The routing autonomous system identifiers respectively corresponding to the equivalent paths indicated by the direct connection equivalent path group include at least one, and any one of the at least one is different from the single routing autonomous system identifier under the federated routing autonomous system to which the local device belongs, which can ensure that in the direct connection equivalent path group, if the same detour network device group is regarded as a routing autonomous system, when the detour packets reach the second upstream network device in each path indicated by the direct connection equivalent path group, they will not detour within the detour network device group, but leave the detour network device group from the second upstream network device, avoiding the transmission path of the detour packets from forming a loop within the group, and further improving the packet transmission efficiency; and / or, The total number of the routing autonomous system identifiers corresponding to the equivalent paths indicated by the direct connection equivalent path group is less than a quantity threshold. For example, the total number is relatively the least (such as 1), so as to control the number of detour network device groups that the detour packets need to pass through when going from the second network device to the destination network device group, and avoid the transmission path of the detour packets from forming a loop outside the group, and further improve the packet transmission efficiency.

[0010] According to the first aspect, or any one of the implementation manners of the above first aspect, the inter-group detour rule includes: the transmission distance corresponding to the equivalent path between the second upstream network device and the destination network device group satisfies a distance threshold.

[0011] In the embodiment of the present application, the inter-group detour rule includes that the equivalent path between the second upstream network device and the destination network device group is a path whose transmission distance satisfies the distance threshold, so as to ensure that the number of inter-group hops of the packets in the network formation belongs to the expected number of hops and ensure the efficiency requirement of packet transmission.

[0012] According to the first aspect, or any one of the implementation manners of the above first aspect, when the detour packets include a first packet and a second packet, and the first packet and the second packet respectively correspond to different destination network device groups, the number of common equivalent path groups is at least one.

[0013] In the embodiment of the present application, when the detour packets include multiple packets going to different destination network device groups, the number of common equivalent path groups can be at least one. That is to say, when the first downstream network device coordinates and schedules the detour packets, it is not necessary to distinguish the destination network device groups of different packets, and a common equivalent path group can be directly shared, which is more convenient and efficient.

[0014] According to the first aspect, or any one of the implementation manners of the above first aspect, the detour packets carry a detour mark, and the detour mark is used for the first upstream network device to identify the detour packets in the received packets; the adding time of the detour mark in the detour packets includes: before the source network device group sends the detour packets to the detour network device group, or before the detour network device group receives the detour packets and sends the detour packets to the first downstream network device.

[0015] In an embodiment of the present application, by adding a detour mark to a detour message before the detour message is sent from the source network device group or when the detour message arrives at the detour network device group, it is ensured that the first upper network device can accurately identify the detour message from the received messages, further improving the performance of message transmission.

[0016] In a second aspect, an embodiment of the present application provides a message transmission method, which is applied to an upper network device. The upper network device is deployed in a detour network device group in a network and is docked with devices outside the detour network device group. The detour network device group includes multiple upper network devices and multiple lower network devices docked with computing devices in the network. Each upper network device and each lower network device are communicatively connected through an intra-group interconnection port. The network also includes a source network device group and a destination network device group. The destination network device group and each upper network device are communicatively connected through an inter-group interconnection port. The method includes: Receiving a detour message, where the detour message includes a message from the source network device group and to be forwarded by the detour network device group to the destination network device; According to a common equivalent path group, forwarding the detour message to a first lower network device among the multiple lower network devices. The common equivalent path group is used to indicate an equivalent path formed by a first group of intra-group interconnection ports that meet the load balancing rule in each intra-group interconnection port, and the first lower network device corresponds to the first group of intra-group interconnection ports; When the equivalent path between the upper network device and the destination network device group meets the inter-group detour rule, receiving the detour message forwarded by the first lower network device according to a coordinated equivalent path group. The coordinated equivalent path group is used to indicate an equivalent path formed by a second group of intra-group interconnection ports in each intra-group interconnection port. The second group of intra-group interconnection ports meets the load balancing rule and corresponds to the upper network device that receives the detour message forwarded by the first lower network device; According to a direct connection equivalent path group, forwarding the detour message received from the first lower network device to the destination network device group. The direct connection equivalent path group is used to indicate an equivalent path formed by a first group of inter-group interconnection ports that meet the inter-group detour rule in each inter-group interconnection port.

[0017] According to the second aspect, when the detour network device group is divided into a single federated routing autonomous system, each upper network device and each lower network device respectively correspond to a single routing autonomous system identifier under the federated routing autonomous system. Each equivalent path indicated by the direct connection equivalent path group satisfies one or more of the following conditions: Any one of the equivalent paths indicated by the direct connection equivalent path group corresponds to at least one routing autonomous system identifier, and any one of the at least one is different from the single routing autonomous system identifier of the federated routing autonomous system to which the local device belongs; The total number of routing autonomous system identifiers corresponding to the equivalent paths indicated by the direct connection equivalent path group is less than a quantity threshold.

[0018] According to the second aspect, or any implementation manner of the above second aspect, the inter-group rotation rule includes that the transmission distance corresponding to the equivalent path between the upper network device and the destination network device group satisfies a distance threshold.

[0019] According to the second aspect, or any implementation manner of the above second aspect, when the detour message includes a first message and a second message, and the first message and the second message respectively correspond to different destination network device groups, the number of common equivalent path groups is at least one.

[0020] According to the second aspect, or any implementation manner of the above second aspect, the detour message carries a detour mark, and the detour mark is used for the first upper network device to identify the detour message in the received message; The timing of adding the detour mark in the detour message includes: before the source network device group sends the detour message to the rotation network device group, or, after the rotation network device group receives the detour message and before sending the detour message to the first lower network device.

[0021] The second aspect and any implementation manner of the second aspect respectively correspond to the first aspect and any implementation manner of the first aspect. For the technical effects corresponding to the second aspect and any implementation manner of the second aspect, reference can be made to the technical effects corresponding to the first aspect and any implementation manner of the first aspect above, which will not be elaborated here.

[0022] Thirdly, an embodiment of the present application provides a message transmission method, which is applied to a lower network device. The lower network device is deployed in a rotating network device group in a network and is docked with a computing device in the network. The rotating network device group includes a plurality of lower network devices and a plurality of upper network devices that are docked with devices outside the rotating network device group. Each upper network device and each lower network device are communicatively connected through an intra-group interconnection port. The network further includes a source network device group and a destination network device group. The destination network device group and each upper network device are communicatively connected through an inter-group interconnection port. The method includes: forwarding a detouring message received from a first upper network device to a second upper network device among the plurality of upper network devices according to a coordinated equivalent path group, where the coordinated equivalent path group is used to indicate an equivalent path formed by a second intra-group interconnection port in each intra-group interconnection port, and the second intra-group interconnection port satisfies a load balancing rule and corresponds to the second upper network device; where the first upper network device includes any one of the plurality of upper network devices, and the detouring message includes a message from the source network device group and to be forwarded by the rotating network device group to the destination network device; the equivalent path between the second upper network device and the destination network device group satisfies an inter-group detouring rule and is used to forward the received detouring message to the destination network device group according to a direct-connected equivalent path group, where the direct-connected equivalent path group is used to indicate an equivalent path formed by a first inter-group interconnection port that satisfies the inter-group detouring rule in each inter-group interconnection port.

[0023] According to the third aspect, in the case where the rotating network device group is divided into a single federated routing autonomous system, each upper network device and each lower network device respectively correspond to a single routing autonomous system identifier under the federated routing autonomous system to which the local machine belongs. Each equivalent path indicated by the direct-connected equivalent path group satisfies one or more of the following conditions: any equivalent path indicated by the direct-connected equivalent path group corresponds to at least one routing autonomous system identifier, and any one of the at least one is different from the single routing autonomous system identifier under the federated routing autonomous system to which the local machine belongs; the total number of routing autonomous system identifiers corresponding to each equivalent path indicated by the direct-connected equivalent path group is less than a quantity threshold.

[0024] According to the third aspect, or any implementation manner of the above third aspect, the inter-group detouring rule includes: the transmission distance corresponding to the equivalent path between the upper network device and the destination network device group satisfies a distance threshold.

[0025] The third aspect and any implementation manner of the third aspect respectively correspond to the first aspect and any implementation manner of the first aspect. The technical effects corresponding to the third aspect and any implementation manner of the third aspect can be referred to the technical effects corresponding to the first aspect and any implementation manner of the first aspect above, and will not be elaborated here.

[0026] Fourth aspect, an embodiment of the present application provides a message transmission system. The system includes a rotating network device group, a source network device group, and a destination network device group. The rotating network device group includes a plurality of upper network devices that are docked with devices outside the rotating network device group, and a plurality of lower network devices that are docked with computing devices in the network. Each upper network device and each lower network device are communicatively connected through an intra-group interconnection port. The destination network device group and each upper network device are communicatively connected through an inter-group interconnection port. A first upper network device is configured to receive a detour message, and forward the detour message to a first lower network device among the plurality of lower network devices according to a common equivalent path group. The first upper network device includes any one of the plurality of upper network devices. The detour message includes a message from the source network device group and to be forwarded by the rotating network device group to the destination network device. The common equivalent path group is used to indicate an equivalent path formed by a first group of intra-group interconnection ports that meet the load balancing rule in each intra-group interconnection port. The first lower network device corresponds to the first group of intra-group interconnection ports. The first lower network device is configured to forward the received detour message to a second upper network device among the plurality of upper network devices according to a coordinated equivalent path group. The coordinated equivalent path group is used to indicate an equivalent path formed by a second group of intra-group interconnection ports in each intra-group interconnection port. The second group of intra-group interconnection ports meet the load balancing rule and correspond to the second upper network device. The equivalent path between the second upper network device and the destination network device group meets the inter-group detour rule. The second upper network device is configured to forward the received detour message to the destination network device group according to a direct connection equivalent path group. The direct connection equivalent path group is used to indicate an equivalent path formed by a first group of inter-group interconnection ports that meet the inter-group detour rule in each inter-group interconnection port.

[0027] According to the fourth aspect, in the case where the rotating network device group is divided into a single federated routing autonomous system, each upper network device and each lower network device respectively correspond to a single routing autonomous system identifier under the federated routing autonomous system. Each of the equivalent paths indicated by the direct connection equivalent path group satisfies one or more of the following conditions: any one of the equivalent paths indicated by the direct connection equivalent path group corresponds to at least one routing autonomous system identifier, and any one of the at least one is different from the single routing autonomous system identifier of the federated routing autonomous system to which the local device belongs; the total number of routing autonomous system identifiers corresponding to the equivalent paths indicated by the direct connection equivalent path group is less than a quantity threshold.

[0028] According to the fourth aspect, or any one of the implementation manners of the above fourth aspect, the inter-group detour rule includes: the transmission distance corresponding to the equivalent path between the fourth upper network device and the destination network device group satisfies a distance threshold.

[0029] According to the fourth aspect, or any implementation manner of the above fourth aspect, when the detour message includes a first message and a fourth message, and the first message and the fourth message respectively correspond to different destination network device groups, the number of common equivalent path groups is at least one.

[0030] According to the fourth aspect, or any implementation manner of the above fourth aspect, the detour message carries a detour flag, and the detour flag is used for the first upstream network device to identify the detour message in the received message; the adding time of the detour flag in the detour message includes: before the source network device group sends the detour message to the detouring network device group, or, before the detouring network device group receives the detour message and sends the detour message to the first downstream network device.

[0031] The fourth aspect and any implementation manner of the fourth aspect respectively correspond to the first aspect and any implementation manner of the first aspect. For the technical effects corresponding to the fourth aspect and any implementation manner of the fourth aspect, reference can be made to the technical effects corresponding to the first aspect and any implementation manner of the first aspect above, which will not be elaborated here.

[0032] Fifth aspect, an embodiment of the present application provides a network device, including: a processor and a memory; the memory is used for storing instructions; the processor is used for executing the instructions stored in the memory, so that the network device executes the method as described in any item of the first aspect and any implementation manner of the first aspect.

[0033] Sixth aspect, an embodiment of the present application provides a computing device cluster, including at least one computing device, each computing device is connected to at least one downstream network device, the downstream network devices belong to the detouring network device group, the detouring network device group further includes multiple upstream network devices, and each upstream network device and each downstream network device are communicatively connected through an intra-group interconnection port, the upstream network device is connected to a device outside the detouring network device group, and each downstream network device and each upstream network device respectively include a processor and a memory; The processor of the downstream network device is used for executing the instructions stored in the memory of the downstream network device, and the processor of the upstream network device is used for executing the instructions stored in the memory of the upstream network device, so that the computing device cluster executes the instructions of the method in the first aspect or any possible implementation manner of the first aspect.

[0034] Seventh aspect, an embodiment of the present application provides a computer program product containing instructions, when the instructions are run by the computing device cluster, the network device for transmitting messages in the computing device cluster executes the instructions of the method in the first aspect or any possible implementation manner of the first aspect.

[0035] In an eighth aspect, an embodiment of the present application provides a computer program, including a computer program, which, when executed by a network device, causes the network device to execute the instructions of the method in the first aspect or any possible implementation manner of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0037] Figure 1 is a scenario example diagram of a message being transmitted through multiple aggregation switch groups in a network; Figure 2a is an example process of a message being transmitted through multiple aggregation switch groups in a network Figure 1 ; Figure 2b is the second example diagram of the process of a message being transmitted through multiple aggregation switch groups in a network; Figure 3 is an example diagram of an aggregation switch in an aggregation switch group; Figure 4 is one of the example diagrams of the entry cost of an equivalent path table used by an upper aggregation switch in an aggregation switch group; Figure 5 is one of the example diagrams of the entry cost of an equivalent path table used by an upper aggregation switch in an aggregation switch group; Figure 6 is one of the example diagrams of the entry cost of an equivalent path table used by an upper aggregation switch in an aggregation switch group; Figure 7 is one of the example diagrams of the entry cost of an equivalent path table used by an upper aggregation switch in an aggregation switch group; Figure 8 is an example diagram of an application scenario of a message transmission method provided by an embodiment of the present application; Figure 9 is a schematic flow diagram of a message transmission method provided by an embodiment of the present application; Figure 10 is a comparison diagram of an example of a message transmission path and an example of a message transmission path in the related art under the message transmission method provided by an embodiment of the present application; Figure 11 is a schematic framework diagram of a network device provided by an embodiment of the present application; Figure 12 is a schematic framework diagram of a computing device cluster provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0039] The term "and / or" in this document is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0040] The terms "first", "second", etc. in the description and claims of the embodiments of the present application are used to distinguish different objects, rather than to describe a specific order of the objects. For example, the first target object and the second target object are used to distinguish different target objects, rather than to describe the specific order of the target objects.

[0041] In the embodiments of the present application, words such as "exemplary" or "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the embodiments of the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "exemplary" or "for example" is intended to present relevant concepts in a specific manner.

[0042] In the description of the embodiments of the present application, unless otherwise specified, the meaning of "a plurality of" refers to two or more. For example, a plurality of processing units refers to two or more processing units; a plurality of systems refers to two or more systems. To facilitate the understanding of this embodiment, some technical terms and background technologies involved in this embodiment will be introduced first: (1) Optical cross-connect (OXC): A device that can implement signal cross-connection and routing selection in the optical domain, also known as an optical cross-connector. It directly processes optical signals without the need for optical-electrical-optical conversion.

[0043] (2) Optical circuit switch (OCS): A device that implements signal exchange in the optical domain. The core function is to accurately switch the optical signal of the input port to the specified output port according to certain rules and instructions, and to establish and remove the optical path to meet the communication needs of different users and services. OCS performs switching directly in the optical domain, avoiding the delay caused by electronic devices in the optical-electrical-optical conversion process. The signal transmission delay is usually in microseconds or even lower.

[0044] (3) Dragonfly+ topology networking: A network topology used in data centers and high-performance computing. It is an extension of the Dragonfly topology with better performance and scalability. The Dragonfly topology is a new type of ultra-large-scale direct connection topology.

[0045] (4) Equal-cost paths: In routing selection, equal-cost paths refer to multiple paths with the same metric to the same destination network. When a router selects a path to reach a destination network, it calculates the metrics of different paths based on the routing protocol. Common metric types include hop count, bandwidth, delay, cost, etc. If multiple paths have the same metric, then these paths are considered equal-cost paths.

[0046] (5) Entry: It can also be called an entry or item. For example, in a routing scenario, each record in a routing table (such as an equal-cost path table) is an entry. The entry contains information about the destination network, next-hop address, path metric, etc., which is used to guide the forwarding of data packets in the network.

[0047] (6) Trunk group: Port grouping achieved through link aggregation technology: multiple physical ports are bundled into one logical port, so that the physical links corresponding to these physical ports can work as one link. Data will be distributed on these bundled physical links to achieve load balancing, and also provide redundant backup function.

[0048] (7) Weighted cost multipath routing (WCMP): Packets are distributed in proportion to the weights of the links. The router assigns a weight value to each path, which represents the relative merits of the path. Factors such as link bandwidth, latency, and cost can affect the weight setting. When packets are transmitted, they are distributed to different links according to the weight ratio of the path.

[0049] (8) Hash polarization: When implementing load balancing in a network using a hash algorithm, a hash operation is usually performed based on certain packet information (such as source IP address, destination IP address, port number, etc.) to determine the path through which a data packet, such as a message, is transmitted. If, in a scenario of multi-level device networking or multiple hash calculations, the hash calculation result of a subsequent device is affected by the hash result of a previous device, it may cause messages to be unevenly distributed across links, resulting in Hash polarization.

[0050] (9) Border Gateway Protocol (BGP): A path-vector routing protocol used to exchange routing information between different routing autonomous systems (AS).

[0051] (10) Routing autonomous system (AS): Also known as an autonomous system, it refers to a collection of network devices (such as routers, switches, etc.) under the control of a single technical management agency. These network devices use an Interior Gateway Protocol (IGP) and a common metric to determine the routing of packets within the system and exchange routing information with other autonomous systems through the Border Gateway Protocol (BGP).

[0052] (11) Federated AS: A larger-scale network entity formed by combining multiple autonomous systems through specific technologies and strategies. The federated AS allows these autonomous systems to maintain a certain degree of independence and autonomy while enabling communication and collaboration among them to jointly provide a certain network service or achieve specific network functions. Each autonomous system within the federated AS can independently perform operations such as internal routing selection, network management, and policy formulation. However, at the federated level, there are also some unified management and coordination mechanisms to regulate the interaction and cooperation among autonomous systems and ensure the normal operation of the entire federated AS as a whole.

[0053] (12) Single autonomous system identifier (local autonomous system number, local ASN): Also known as the local autonomous system identifier or local AS identifier, it is the unique identifier of a single autonomous system (AS) used to identify the routing boundary of the AS in a global or local network. Each AS uses its own local ASN for routing management within the system.

[0054] (13) The forwarding control method of network devices (such as routers or switching devices) in routing, taking a router as an example of a network device: Redirect: A means that allows data packets to change their original forwarding paths and be forwarded according to new instructions. After configuring the redirection policy, the router will perform special processing on data packets that meet specific conditions according to these policies, redirecting them to other paths or interfaces for forwarding, rather than following the instructions in the conventional routing table.

[0055] Forwarding control based on MPLS (Multi-Protocol Label Switching): MPLS assigns a label to a data packet when it enters the network, and then the router forwards the packet according to the label, rather than performing a lookup and forwarding based on the IP address as in traditional routing. The assignment of labels and the forwarding paths are predetermined by the control plane in the network.

[0056] Forwarding control based on Access Control List (ACL): ACL filters data packets according to a series of defined rules based on information such as the source IP address, destination IP address, and port number of the data packet. An ACL can be applied to the interfaces of a router to allow or deny data packets that meet specific conditions to pass through, thereby achieving control over routing forwarding.

[0057] (14) Virtual Local Area Network (VLAN): A technology that logically divides the devices within a local area network rather than physically into different network segments. Through VLAN technology, a physical LAN can be divided into multiple logically isolated VLANs, and each VLAN can be regarded as an independent broadcast domain. Devices in different VLANs cannot communicate directly by default, and communication can only be achieved through routing forwarding by devices such as routers or layer-3 switches.

[0058] (15) priority field: Usually appears in the headers of some network protocols and is used to indicate the priority of data packets. The priority field is a numerical field, and the specific meaning and value range of the priority field may vary in different protocols.

[0059] In this application, the network device group used for message transmission can be, for example, an aggregation network device group. The aggregation network device group can aggregate the messages at the access layer in the network and send them to the core layer of the network. It can be understood that the aggregation network device group, such as an aggregation switch group, an aggregation router group, etc. are only examples, and the embodiments of this application do not limit the layer where the network device group is located and the specific network devices within the network device group. For the convenience of understanding and description, the aggregation switch group and the aggregation switch will be used as examples for subsequent description. For example, in the subsequent description, the rotation network device group can be, for example, a rotation aggregation switch group, the upper network device can be, for example, an upper aggregation switch, the lower network device can be, for example, a lower aggregation switch, the source network device group can be, for example, a source aggregation switch group, and the destination network device group can be, for example, a destination aggregation switch group. In a specific application, the network device can be, for example, a switch, a router, a hub, a bridge, a layer-3 switch, or a wireless access point, etc. The embodiments of this application do not limit the specific form of the network device and can be set according to application requirements.

[0060] The network (such as a data center, Ethernet, etc.) can form a topology through multiple network device groups to transmit messages (such as in a dragonfly+ topology network) between the host clusters in the network and / or between the host clusters in the network and the devices outside the network. Among them, the network device group refers to a device group that contains multiple network devices (such as switches or routers, etc.). The topology between multiple network device groups can be a fixed uniform topology, or the optical path can be changed through an optical switching device (such as an optical cross-connect or an optical circuit switch, etc.) to flexibly adjust the inter-group topology connection relationship, taking into account lower transmission costs and higher transmission performance.

[0061] Exemplarily, Figure 1 is an example diagram of the scenario where messages in the network are transmitted through multiple aggregation switch groups. As Figure 1 shown, in addition to using the direct connection path between groups for message transmission, the rotation path between groups can also be used. For example, the message transmission from host H1 to host H2 can be achieved through the direct connection path between aggregation switch group SG1 and aggregation switch group SG3, or by using aggregation switch group SG2 for rotation and through the rotation path from aggregation switch group SG1 to aggregation switch group SG3. When the message is rotated and forwarded through aggregation switch group SG2, since the connection ports between aggregation switch group SG2 and aggregation switch group SG3 may be distributed on multiple different upper aggregation switches, load sharing often needs to be performed according to the load congestion conditions of these ports. For the scenario where the rotation aggregation switch group can independently evaluate the routing forwarding decision, aggregation switch group SG2 needs to try its best to meet the forwarding and load sharing requirements of these detoured messages according to its own capabilities.

[0062] In the related art, after a packet arrives at a rotary switching unit, each switch sets weights according to equivalent paths to perform load sharing on the packet. Affected by the entry specifications within the equivalent path group in the forwarding chip, the load sharing is usually best-effort, and there is a large deviation between the weights issued in the actual chip and the ideal setting. Exemplarily, Figure 2a and Figure 2b are example diagrams of the process of a packet being transmitted through multiple aggregation switching units in a network. As Figure 2a and 2b show, taking a simplified network model as an example: Each aggregation switching unit consists of 4 devices, including two upper aggregation switches and two lower aggregation switches. Each upper aggregation switch is connected to one lower aggregation switch by 1 link; for example, in Figure 2a , the upper aggregation switch 1 in the aggregation switching unit SG2 is connected to the lower aggregation switch 1 through port 3; in Figure 2b , the upper aggregation switch 1 in the aggregation switching unit SG2 is connected to the lower aggregation switch 2 through port 4. Among them, ports 1 and 2 are output ports. Assuming that both ports 1 and 2 are idle, the aggregation switching unit SG2 hopes to load share half of the packet from each of ports 1 and 2. Then, first, at the first-hop switch of the aggregation switching unit SG2 where the packet arrives, that is, the upper aggregation switch 1, control the forwarding and load sharing.

[0063] The upper aggregation switch 1 receives the route issued by the aggregation switching unit SG3 through port 1, and the packet can be forwarded out through the transmission path P1. The upper aggregation switch 2 receives the route issued by the aggregation switching unit SG3 through port 2, and further issues it to the lower aggregation switch 1 and the lower aggregation switch 2, and then sends it to the upper aggregation switch 1. The packet can be forwarded out through the transmission path P2 and the transmission path P3. In this way, there are 3 equivalent paths in the equivalent path group to the destination, where the weight of the transmission path P1 is 50%, and the weights of the transmission paths P3 and P2 are 25% respectively. To achieve such a weight effect, the equivalent path group used by the upper aggregation switch 1 (such as stored in the forwarding chip of the upper aggregation switch 1), that is, the entries filled in the equivalent path table, can be, for example, as shown in Table 1 below: Table 1 Example of Entries in the Equivalent Path Table

[0064] Among them, the ratio of the number of entries occupied by the next hop corresponding to the transmission path P1 to the number of entries occupied by the next hop corresponding to the transmission paths P2 and P3 is 2:1:1.

[0065] When the load congestion situations of the above-mentioned port 1 and the above-mentioned port 2 are inconsistent, the situation will change. For example, port 1 is idle and port 2 is congested. Assuming that the ratio of the packets forwarded through port 1 to the packets forwarded through port 2 needs to be set to 4:1 at this time, an example of the entries filled in the forwarding chip of the upper aggregation switch 1 corresponding to this equivalent path group is shown in Table 2 below: Table 2 Example 2 of the Entries in the Equivalent Path Table

[0066] Among them, the ratio of the number of entries occupied by the next hop corresponding to the transmission path P1 to the number of entries occupied by the next hops corresponding to the transmission paths P2 and P3 is 8:1:1. It can be understood that the index distribution of the next hop corresponding to the transmission path P1 in the above equivalent path table may not be continuous and can be more discrete, so that the hash result will be better. Table 2 is a simple example of continuous distribution given for the convenience of explaining the ratio of the number of entries occupied by each next hop in the table.

[0067] The above Figure 2a and Figure 2b shown is a simplified networking model. It can be understood that in actual application scenarios, for example, in the application scenario of a data center, the scale of the aggregation switch group is usually much larger than the example shown in Figure 2a and Figure 2b For example, a typical Ethernet aggregation switch has 64 800GE ports. After splitting, 32 ports in each aggregation switch are connected to the outside, and 32 ports are connected to the aggregation switches within the group. The upper aggregation switch is connected to the lower aggregation switch through 32 ports. 8 upper aggregation switches and 8 lower aggregation switches can be used for networking, or 32 upper aggregation switches and 32 lower aggregation switches can be used to form an aggregation switch group. Exemplarily, Figure 3 is an example diagram of the aggregation switch in the aggregation switch group. As shown in Figure 3 , for the aggregation switch, such as the upper aggregation switch 1, ports 33 to 64 have the same role and status as ports 3 in Figure 2a and port 4 in Figure 2b . At this time, there are 32 paths equivalent to the transmission path P2 or the transmission path P3.

[0068] Exemplarily, Figure 4 is one of the example diagrams of the entry cost of the equivalent path table used by the upper aggregation switch in the aggregation switch group. As shown in Figure 4 , for Figure 3In the large-scale scenario shown, if the upper aggregation switch 1 wants to achieve a ratio of 1:1 for the packets forwarded through port 1 to those forwarded through port 2, then the number of entries occupied by the next hop corresponding to the path directly forwarded from port 1 is at least 32.

[0069] Exemplarily, Figure 5 is one of the diagrammatic examples of the entry cost of the equivalent path table used by the upper aggregation switch in the aggregation switch group. As Figure 5 shown, for Figure 3 the large-scale scenario shown, if it is desired to achieve a ratio of 4:1 for the packets forwarded through port 1 to those forwarded through port 2, then the number of entries occupied by the next hop corresponding to the path directly forwarded from port 1 is at least 128.

[0070] In this way, for the upper aggregation switch 1 to achieve load sharing among the ports distributed on different upper aggregation switches, a large number of entries in the equivalent path table need to be consumed. And this is only for the packets going to one destination. In a large-scale networking scenario, such as a network composed of 256 aggregation switch groups, it is necessary to support the packet forwarding to a maximum of 255 destination aggregation switch groups.

[0071] Exemplarily, Figure 6 is one of the diagrammatic examples of the entry cost of the equivalent path table used by the upper aggregation switch in the aggregation switch group. As Figure 6 shown, if considering consistent hashing again and keeping the hash corresponding entry index result of the packets that have already been load-shared unchanged when adjusting the weights of each path, then the total number of entries in the equivalent path table needs to be fixed. At this time, in order to balance flexible ratio adjustment, a relatively large total number of entries needs to be set, such as 400. Then, if it is desired to achieve a ratio of 1:1 for the packets forwarded through port 1 to those forwarded through port 2, the next hop corresponding to port 1 occupies 200 entries. Exemplarily, Figure 7 is one of the diagrammatic examples of the entry cost of the equivalent path table used by the upper aggregation switch in the aggregation switch group. As Figure 7 shown, in the same scenario as Figure 6 above, if the ratio of the packets forwarded through port 1 to those forwarded through port 2 is adjusted to 4:1, then the next hop corresponding to port 1 occupies 320 entries.

[0072] In actual common application scenarios, they are all large-scale networking. The networking specifications within the group (such as forming a group with 64 aggregation switches) and the inter-group specifications (such as 256 groups) are relatively high. In a fixed uniform topology or in a flexible topology scenario constructed by OXC networking, the ports for the interconnection between one aggregation switch group and another aggregation switch group will never be deployed on only one upper aggregation switch, but will be dispersed as much as possible to ensure high reliability.

[0073] As in the above examples, when a detour packet, that is, a packet that needs to be forwarded by a detour aggregation switch group, enters the upper aggregation switch of the detour aggregation switch group, based on the traditional equal-cost path grouping rules in the above examples, according to the determination condition of whether it goes to the same destination aggregation switch group or according to whether it is a service packet with the same characteristics, it is very likely that the next hops with the internal interconnection port as the output port and the next hops with the external interconnection port as the output port are mixed in an equal-cost path group for weighted load sharing processing. In this way, for a certain upper aggregation switch, after receiving a detour packet from the external interconnection port, the number of next hops of the forwarding path through the internal interconnection port is often dozens more than that of the path directly connected to the destination aggregation switch group by this device. However, the sum of the weights of the numerous next hops of the internal interconnection port forwarding is very likely to be much smaller than the weight of the path directly connected to the destination aggregation switch group by this device, that is, the load sharing is not balanced enough.

[0074] In summary, during the internal forwarding process of the aggregation switch group, in the prior art, the forwarding and load sharing responsibilities of the aggregation switch members are often not analyzed in combination with the characteristics of the actual network topology. A very small number of external interconnection ports and a very large number of internal interconnection ports are mixed in an equal-cost path group for load sharing, resulting in the following two major drawbacks: Drawback 1: As the network topology scale increases, the consumption of equal-cost path entries in the chip implementation is huge: In order to achieve load balancing of the output ports on different aggregation switches in the aggregation switch group, a large number of equal-cost path table entries need to be consumed. For example, in a data center network topology composed of 256 aggregation switch groups, as an upper aggregation switch in the detour aggregation switch group, to support going to 255 destination aggregation switch groups, even if only one equal-cost path group (usually multiple) is established for each destination aggregation switch group, and each group consumes 400 entries, then a total of at least 255×400 = 102k entries are required for 255 destination groups. In addition, this switch group may also consume relevant entry numbers as members of the source and destination aggregation switches. However, the total scale of the equal-cost path table entries that the network device can support is limited. For example, the total number of equal-cost path table entries supported by a 51.2T switch does not exceed 128k.

[0075] Disadvantage 2: Uneven load sharing during the forwarding process from the upper aggregation switch to the lower aggregation switch: Since there are 32 or even more ports for the upper aggregation switch to interconnect with the lower aggregation switch, in the case of a given total number of entries, it is often impossible to ensure that the load on each output port is basically uniform. As in the 4:1 packet ratio scenario in the previous section, 32 output ports need to share 80 entries. At this time, the next hops of half of the output ports occupy 3 entries, and the next hops of the other half of the output ports occupy 2 entries, 16 × 3 entries + 16 × 2 entries = 80. Then the output ports that occupy 3 entries are more likely to hash to more packets, while the output ports that occupy 2 entries may hash to fewer packets. As the scale within the aggregation switch group continues to increase, the problem of uneven load sharing will become more serious. Some networks choose to create a trunk group for the ports connecting to the same aggregation switch, which will also result in the situation of two-level hashing of WCMP and the trunk group, possibly increasing the probability of hash polarization and leading to uneven load.

[0076] It can be seen that the number of entries in the equivalent path table that a network device can support, that is, the table entry scale, is limited, while the trend of the networking scale is getting larger and larger. Then there are great limitations in the number of entries in the equivalent path table allocated to each destination, making the problem of uneven load sharing more obvious. That is to say, the load sharing between different types of next hops extremely wastes the number of equivalent path entries: The number of internal interconnection ports of an aggregation switch is generally at least in the dozens. As a result, among the next hop members of the equivalent path group, once the output port included is the next hop of the internal interconnection port, normally there will be dozens of equivalent next hops of different output ports. For example, when a next hop of an external interconnection port shares the load with dozens of next hops of internal interconnection ports, even if the ratio is 1:1, it will occupy twice the number of entries of the next hops of the internal interconnection ports.

[0077] In response to this, the embodiments of the present application provide a packet transmission method to solve the above problems. In this method, the first upper network device in the detour network device group that is used to receive the detour packets sent by the source network device group forwards the detour packets to the first lower network device, so as to ensure that the types of the interconnected ports in the common equivalent path group indicating the transmission path of the detour packets during this process are all intra-group interconnected ports. On this basis, the first lower network device forwards the detour packets to the second upper network device in the detour network device group that docks with the destination network device group, so as to ensure that the types of the interconnected ports in the overall equivalent path group indicating the transmission path of the detour packets during this process are all intra-group interconnected ports. In this way, when the second upper network device forwards the detour packets to the destination network device group, it can ensure that the types of the interconnected ports in the direct connection equivalent path group indicating the transmission path of the detour packets during this process are all extra-group interconnected ports. That is to say, through the forwarding or overall scheduling of the lower network device, the various equivalent path tables corresponding to the transmission path of the detour packets can be adjusted to have the same type of interconnected ports involved, so as to avoid different types of interconnected ports gathering in an equivalent path group, that is, an equivalent path table, and greatly reduce the consumption of table entries in an equivalent path group. Therefore, the solution of the present application can achieve load sharing of the transmitted packets at the cost of fewer table entries in the equivalent path group, so as to more fully meet the load sharing requirements of packet transmission in the network under the limited table entry scale of the equivalent path table.

[0078] Before describing the technical solutions of the embodiments of the present application, for the sake of easy understanding, first, the application scenario of the packet transmission method of the embodiments of the present application will be described with reference to the accompanying drawings. The packet transmission method of the embodiments of the present application can be applied to a network, such as a data center, an Ethernet, etc. This network is not limited to a flexible topology network with OXC or OCS. It can also be applied to the scenario where packets can be detoured and forwarded between aggregation switch groups in a fixed topology network, such as a typical dragonfly+ network. For example, Figure 8 is an example diagram of the application scenario of a packet transmission method provided by the embodiments of the present application. As Figure 8 shown, taking the network as a data center as an example, multiple aggregation switch groups (such as aggregation switch group SG1 to aggregation switch group SG3) form the internal network of the data center, which is used to implement packet transmission between various computing devices in the data center, such as between multiple host cabinets, and packet transmission between the data center and devices and / or networks outside the data center. Among them, different aggregation switch groups can be connected through an optical cross-connect device group to flexibly adjust the topology structure. The optical cross-connect device group can specifically be a device group formed by multiple OXC or multiple OCS.

[0079] Among them, each aggregation switch group may include a plurality of upper aggregation switches and a plurality of lower aggregation switches. For example, upper aggregation switch 1 to upper aggregation switch n, lower aggregation switch 1 to lower aggregation switch n, where n is an integer greater than 1. The upper aggregation switch is also referred to as a spine switch, and the lower aggregation switch is also referred to as a leaf switch. The present application does not limit the names of the aggregation switches.

[0080] In the embodiments of the present application, the aggregation switch group may include a network device group that forwards and communicates between a host cluster locally accessed by the aggregation switch group and a host cluster outside the aggregation switch group. The upper aggregation switch is used to connect to devices outside the aggregation switch group where it is located. For example, other aggregation switches and / or devices outside the network formation, etc. In one possible implementation, the lower aggregation switch is used to connect to computing devices in the network formation, such as Figure 8 the shown main cabinet.

[0081] The network device provided by the embodiments of the present application supports core-layer-free forwarding of packets, and the packets detour between different network device groups, that is, inter-group detouring. For example, the network device may support packet detouring between different pods; among them, each pod includes at least one network device group, and a network formation may be divided into multiple pods.

[0082] The network device group provided by the embodiments of the present application supports independently determining the packet forwarding path, and may also support a control device (such as a cloud management platform, a computing device with a cluster management function, a network device with a routing control function, etc.) to set routing rules through the packet header, and the network device group parses the routing rules for packet transmission.

[0083] In addition, the network device provided by the embodiments of the present application may support controlling the number of detouring aggregation switch groups or detouring pods through an equivalent path group. It can be understood that the application scenario of the detouring pod can refer to the application scenario of the detouring aggregation switch group, and the detouring pod includes the detouring aggregation switch group.

[0084] It should be understood that Figure 8 the shown network formation is only an example of the network formation, and moreover, the network formation or the packet transmission system may have more or fewer components and / or devices than those shown in the figure. Figure 8 The various components and / or devices shown in can be implemented in hardware, software, or a combination of hardware and software including one or more signal processing and / or application specific integrated circuits. Figures 9 to 10 The following specifically describes the packet transmission method provided by the embodiments of the present application in conjunction with

[0085] Exemplarily, Figure 9It is a schematic flowchart of a message transmission method provided by an embodiment of the present application. As Figure 9 shown, an embodiment of the present application provides a message transmission method, which is applied to a rotating network device group in a network. For example, in the network device group in the above Figure 8 shown network, the rotating network device group includes multiple upper network devices docked with devices outside the rotating network device group, and multiple lower network devices docked with computing devices in the network. Each upper network device is communicatively connected to each lower network device through an intra-group interconnection port. The network also includes a source network device group and a destination network device group. The destination network device group is communicatively connected to each upper network device through an inter-group interconnection port. The method may include: S901, a first upper network device receives a detour message; wherein, the first upper network device includes any one of the multiple upper network devices, and the detour message includes a message from the source network device group and to be forwarded by the rotating network device group to the destination network device.

[0086] In an optional implementation manner, the detour message carries a detour mark, and the detour mark is used for the first upper network device to identify the detour message in the received message; The timing of adding the detour mark in the detour message includes: before the source network device group sends the detour message to the rotating network device group, or, after the rotating network device group receives the detour message and before sending the detour message to the first lower network device.

[0087] Exemplarily, in order to identify and match the detour message in the rotating aggregation switch group, one method is that the detour message can add a mark when leaving the source aggregation switch group, and another method is to add a mark when arriving at the rotating aggregation switch group and erase the mark when leaving the rotating aggregation switch group. General data center switches have the ability to do this. For the case where all ports of the upper aggregation switch and the internal interconnection ports of the lower aggregation switch are all three-layer main interfaces, the detour message can be easily marked by using the method of adding a reserved VLAN (Virtual Local Area Network) and Priority field, and then when matching the message with an access control list (ACL) on the switch device, combining the detour mark of the message and the role of the incoming port (internal interconnection port, external interconnection port) of the device entering the device can identify the detour message.

[0088] In the embodiment of the present application, by adding a detour mark to the detour message before the detour message is sent from the source network device group or when the detour message arrives at the rotating network device group, it is ensured that the first upper network device can accurately identify the detour message from the received messages, further improving the performance of message transmission.

[0089] S902, the first upstream network device forwards the detour packet to the first downstream network device among multiple downstream network devices according to a common equivalent path group, where the common equivalent path group is used to indicate an equivalent path formed by the first group of interconnected ports that meet the load balancing rule among the interconnected ports within each group; Among them, the first downstream network device corresponds to the first group of interconnected ports. In a specific application, in a scenario where the detour aggregation switch group can independently determine the path for packet forwarding, the detour aggregation switch group can determine each path group by itself, such as the above-mentioned common equivalent path group; in the case where the detour aggregation switch group or the aggregation switch determines each path group by parsing the packet header of the detour data, the switch obtains the path group by parsing the packet header.

[0090] Exemplarily, after receiving the detour packet, the upstream aggregation switch does not distinguish whether the destination is the same aggregation switch group, and uniformly forwards it through the internal interconnected port to the downstream aggregation switch. Specifically, the upstream aggregation switch combines the next hops corresponding to the active internal interconnected ports (ports docking the downstream aggregation switches within this group) into an equivalent path group, that is, the common equivalent path group. The common equivalent path group is the path group hit by the detour packet received by the first upstream aggregation switch in the detour aggregation switch group. The common equivalent path group is used to indicate the equivalent path for forwarding the packet to each downstream aggregation switch. The common equivalent path group can be simply referred to as the detour common group, which matches the detour packet received from the external interconnected port, that is, the inter-group interconnected port (port docking the external aggregation switch group), and redirects it to the detour common group. The packet is hashed within this group for load sharing and then forwarded to each downstream aggregation switch, that is, the first downstream network device.

[0091] In an optional implementation manner, when the detour packet includes a first packet and a second packet, and the first packet and the second packet respectively correspond to different destination network device groups, the number of common equivalent path groups is at least one.

[0092] Exemplarily, still referring to the above Figure 3 , a aggregation switch group composed of 32 upstream aggregation switches and 32 downstream aggregation switches. The packet enters this group through the upstream aggregation switch 1 of this aggregation switch group, and is detoured and forwarded to the destination aggregation switch group. There are two direct connection paths between this aggregation switch group and the destination aggregation switch group. The paths corresponding to the out ports of this aggregation switch group are port 1 of the upstream aggregation switch 1 and port 2 of the upstream aggregation switch 2 respectively.

[0093] First, the first-hop network device of the detour packet arriving at the detour aggregation and switching unit must be the upper aggregation switch. The upper aggregation switch identifies the detour packet and sends it to the detour common group through redirection or other forwarding control means. If all the next-hop output ports in the detour common group are idle, uniform load sharing can be achieved. If the internal interconnection ports 33 to 64 of the upper aggregation switch 1 are all idle, at least 32 next hops can be used to achieve load balancing. Considering possible congestion and supporting consistent hashing, the number of occupied entries can be increased, such as occupying 128 entries, with 4 entries occupied by the next hop of each internal interconnection port. When a certain internal interconnection port is congested, some or all of the 4 entries occupied by this port can be replaced with other idle internal interconnection ports. An upper aggregation switch supports at least one detour common group, that is, the packets going to different destinations on this upper aggregation switch can share a detour common group. The upper aggregation switch 1 receives the detour packet and sends it to each lower aggregation switch through the detour common group.

[0094] In the embodiment of the present application, when the detour packet includes multiple packets going to different destination network device groups, the number of common equivalent path groups can be at least one; in this way, there is no need to distinguish the destination network device groups of different packets, and a common equivalent path group can be directly shared, which is more convenient and efficient.

[0095] S903. The first lower network device forwards the received detour packet to the second upper network device among the multiple upper network devices according to the overall equivalent path group. The overall equivalent path group is used to indicate the equivalent path formed by the second group of internal interconnection ports in each group of internal interconnection ports. The second group of internal interconnection ports satisfies the load balancing rule and corresponds to the second upper network device. The equivalent path between the second upper network device and the destination network device group satisfies the inter-group detour rule.

[0096] Exemplarily, the lower aggregation switch matches the detour packet and sends the detour packet to the upper aggregation switch indicated by the overall equivalent path group, that is, the second upper network device, through redirection or other forwarding control means. The output ports of the direct connection path between the local aggregation and switching unit and the destination aggregation and switching unit are port 1 of the upper aggregation switch 1 and port 2 of the upper aggregation switch 2. Each lower aggregation switch only needs to include the internal interconnection ports of the group connected to the upper aggregation switch 1 and the upper aggregation switch 2 in the next-hop output port in the overall equivalent path group.

[0097] Generally, the number of ports of each lower aggregation switch connected to each upper aggregation switch is uniformly consistent. Figure 3In the networking example, each lower aggregation switch is directly connected to an upper aggregation switch through one link. That is to say, the lower aggregation switch can adjust the weights of the next hops corresponding to the two egress ports within the overall equivalent path group according to the congestion conditions of port 1 of the upper aggregation switch 1 and port 2 of the upper aggregation switch 2 sensed normally, as well as the local congestion condition. Packets are load-balanced from the overall equivalent path group of the lower aggregation switch, and some packets are forwarded to the upper aggregation switch 1, and some packets are forwarded to the upper aggregation switch 2.

[0098] In an optional implementation manner, the inter-group detour rule includes: the transmission distance corresponding to the equivalent path between the second upper network device and the destination network device group meets the distance threshold.

[0099] Exemplarily, when the lower aggregation switch matches a detour packet, it assigns an equivalent path group to each packet going to each destination, that is, the overall equivalent path group. The next hops indicated in the group all correspond to the equivalent shortest paths to the destination aggregation switch group. Generally speaking, the next-hop switches indicated are all upper aggregation switches that have a direct connection path to the destination aggregation switch group. It can be understood that when there is a failure in the network device, the shortest path corresponding to the next hop indicated by the overall equivalent path group may not be a direct connection path, but a path that detours to avoid the shortest route corresponding to the failure, that is, it meets the distance threshold.

[0100] In the embodiment of the present application, the inter-group detour rule includes that the equivalent path between the second upper network device and the destination network device group is a path whose transmission distance meets the distance threshold, so as to ensure that the inter-group hop count of the packet in the network is within the expected hop count and meet the efficiency requirement of packet transmission.

[0101] S904, the second upper network device forwards the received detour packet to the destination network device group according to the direct connection equivalent path group, and the direct connection equivalent path group is used to indicate the equivalent path formed by the first group of external interconnection ports that meet the inter-group detour rule among the external interconnection ports of each group.

[0102] Exemplarily, still taking Figure 3 shown as an example, the upper aggregation switches 1 and 2 match the detour packets entering the device from the internal interconnection port, and send them to the direct connection equivalent path group through redirection or other forwarding control means. In the upper aggregation switch 1, there is only 1 direct connection path egress port, which is port 1. In the upper aggregation switch 2, there is only 1 direct connection path egress port, which is port 2. The upper aggregation switches 1 and 2 can respectively occupy the least number of equivalent path entries to forward the packets from the direct connection equivalent path group out of the local aggregation switch group.

[0103] It can be seen that in the above forwarding load sharing process, the next-hop egress ports that need to be placed in an equivalent path group all belong to the same type, either all external interconnection ports or all internal interconnection ports. There is no situation where a small number of external interconnection ports are mixed with a large number of internal interconnection ports for load sharing, reducing the waste of the number of next-hop entries.

[0104] In an alternative embodiment, when the loop network device group is divided into a federated routing autonomous system, each upper network device and each lower network device respectively correspond to a single routing autonomous system identifier under the federated routing autonomous system, and each equivalent path indicated by the directly connected equivalent path group satisfies one or more of the following conditions: The routing autonomous system identifier corresponding to any one equivalent path indicated by the directly connected equivalent path group includes at least one, and any one of the at least one is different from the single routing autonomous system identifier under the federated routing autonomous system to which the local machine belongs; The total number of routing autonomous system identifiers corresponding to the equivalent paths indicated by the directly connected equivalent path group is less than a quantity threshold; the quantity threshold is a positive integer greater than or equal to 2, and the specific value can be set according to actual needs, and is not limited in this application example.

[0105] Exemplarily, each device publishes and obtains routing information according to a general routing protocol, such as BGP. Assuming that a federated AS is allocated based on an aggregation switch group, and each aggregation switch is allocated a local AS in the federated AS. The usual path characteristics in the directly connected equivalent path group are that it contains 0 local ASNs and at least one, for example, 1 federated ASN. That is to say, the routing autonomous system identifiers corresponding to the equivalent paths indicated by the directly connected equivalent path group do not include the single routing autonomous system identifier under the federated routing autonomous system to which the local belongs.

[0106] In the embodiment of the present application, when the loop network device group is divided into a federated routing autonomous system, each upper network device and each lower network device respectively correspond to a single routing autonomous system identifier under the federated routing autonomous system: The routing autonomous system identifier corresponding to each equivalent path indicated by the directly connected equivalent path group includes at least one, and any one of the at least one is different from the single routing autonomous system identifier under the federated routing autonomous system to which the local machine belongs, which can ensure that in the directly connected equivalent path group, if the same loop network device group is used as a routing autonomous system, when the detour packets reach the second upper network device in each path indicated by the directly connected equivalent path group, they will no longer detour within the loop network device group, but leave the loop network device group from the second upper network device, avoiding the transmission path of the detour packets forming a loop within the group and further improving the packet transmission efficiency; and / or, The total number of routing autonomous system identifiers corresponding to each equivalent path indicated by the directly-connected equivalent path group is less than the quantity threshold. For example, the total number is relatively the smallest (such as 1), so as to control the number of network device groups that the detour message needs to pass through when going from the second upstream network device to the destination network device group, and avoid the loop formation of the detour message transmission path outside the group, further improving the message transmission efficiency.

[0107] In addition, the embodiments of the present application do not limit the load balancing rules adopted by the network device to determine each equivalent path group. For example, the network device can refine the grouping under the public equivalent path group, the overall equivalent path group, and the directly-connected equivalent path group according to characteristics such as the service message type, so as to achieve more targeted load balancing. For example, for the public equivalent path group, the network device can further divide it into path subgroups such as a delay-sensitive group and a bandwidth-sensitive group, and thus adjust the weight of each next hop by adjusting the number of equivalent path entries occupied by the next hop in each path subgroup according to the load congestion situation of the path.

[0108] For the sake of easy understanding, the following combines Figure 10 to compare and illustrate the message transmission method provided by the embodiments of the present application with the message transmission path of the related art.

[0109] Exemplarily, Figure 10 is a comparison diagram of the example of the message transmission path under the message transmission method provided by the embodiments of the present application and the example of the message transmission path in the related art. As Figure 10 shown, the effect that the detour message hopes to achieve through the aggregation switch group is that the detour messages going to the same destination group can be load-shared to the ports docking with each destination group for forwarding as expected. For example, in the related art, each equivalent path group corresponding to the transmission path P4 and the transmission path P5 includes different port types: intra-group interconnection ports and extra-group interconnection ports, while the port type of the equivalent path group corresponding to the transmission path P6 of the embodiments of the present application is intra-group interconnection ports, and the port type of the equivalent path group corresponding to the transmission path P7 before sending out through the expected port is intra-group interconnection ports.

[0110] It can be seen that the embodiments of the present application use time to exchange for space. Although the first hop of the detour message entering the group on the aggregation device does not directly forward out from the expected target port, through the overall scheduling of the lower aggregation switch, the detour message returns to the upper aggregation device after being forwarded by the lower aggregation and is forwarded out through the expected port, avoiding the load sharing of the next hop corresponding to the internal interconnection port and the next hop corresponding to the external interconnection port in the upper aggregation switch in an equivalent path group, saving a large amount of equivalent path entry resources, and being beneficial to supporting a larger networking scale under limited chip capabilities.

[0111] That is to say, instead of being eager to perform load sharing by putting together all equivalent paths with different types of next hops on a single aggregation switch, hierarchical load sharing is adopted: when the detour packet arrives at the first-hop device (the upper aggregation switch) of the detour aggregation switch group, load sharing is only performed among the next-hop members corresponding to each internal interconnection port; when it arrives at the second-hop device (the lower aggregation switch) of the detour aggregation switch group according to unified scheduling, load sharing is only performed among the next-hop members corresponding to the relevant internal interconnection ports; when it arrives at the third-hop device (the upper aggregation switch) of the detour aggregation switch group, load sharing is only performed among the next hops corresponding to the relevant external interconnection ports.

[0112] The present application also provides a network device. Exemplarily, Figure 11 is a schematic diagram of the framework structure of the network device provided by an embodiment of the present application. As Figure 11 shown, the network device 100 includes: a bus 102, a processor 104, a memory 106, and a communication interface 108. The processor 104, the memory 106, and the communication interface 108 communicate with each other through the bus 102. The network device 100 may be the upper network device or the lower network device (such as a switch, a router, etc.) in the above embodiments. It should be understood that the present application does not limit the number of processors and memories in the network device 100.

[0113] The bus 102 may be a peripheral component interconnect (PCI) bus, an extended industry standard architecture (EISA) bus, or the like. The bus may be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, Figure 3 only one line is shown in [figure], but it does not mean that there is only one bus or one type of bus. The bus 102 may include a path for transmitting information among various components (such as the memory 106, the processor 104, and the communication interface 108) of the network device 100.

[0114] The processor 104 may include any one or more of a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP), etc.

[0115] The memory 106 may include volatile memory, such as random access memory (RAM). The processor 104 may also include non-volatile memory, such as read-only memory (ROM), flash memory, a hard disk drive (HDD), or a solid state drive (SSD).

[0116] Exemplarily, as Figure 11 shown, executable program code is stored in the memory 106, and the processor 104 executes the executable program code to respectively implement the functions of the foregoing first upper network device, first lower network device, or second upper network device, thereby implementing the message transmission method. In one example, the processor 104 may also execute the executable program code to respectively implement the functions of the foregoing first upper network device, first lower network device, or second upper network device. That is, instructions for executing the message transmission method are stored on the memory 106.

[0117] The communication interface 108 uses a transceiver module such as, but not limited to, a network interface card or a transceiver to implement communication between the network device 100 and other devices or a communication network.

[0118] Exemplarily, an embodiment of the present application further provides a message transmission system, which includes a rotating network device group, a source network device group, and a destination network device group. The rotating network device group includes a plurality of upper network devices docked with devices outside the rotating network device group and a plurality of lower network devices docked with computing devices in the network. Each upper network device and each lower network device are communicatively connected through an intra-group interconnection port. The destination network device group and each upper network device are communicatively connected through an inter-group interconnection port; A first upper network device, configured to receive a detour message, and forward the detour message to a first lower network device among a plurality of lower network devices according to a common equivalent path group. The first upper network device includes any one of the plurality of upper network devices. The detour message includes a message from the source network device group and to be forwarded by the rotating network device group to the destination network device. The common equivalent path group is used to indicate an equivalent path formed by a first group of intra-group interconnection ports that meet the load balancing rule among the intra-group interconnection ports. The first lower network device corresponds to the first group of intra-group interconnection ports; A first lower network device is configured to forward the received detour message to a second upper network device among multiple upper network devices according to a coordinated equivalent path group. The coordinated equivalent path group is used to indicate an equivalent path formed by second intra-group interconnected ports in each group of intra-group interconnected ports. The second intra-group interconnected ports satisfy the load balancing rule and correspond to the second upper network device. The equivalent path between the second upper network device and the destination network device group satisfies the inter-group detour rule. The second upper network device is configured to forward the received detour message to the destination network device group according to a direct-connected equivalent path group. The direct-connected equivalent path group is used to indicate an equivalent path formed by first extra-group interconnected ports that satisfy the inter-group detour rule among each group of extra-group interconnected ports.

[0119] Exemplarily, the first upper network device, the first lower network device, and the second upper network device in the above message transmission system may respectively be, for example, the structures shown above Figure 11 as shown. An embodiment of the present application further provides a computing device cluster. The computing device cluster includes at least one computing device. Each computing device is connected to at least one lower network device. The lower network devices belong to a detour network device group, and the detour network device group further includes multiple upper network devices. Each upper network device and each lower network device are communicatively connected through intra-group interconnected ports. The upper network device is connected to a device outside the detour network device group. Each lower network device and each upper network device respectively include a processor and a memory. The processor of the lower network device is configured to execute instructions stored in the memory of the lower network device, and the processor of the upper network device is configured to execute instructions stored in the memory of the upper network device, so that the computing device cluster executes the message transmission method provided in any of the above embodiments.

[0120] Among them, the computing device may be directly connected to at least one lower network device, or may be connected to at least one lower network device through a top of rack (TOR) switch. The upper network device and a device outside the detour network device group may be directly connected; alternatively, the upper network device may be connected to an OXC or an OCS, and different detour network device groups are interconnected through an OXC or an OCS.

[0121] Exemplarily, Figure 12 is a schematic diagram of the framework structure of the computing device cluster provided by an embodiment of the present application. As Figure 12 shown, the computing device cluster includes at least one computing device and at least two network device groups. Instructions for executing the message transmission method may be stored in the memory 106 of one or more network devices in the computing device cluster.

[0122] In some possible implementations, part of the instructions for executing the message transmission method may also be stored separately in the memory 106 of one or more network devices in the computing device cluster. In other words, a combination of one or more computing devices can jointly execute the instructions for executing the message transmission method.

[0123] In one example, the structure of the above message transmission system may refer to the structure of the computing device cluster shown above Figure 12 shown in the computing device cluster.

[0124] The embodiments of the present application also provide a computer program product containing instructions. The computer program product may be software or a program product containing instructions that can run on a computing device or be stored in any available medium. When the computer program product runs on the computing device cluster, it causes the computing device cluster to execute the above message transmission method.

[0125] The embodiments of the present application also provide a computer-readable storage medium. The computer-readable storage medium may be any available medium that a computing device can store or a data storage device such as a data center containing one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc. The computer-readable storage medium includes instructions that direct the computing device cluster to execute the message transmission method.

[0126] This embodiment also provides a computer program product that, when run on a computer, causes the computer to execute the above-related steps to implement the message transmission method in the above embodiment.

[0127] Among them, the network device, computer storage medium, or computer program product provided in this embodiment are all used to execute the corresponding method provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method provided above, and will not be elaborated here.

[0128] Any content of each embodiment of the present application, as well as any content of the same embodiment, can be freely combined. Any combination of the above content is within the scope of the present application.

[0129] Those skilled in the art should be able to realize that in one or more of the above examples, the functions described in the embodiments of the present application can be implemented by hardware, software, firmware, or any combination thereof. When implemented using software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. The computer-readable medium includes computer storage media and communication media, where the communication media includes any medium that facilitates the transfer of a computer program from one place to another. The storage media can be any available medium accessible by a general-purpose or special-purpose computer.

[0130] The embodiments of the present application have been described above in conjunction with the accompanying drawings. However, the present application is not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative and not restrictive. Under the inspiration of the present application, those of ordinary skill in the art can also make many forms without departing from the purpose of the present application and the scope protected by the claims, and all of them fall within the protection scope of the present application.

Claims

1. A message transmission method, characterized in that: A circumventing network device group applied to a network, the circumventing network device group includes a plurality of upper network devices connected to devices outside the circumventing network device group, and a plurality of lower network devices connected to computing devices in the network, each of the plurality of upper network devices is communicatively connected to each of the plurality of lower network devices via an intra-group interconnection port, the network further includes a source network device group and a destination network device group, the destination network device group is communicatively connected to each of the upper network devices via an extra-group interconnection port, the method includes: A first upper network device receives a detour message, wherein the first upper network device includes any upper network device among the plurality of upper network devices, and the detour message includes a message from the source network device group and to be forwarded by the detour network device group to the destination network device; The first upper network device forwards the detour message to a first lower network device among the multiple lower network devices according to a common equal-cost path group, wherein the common equal-cost path group is used to indicate an equal-cost path formed by a first intra-group interconnection port among the intra-group interconnection ports that meets a load balancing rule, and the first lower network device corresponds to the first intra-group interconnection port; The first lower network device forwards the received detour message to a second upper network device among the multiple upper network devices according to a coordinated equal-cost path group, wherein the coordinated equal-cost path group is used to indicate an equal-cost path formed by a second intra-group interconnection port among the intra-group interconnection ports, wherein the second intra-group interconnection port satisfies a load balancing rule and corresponds to the second upper network device, and the equal-cost path between the second upper network device and the destination network device group satisfies an inter-group detour rule; The second upper network device forwards the received detour message to the destination network device group according to a direct equal-cost path group, wherein the direct equal-cost path group is used to indicate an equal-cost path formed by the first group of external interconnected ports among each group of external interconnected ports that meets the inter-group detour rule.

2. The method according to claim 1, characterized in that In the case where the detour network device group is divided into a federal routing autonomous system, each of the upper network devices and each of the lower network devices respectively corresponds to a single routing autonomous system identifier under the federal routing autonomous system, and each equal-cost path indicated by the direct equal-cost path group satisfies one or more of the following conditions: The routing autonomous system identifier corresponding to any equal-cost path indicated by the directly connected equal-cost path group includes at least one, and any one of the at least one is different from a single routing autonomous system identifier under the federal routing autonomous system to which the local device belongs; The total number of routing autonomous system identifiers corresponding to the equal-cost paths indicated by the directly connected equal-cost path group is less than the number threshold.

3. The method according to claim 1 or 2, characterized in that: The inter-group detour rule includes: a transmission distance corresponding to an equal-cost path between the second upper network device and the destination network device group satisfies a distance threshold.

4. The method according to claim 1 or 2, characterized in that: When the detour message includes a first message and a second message, and the first message and the second message correspond to different destination network device groups respectively, the number of the public equal-cost path groups is at least one.

5. The method according to claim 1 or 2, characterized in that: The detour message carries a detour tag, and the detour tag is used by the first upper network device to identify the detour message in the received message; The timing for adding the detour mark in the detour message includes: before the source network device group sends the detour message to the detour network device group, or before the detour network device group receives the detour message and sends the detour message to the first downstream network device.

6. A message transmission method, characterized in that: Applied to an upper network device, the upper network device is deployed in a circumventing network device group in a network and is connected to a device outside the circumventing network device group, the circumventing network device group includes a plurality of upper network devices and a plurality of lower network devices connected to a computing device in the network, each of the plurality of upper network devices is communicatively connected to each of the plurality of lower network devices via an intra-group interconnection port, the network further includes a source network device group and a destination network device group, the destination network device group is communicatively connected to each of the upper network devices via an extra-group interconnection port, the method includes: receiving a detour message, wherein the detour message includes a message from the source network device group and to be forwarded by the detour network device group to the destination network device; Forwarding the detour message to a first lower network device among the multiple lower network devices according to a common equal-cost path group, wherein the common equal-cost path group is used to indicate an equal-cost path formed by a first intra-group interconnection port among the intra-group interconnection ports that meets a load balancing rule, and the first lower network device corresponds to the first intra-group interconnection port; In the case where the equal-cost path between the first lower network device and the destination network device group satisfies the inter-group detour rule, receiving the detour message forwarded by the first lower network device according to the coordinated equal-cost path group, the coordinated equal-cost path group is used to indicate the equal-cost path formed by the second intra-group interconnection port among the intra-group interconnection ports, the second intra-group interconnection port satisfies the load balancing rule and corresponds to the upper network device that receives the detour message forwarded by the first lower network device; The detour message received from the first downstream network device is forwarded to the destination network device group according to the direct equivalent-cost path group, wherein the direct equivalent-cost path group is used to indicate an equivalent-cost path formed by the first group of external interconnection ports among the external interconnection ports of each group that meets the inter-group detour rule.

7. The method according to claim 6, characterized in that In the case where the detour network device group is divided into a federal routing autonomous system, each of the upper network devices and each of the lower network devices respectively corresponds to a single routing autonomous system identifier under the federal routing autonomous system, and each equal-cost path indicated by the direct equal-cost path group satisfies one or more of the following conditions: The routing autonomous system identifier corresponding to any equal-cost path indicated by the directly connected equal-cost path group includes at least one, and any one of the at least one is different from a single routing autonomous system identifier under the federal routing autonomous system to which the local device belongs; The total number of routing autonomous system identifiers corresponding to the equal-cost paths indicated by the directly connected equal-cost path group is less than the number threshold.

8. The method according to claim 6 or 7, characterized in that: The inter-group detour rule includes: a transmission distance corresponding to an equal-cost path between the upper network device and the destination network device group satisfies a distance threshold.

9. The method according to claim 6 or 7, characterized in that: When the detour message includes a first message and a second message, and the first message and the second message correspond to different destination network device groups respectively, the number of the public equal-cost path groups is at least one.

10. The method according to claim 6 or 7, characterized in that: The detour message carries a detour tag, and the detour tag is used by the first upper network device to identify the detour message in the received message; The timing for adding the detour mark in the detour message includes: before the source network device group sends the detour message to the detour network device group, or before the detour network device group receives the detour message and sends the detour message to the first downstream network device.

11. A message transmission method, characterized in that: Applied to a lower network device, the lower network device is deployed in a circumventing network device group in a network and is connected to a computing device in the network, the circumventing network device group includes a plurality of lower network devices, and a plurality of upper network devices connected to devices outside the circumventing network device group, each of the plurality of upper network devices is communicatively connected to each of the plurality of lower network devices via an intra-group interconnection port, the network further includes a source network device group and a destination network device group, the destination network device group is communicatively connected to each of the upper network devices via an extra-group interconnection port, the method includes: forwarding the detour message received from the first upper network device to a second upper network device among the plurality of upper network devices according to the coordinated equal-cost path group, wherein the coordinated equal-cost path group is used to indicate an equal-cost path formed by a second group of interconnected ports among the interconnected ports within each group, wherein the second group of interconnected ports satisfies a load balancing rule and corresponds to the second upper network device; The first upper network device includes any upper network device among the plurality of upper network devices, and the detour message includes a message from the source network device group to be forwarded by the detour network device group to the destination network device; The equal-cost path between the second upper network device and the destination network device group satisfies the inter-group detour rule and is used to forward the received detour message to the destination network device group according to the direct-connected equal-cost path group, and the direct-connected equal-cost path group is used to indicate the equal-cost path formed by the first group of external interconnection ports among each group of external interconnection ports that satisfies the inter-group detour rule.

12. The method according to claim 11, characterized in that In the case where the detour network device group is divided into a federal routing autonomous system, each of the upper network devices and each of the lower network devices respectively corresponds to a single routing autonomous system identifier under the federal routing autonomous system, and each equal-cost path indicated by the direct equal-cost path group satisfies one or more of the following conditions: The routing autonomous system identifier corresponding to any equal-cost path indicated by the directly connected equal-cost path group includes at least one, and any one of the at least one is different from a single routing autonomous system identifier under the federal routing autonomous system to which the local device belongs; The total number of routing autonomous system identifiers corresponding to the equal-cost paths indicated by the directly connected equal-cost path group is less than the number threshold.

13. The method according to claim 11 or 12, characterized in that: The inter-group detour rule includes: a transmission distance corresponding to an equal-cost path between the upper network device and the destination network device group satisfies a distance threshold.

14. A message transmission system, characterized in that: The system comprises a circumvention network device group, a source network device group and a destination network device group, wherein the circumvention network device group comprises a plurality of upper network devices connected to devices outside the circumvention network device group, and a plurality of lower network devices connected to computing devices in the network, each of the plurality of upper network devices is communicatively connected to each of the plurality of lower network devices via an intra-group interconnection port, and the destination network device group is communicatively connected to each of the upper network devices via an extra-group interconnection port; A first upper network device, configured to receive a detour message, and forward the detour message to a first lower network device among the multiple lower network devices according to a public equivalent-cost path group, wherein the first upper network device includes any upper network device among the multiple upper network devices, the detour message includes a message from the source network device group to be forwarded by the detour network device group to the destination network device, the public equivalent-cost path group is used to indicate an equivalent path formed by a first intra-group interconnection port among the intra-group interconnection ports that meets a load balancing rule, and the first lower network device corresponds to the first intra-group interconnection port; The first lower network device is used to forward the received detour message to a second upper network device among the multiple upper network devices according to a coordinated equal-cost path group, the coordinated equal-cost path group is used to indicate an equal-cost path formed by a second intra-group interconnection port among the intra-group interconnection ports, the second intra-group interconnection port satisfies a load balancing rule and corresponds to the second upper network device, and the equal-cost path between the second upper network device and the destination network device group satisfies an inter-group detour rule; The second upper network device is used to forward the received detour message to the destination network device group according to the direct equal-cost path group, and the direct equal-cost path group is used to indicate the equivalent path formed by the first group of external interconnection ports among each group of external interconnection ports that meets the inter-group detour rule.

15. The system according to claim 14, characterized in that In the case where the detour network device group is divided into a federal routing autonomous system, each of the upper network devices and each of the lower network devices respectively corresponds to a single routing autonomous system identifier under the federal routing autonomous system, and each equal-cost path indicated by the direct equal-cost path group satisfies one or more of the following conditions: The routing autonomous system identifier corresponding to any equal-cost path indicated by the directly connected equal-cost path group includes at least one, and any one of the at least one is different from a single routing autonomous system identifier under the federal routing autonomous system to which the local device belongs; The total number of routing autonomous system identifiers corresponding to the equal-cost paths indicated by the directly connected equal-cost path group is less than the number threshold.

16. The system according to claim 14 or 15, characterized in that The inter-group detour rule includes: a transmission distance corresponding to an equal-cost path between the second upper network device and the destination network device group satisfies a distance threshold.

17. The system according to claim 14 or 15, characterized in that When the detour message includes a first message and a second message, and the first message and the second message correspond to different destination network device groups respectively, the number of the public equal-cost path groups is at least one.

18. The system according to claim 14 or 15, characterized in that The detour message carries a detour tag, and the detour tag is used by the first upper network device to identify the detour message in the received message; The timing for adding the detour mark in the detour message includes: before the source network device group sends the detour message to the detour network device group, or before the detour network device group receives the detour message and sends the detour message to the first downstream network device.

19. A network device, characterized in that: The device comprises: a processor and a memory; The memory is used to store instructions; The processor is configured to execute instructions stored in the memory so that the network device executes the method according to any one of claims 1 to 13.

20. A computing device cluster, characterized in that: The method comprises at least one computing device, each computing device is connected to at least one lower network device, the lower network device belongs to a circumvention network device group, the circumvention network device group also comprises a plurality of upper network devices, each upper network device is communicatively connected to each lower network device via an intra-group interconnection port, the upper network device is connected to a device outside the circumvention network device group, and each lower network device and each upper network device respectively comprises a processor and a memory; The processor of the lower network device is used to execute instructions stored in the memory of the lower network device, and the processor of the upper network device is used to execute instructions stored in the memory of the upper network device, so that the computing device cluster executes the method according to any one of claims 1 to 13.

21. A computer program product comprising instructions, characterized in that When the instruction is executed by the computing device cluster, the network device for transmitting messages within the computing device cluster executes the method according to any one of claims 1 to 13.

22. A computer program product, characterized in that The method comprises a computer program, which, when executed by a network device, enables the network device to perform the method according to any one of claims 1 to 13.

Images