WO2022083503A1 - 数据处理方法及装置 - Google Patents

数据处理方法及装置 Download PDF

Info

Publication number
WO2022083503A1
WO2022083503A1 PCT/CN2021/123930 CN2021123930W WO2022083503A1 WO 2022083503 A1 WO2022083503 A1 WO 2022083503A1 CN 2021123930 W CN2021123930 W CN 2021123930W WO 2022083503 A1 WO2022083503 A1 WO 2022083503A1
Authority
WO
WIPO (PCT)
Prior art keywords
network node
computing task
routing table
node
received
Prior art date
Application number
PCT/CN2021/123930
Other languages
English (en)
French (fr)
Inventor
林钦亮
田阳
廖怡
游俊
熊艳辉
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Publication of WO2022083503A1 publication Critical patent/WO2022083503A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00—Routing or path finding of packets in data switching networks
    • H04L45/74—Address processing for routing
    • H04L45/745—Address table lookup; Address filtering
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00—Routing or path finding of packets in data switching networks
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00—Routing or path finding of packets in data switching networks
    • H04L45/28—Routing or path finding of packets in data switching networks using route fault recovery
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00—Routing or path finding of packets in data switching networks
    • H04L45/54—Organization of routing tables

Definitions

  • the present application relates to the field of in-network computing, in particular to a data processing method.
  • routers and switches may also have service packet processing capabilities.
  • Some computing tasks suitable for processing by routers and switches can be offloaded to network nodes such as routers and switches for execution, thereby reducing the pressure on computing nodes.
  • the service packets belonging to the same computing task need to be allocated to the same network node for execution.
  • the calculation task ID carried in the service packet is changed to the number of network nodes in the network. Hash calculation is performed for the length to calculate the network node processing the computing task, and recording the network node processing the computing task in the routing table in the network node receiving the computing task.
  • the computing task when the number of network nodes in the network changes, such as a network node failure or a new network node is added, due to the change in the hash length, when the computing task has been received before receiving it, it is determined by the hash calculation.
  • the network node processing the computing task is different from the network node processing the computing task calculated before the network node changes.
  • the sending of computing tasks is stopped, and the sending of computing tasks is restarted after the routing tables in all network nodes are synchronized.
  • the embodiment of the present application aims to provide a data processing method, so that when the structure of the distributed network changes, the execution of the computing task does not need to be interrupted.
  • the first aspect of the embodiments of the present invention provides a data processing method, which is executed by a main network node, the main network node is used to connect a first network node and a second network node, and the second network node
  • the method includes: when the second network node is detected, sending a notification message to the first network node; receiving the first network node sent by the first network node. A computing task that a network node has never received before receiving the notification message; processing the computing task sent by the first network node.
  • the method further includes: receiving confirmation information sent by the first network node, where the confirmation information is used to indicate the the first network node has received the notification information; send a global routing table to the first network node, the global routing table includes the first computing task received by the main network node and processing the first A network node for computing tasks, and a second computing task received by the first network node before receiving the notification message, and a network node for processing the second computing task.
  • the method further includes: after sending a notification message to the first network node, receiving a node routing table in the first network node sent by the first network node, the first network node A node routing table in a network node includes a second computing task received by the first network node before receiving the notification message and a network node that processes the second computing task; the first network node The node routing table in the main network node and the node routing table in the main network node are merged into the global routing table, and the routing table of the main network node includes the first computing task received by the main network node and the processing of the first computing task received by the main network node.
  • a network node for computing tasks is merged into the global routing table, and the routing table of the main network node includes the first computing task received by the main network node and the processing of the first computing task received by the main network node.
  • the method further includes: during normal operation, when the first network node receives the first message of the computing task, updating the node routing table according to the message, and updating the The content of the node routing table update is sent to the main network node.
  • the pressure of the master node to synchronize the node routing tables of all nodes during the routing table synchronization period is avoided.
  • the method further includes: storing the computing task and the network node processing the computing task in the host computer The routing table of the network node.
  • the main network node saves the network node information of the computing task in the routing table of the main network node in time to ensure the accuracy of the network-wide routing table generated according to the routing table of the main network node.
  • the method further includes: sending the confirmation information to the first network node.
  • the node sends a hash routing table, where the hash routing table is used to calculate and process the computing task with the number of network nodes in the network system where the main network node and the first network node are located as the hash length.
  • the notification message is sent to the first network node, if the confirmation message from the first network node is not received after a first time threshold is exceeded, the notification message is re-sent to the first network node .
  • each first network node obtains the notification message of the network structure change.
  • a second aspect of the present invention provides a data processing method, which is performed by a first network node, the first network node is connected to a main network node, the method includes: receiving a notification message sent by the main network node, the The notification message is used to notify the primary network node to detect a second network node, and the second network node is a newly added network node or a faulty network node; to receive the first computing task; after determining that the received computing task is in the When a computing task that has never been received before the notification message is received, the computing task is sent to the main network node.
  • the first network node includes a node routing table, and the node routing table records the computing tasks and processing tasks received by the first network node before receiving the notification message.
  • the network node that describes the computing task.
  • the method further includes: when it is determined that the received computing task is a computing task received before the notification message is received, according to a node routing table in the first network node The computing task is forwarded to the network node corresponding to the computing task.
  • the original service is continued to be processed at the first network node itself, so as to reduce the load of the main network node.
  • the method includes: after receiving the notification message, sending the node routing table to the primary network node.
  • the method further includes: during normal operation, when the first network node receives the first message of the computing task, updating the node routing table according to the message, and updating the The content of the node routing table update is sent to the main network node.
  • the method includes: receiving a global routing table sent by the primary network node, where the global routing table includes a first computing task received by the primary network node and processing the first computing task received by the primary network node.
  • the method includes: receiving a global routing table sent by the primary network node, where the global routing table includes a first computing task received by the primary network node and processing the first computing task received by the primary network node.
  • the task's network node is not a calculation task in the global routing table, the second calculation is processed with the number of network nodes in the network system where the main network node and the first network node are located as the hash length. The task's network node.
  • a third aspect of the present invention provides a data processing apparatus for performing the method described in the first aspect.
  • a fourth aspect of the present invention provides a data processing apparatus for performing the method described in the second aspect.
  • a fifth aspect of the present invention provides a network node system, the network node system includes a main network node and a first network node, and is characterized in that:
  • the primary network node is configured to perform the method described in the first aspect
  • the first network node is configured to perform the method of the second aspect.
  • a fifth aspect of the present invention provides a network node, characterized in that the node includes: a processor and a memory, wherein program instructions are stored in the memory, and the processor executes the program instructions to implement the first aspect or The method described in the second aspect.
  • a sixth aspect of the present invention provides a computer-readable storage medium, on which a computer program or instruction is stored, and when the computer program or instruction is executed in a computer, causes the computer to execute the first or The method described in the second aspect.
  • a seventh aspect of the present invention provides a computer program product comprising instructions which, when run on a computer, cause the computer to perform the method of the first or second aspect above.
  • Fig. 1 shows the structural schematic diagram of the distributed network of the present invention
  • Fig. 2 shows the structure schematic diagram of the network node in the distributed network of the present invention
  • Fig. 3 shows the flow chart of network node service message processing in the present invention
  • Fig. 4 shows the flow chart of routing table synchronization after a network node structure of the present invention is changed
  • Fig. 5a shows the situation of removing a network node in Embodiment 1 of the present invention
  • Fig. 5b shows the situation of adding network nodes in Embodiment 1 of the present invention
  • Fig. 6 shows the flow chart of the message processing in Fig. 4 of the present invention
  • Fig. 7 shows the flow chart of routing table synchronization after another distributed network structure of the present invention is changed
  • Fig. 8 shows the structural block diagram of the data processing apparatus of the present invention
  • FIG. 9 is a block diagram showing the structure of another data processing apparatus of the present invention.
  • the distributed network 100 includes a plurality of network nodes and computing nodes. Four network nodes and four computing nodes are exemplarily taken as an example. Node 102 , network node 103 , network node 104 and computing node 105 , computing node 106 , computing node 107 , computing node 108 . The network node is connected with the corresponding computing node.
  • the network node 101 is connected with the computing node 105
  • the network node 102 is connected with the computing node 106
  • the network node 103 is connected with the computing node 107
  • the network node 104 is connected with the computing node 108;
  • the network nodes in the distributed network 100 are fully interconnected.
  • the computing nodes are connected through network nodes, and the computing nodes are used for data processing and data transmission is performed through the network nodes.
  • the computing node is a terminal device with data processing capability in the network, such as a host or a storage device.
  • Network nodes are used to forward data from computing devices, such as network nodes routers or switches.
  • the current network nodes can also perform some specific calculations. In this way, when the service packets transmitted between computing nodes are transmitted through the network nodes, the network nodes can perform certain calculations.
  • the transmitted service packets are processed in a specific manner, thereby reducing the load of the computing nodes, which is the in-network computing technology.
  • in-network computing technology some computing tasks suitable for processing by network nodes can be offloaded to network nodes for execution.
  • the service packets belonging to the same computing task need to be allocated to the same network node for execution. After the network node receives the service packet, it will identify the computing task carried in the service packet with the number of network nodes in the network.
  • Hash calculation is performed for the length to calculate the network node processing the computing task, and recording the network node processing the computing task in the routing table in the network node receiving the computing task, and then forwarding the service message
  • the network node that processes the service packet can be determined by directly querying the routing table according to the calculation task ID when the service packet of the same computing task is received next time.
  • FIG. 2 is a structural diagram of the network node 200 with computing capability.
  • the network node 200 includes a processor 201 connected to one or more data storage units.
  • the data storage unit may include a storage medium 203 and a memory unit (cache) 204 .
  • the storage medium 203 may be read-only, such as read only memory (ROM), or readable and writable, such as a hard disk or flash memory.
  • the memory unit 204 may be integrated with the processor 201 or may be a separate component.
  • the processor 201 is the control center of the network node 200, and is used for executing program codes to realize the functions corresponding to the program instructions.
  • the processor 201 includes one or more network processor units (Network Processor Unit, NPU) for processing service packets, for example, NPU0 and NPU1 as shown in FIG. 2 .
  • NPU Network Processor Unit
  • the network node 200 includes more than one processor, eg, processors 201 and 205 as shown in FIG. 2 .
  • processor refers to one or more devices, circuits, and/or processing cores for processing data such as computer program instructions.
  • the network processor may be a central processing unit (CPU).
  • the program codes executed by the CPUs of the processors 201 and/or 205 are stored in the memory unit 204 or the storage medium 203 .
  • the program code stored in the storage medium 203 may be copied into the memory unit 204 for execution by the processor.
  • the received data message is processed in the memory unit 204 according to the program code.
  • the processor can execute at least one kernel (eg, the kernel in operating systems sold under the brands LINUZ TM , UNIX TM , WINDOWS TM, ANDROID TM , IOS TM , etc.), which is known to control the execution of other programs, control and peripheral Communication of the devices and control of the use of network node resources controls the operation of the network node 200 .
  • the network node 200 also includes a communication interface 202 through which the network node 200 can communicate with another device or system, either directly or through an external network.
  • the above-mentioned elements of the network node 200 may be interconnected by any one or any combination of buses such as a data bus, an address bus, a control bus, an expansion bus, and a local bus.
  • the above-mentioned network node 200 may be a switch with a data processing function, a router with a data processing function, or any other device having a similar structure as shown in FIG. 2 for forwarding and processing service packets.
  • the service packet For a service packet that needs to be calculated in the network, the service packet will carry the identifier (ID) of the calculation task that needs to be performed by the network node.
  • ID the identifier
  • the administrator When the distributed network 100 supporting in-network computing is established, the administrator will configure two routing tables for each network node in the network.
  • the distributed network network 100 is working normally, after the service packet reaches any network node in the network 100, the address of the target network node for executing the calculation task is obtained according to the two routing tables of the ID of the calculation task carried in the service packet, And forward the service message to the target network node for processing.
  • the two routing tables are a node routing table and a hash routing table, respectively.
  • the node routing table includes the corresponding relationship between the calculation task identifier and the target network node, as shown in Table 1;
  • the hash routing table includes the corresponding relationship between the hash value and the target network node, as shown in Table 2. The method of how to determine the target node for performing the computing task carried in the service message according to the above two routing tables will be described with reference to the flowchart shown in FIG. 3 .
  • Compute task ID Network node IP address 00001 10.0.0.1 (node 101) 00002 10.0.0.2 (node 102)
  • Step S301 When the service packet arrives at the network node, determine whether the computing task identifier in the service packet exists in the node routing table. If it does not exist, go to step S302; if it exists, go to step S303.
  • the computing task identifier in the service packet is not included in the node routing table, it means that the network node has not received the service packet that needs to perform the computing task before. If the computing task identifier in the service packet is not If it is included in the node routing table, it means that the network node has received the service message that needs to perform the computing task before,
  • Step S302 perform hash calculation on the computing task identifier to obtain a hash value, obtain the address of the target network node corresponding to the hash value in the hash routing table according to the hash value, and compare the hash value and the corresponding target network node. The address is updated to the node routing table.
  • the hash length of the hash calculation is the number of network nodes in the distributed network.
  • the number of network nodes is the number of rows in the hash routing table of the network nodes.
  • Step S303 Acquire the address of the target network node corresponding to the computing task identifier, and determine whether the current network node is the target network node according to the address of the target network node. If yes, go to step S304; otherwise, go to step S305.
  • Matching is performed according to the IP address of the target network node and the IP address of the present network node, and if they are the same, the present network node is judged to be the target network node.
  • Step S304 the service packet is processed at the local network node.
  • Step S305 the network node forwards the service packet to the target network node, so that the target network node processes the service packet.
  • the hash length will be calculated according to the new hash length.
  • Hash value is used to determine the target network node that processes the service message.
  • the service message of this computing task may have been received by other network nodes before, and the hash value is calculated using the number of previous network nodes as the hash length.
  • the target network node which will lead to different target network nodes for service packets belonging to the same computing task before and after the number of network nodes changes, that is, the service packets belonging to the same computing task are processed by different network nodes. deal with.
  • the computing node stops sending service packets, and restarts sending computing tasks after the routing tables in all network nodes are synchronized.
  • the technical solution provided by the embodiments of the present invention can, after the number of network nodes in the distributed network changes, the slave network node forwards the received service packet of the computing task that has never been received before to the master network node for processing, thereby After the number of network nodes changes, the computing nodes stop sending service packets, thereby ensuring the continuous execution of computing tasks.
  • FIG. 4 is a schematic flowchart of the routing table synchronization in each network node after the number of network nodes changes in the embodiment of the present invention
  • FIGS. 5a and 5b are the removal of faulty network nodes or adding new ones in 100 in the distributed network.
  • FIG. 6 is a schematic diagram of a network node of the present invention
  • FIG. 6 is a processing flow of a received service packet in the process of synchronizing the routing table of each network node in the embodiment of the present invention.
  • the administrator selects a network node among the network nodes of the distributed network 100 as the master network node, and sends the address of the master network node to the slave network node.
  • the main network node is used to manage the entire routing table update process and data processing in the routing table update process. Through the main network node, it is ensured that the routing tables between the network nodes are consistent after the routing table of the network node is updated, and the data processing of the calculation in the entire network is guaranteed to be uninterrupted.
  • the present invention does not specify the process of selecting the main network node, and the election can be performed by any method or combination of methods in the prior art. For example, the main network node is elected by all network nodes, the administrator designates the main network node, and the network node with the largest bandwidth is selected according to the characteristics of the network nodes, such as the remaining bandwidth, to be the main network node.
  • the administrator can designate a slave network node as the backup network node of the master network node, and periodically back up the node routing table of the master network node to the backup node.
  • the backup network node detects the failure of the master network node, then the backup network node acts as the master network node and broadcasts the address of the node to other slave network nodes.
  • the main network node can be reset after the network is abnormal.
  • Step S401 the master network node detects that the number of network nodes changes, and sends a notification message to the slave network node.
  • the notification message is used to notify slave network nodes that the structure of the distributed network 100 has changed.
  • the master network node can perform heartbeat connection with all the slave network nodes in the network, and when the heartbeat connection of a slave network node is disconnected, it is judged that the network node is faulty; All network nodes and computing nodes broadcast the message, and after receiving the message, the master network node determines that a new node joins the network 100 .
  • Step S402 After receiving the notification message from the network node, send the first confirmation message and the node routing table of the node to the master network node.
  • the first confirmation message is used to notify the master network node that the slave network node has received the notification message.
  • the node routing table is a node routing table generated before the number of network nodes changes.
  • the packet processing method is shown in FIG. 6 .
  • Step S403 the master network node generates a global routing table after receiving the first confirmation message and the node routing table of all the slave network nodes.
  • the main network node summarizes the node routing tables of all network nodes (including the main node itself), and merges them to generate a global routing table.
  • the process of merging is to take the union of multiple node routing tables to obtain the entire network routing table.
  • the corresponding relationship between network nodes and computing task identifiers in the node routing tables of all network nodes is generated according to the original hash routing table.
  • the node routing tables of all network nodes are Summarizing at the main network node ensures that the service packets belonging to the same existing computing task before and after the synchronization of the routing table are all recorded in the global routing table.
  • step S403 the main network node also needs to update the hash routing table of its own network node. In the case of removing or adding nodes, it is necessary to ensure the continuity of the hash values in Table 2.
  • the hash routing table of the master network node is changed to the form shown in Table 3.
  • the number of node_num is decreased by 1.
  • the hash routing table of the main network node is changed to the form shown in Table 4.
  • the number of node_num is incremented by 1.
  • the primary network node If the primary network node has not sent the first confirmation message from the network node beyond the first time threshold, it indicates a timeout.
  • the main network node counts according to the records in the hash routing table.
  • the master network node repeats step S401 to the slave network node that has not sent the confirmation message.
  • Step S404 the master network node sends the global routing table to the slave network node.
  • the main network node processes the service packets arriving at the main network node according to the new hash routing table and the global routing table.
  • Step S405 after receiving the global routing table from the network node, replace the node routing table, and send a second confirmation message to the master network node.
  • the slave network node processes the service packets arriving at the slave network node according to the new hash routing table and the global routing table.
  • the master network node may send the updated hash routing table to the slave network node.
  • the master network node carries the identification information of the newly added or removed node in the notification message and sends it to the slave network node, after step S405, the slave network node updates the hash according to the identification information of the newly added or removed node
  • the specific update process of the routing table is the same as the update process of the main network node in step S404.
  • Step S406 after the master network node receives the second acknowledgment messages from all the slave network nodes, it indicates that the synchronization is completed.
  • the master network node If the master network node does not receive the second acknowledgment message sent by the slave network node beyond the second time threshold, indicating a timeout, the master network node resends the global routing table to the slave network node that has not yet sent the second acknowledgment message.
  • Fig. 6 is a flow chart of the processing of service packets arriving at the slave network nodes of the distributed network between steps S402-S405 in Fig. 4:
  • Step S601 after receiving the service message from the network node, look up in the node routing table according to the computing task identifier of the service message, if there is a record of the target network node corresponding to the computing task identifier in the node routing table, go to: Step S602, if no, go to step S603.
  • Step S602 the slave network node processes the service message according to the node routing table.
  • step S602 is the same as that of S303-S305 in FIG. 3 , and details are not described here.
  • Step S603 the slave network node judges whether the service packet is sent by the master network node according to the master network node field in the service packet; if not, go to step S604, if yes, go to step S606.
  • a primary network node field is added to the service packet of each computing task, and the primary network node field is used to identify that the service packet is taken over by the primary network node during the routing table synchronization process.
  • the primary network node field can be selected and set in the reserved field of the service message.
  • the main network node field is used to identify whether the packet has been processed by the main network node. Exemplarily, 0 indicates that the service packet has not been processed by the main network node, and 1 indicates that the service packet has been processed by the main network node. For any service packet reaching the distributed network 100, this field is set to 0 by default.
  • the primary network node field in the service packet is 1, indicating whether the service packet is sent by the primary network node, and if the primary network node field is 0, it indicates that the service packet has not been processed by the primary network node.
  • Step S604 the slave network node modifies the master network node field in the service packet and forwards the service packet to the master network node for processing.
  • the slave network node modifies the master network node field in the service packet from 0 to 1.
  • Step S605 the primary network node processes the service message.
  • step S605 the process of processing the service message of the computing task by the master network node is the same as that of S301-S305 in FIG. 3, and will not be repeated here.
  • Step S606 the service message is processed at the current node.
  • main network node field indicates that the packet is forwarded by the main network node, indicating that this network node is recognized by the main network node as the target network node, so the packet is processed at this node.
  • the main network node For the service packets arriving at the main network node, the main network node always processes the service packets of the computing task according to the flow in FIG. 3 .
  • the service packets between steps S402 to 403 will be recorded in the global routing table of the master network node to ensure the consistency of subsequent routing table synchronization.
  • the master network node receives the service packets, it changes according to the network structure.
  • the previous node routing table and the unupdated hash routing table process the service message according to the flow of FIG. 3 .
  • the slave nodes will still forward them to the master network node for processing, and the master network node will follow the global routing table and the new hash. It is hoped that the routing table processes the service message according to the flow of FIG. 3 .
  • step S405 the service packet of the computing task that has never been processed by the slave network node that arrives at the slave network node, the slave network node no longer forwards the service packet, and the slave network node follows the global routing table and the new hash routing table according to the service packet.
  • the service message After step S406, it indicates that the routing table update is completed, and all network nodes still process service packets according to the steps in FIG. 3 .
  • the primary network node field is no longer considered when the node processes the message.
  • the packet arriving at the slave network node of the distributed network is the first packet of the computing task and the time is between steps S402-S403, the packet is forwarded to the master network node.
  • the main network node can perform a hash operation on the service packet according to the original hash routing table before the distributed network structure changes. If the IP address of the network node corresponding to the hash value is the IP address of the network node that has not been removed, then Forward the message according to the calculated IP address of the network node and update the node routing table. If the obtained result is the IP address of the removed node, specify any undeleted network node as the target node for the message.
  • the network node judges whether the service message is the first service message of the computing task according to the length of the service message.
  • each network node Before the distributed network node structure changes, if the service packets belonging to the same computing task have been processed on the removed network node, after the distributed network network structure changes, each network node still forwards it to the node routing table according to the node routing table. Remove nodes. After the network node is removed, it is no longer possible to obtain the processing results of the packets belonging to the computing task before the distributed network node structure changes. It is useless to continue processing the packets belonging to the computing task through other network nodes.
  • Embodiment 1 After the distributed network structure changes, the master network node integrates the routing table information of the slave network nodes in the system, and undertakes the service packets of the computing tasks that have never been processed by the slave network nodes, so as to ensure system service forwarding The consistency of the strategy does not interrupt the data processing of the computing task, which improves the data processing efficiency.
  • the master network node is also responsible for the node routing tables of all network nodes.
  • the summary update work of the main network will cause a large load pressure on the main network node.
  • the routing table synchronization strategy in the first embodiment is improved. For details, refer to the flowchart of FIG. 7 :
  • step S401 the difference from the flowchart in FIG. 4 is that before step S401:
  • a master network node When the distributed network is established, a master network node is selected in the distributed network, and the identity of the master network node is broadcast to all slave network nodes.
  • the master network node In the normal communication process without the change of the network structure, the master network node periodically obtains the node routing tables of all the slave network nodes for periodic update. Since periodic updates will occupy a lot of network resources, considering that the format of the first service packet of a new computing task is different from that of subsequent service packets, the first service packet of a new computing task is received at any network node. 3, and actively report the computing task identifier in the message in the local node routing table to the main network node, and the main network node updates the node routing table of the main network node after receiving the information .
  • the master network node may designate a slave network node as a backup network node, and periodically back up the routing table information of the master network node to the backup network node.
  • the backup network node is used as the primary network node.
  • a corresponding field may be set in the packet to identify the first packet or non-first packet of the service.
  • Step S702 after receiving the abnormal message from the slave network node, send a first confirmation message to the master network node.
  • the slave network node Since the master network node has already collected information of all services in the entire network during the normal communication process, the slave network node does not need to send the node routing table to the master network node in step S702.
  • Step S703 specifically takes the node routing table of the master network node as the global routing table for the master network node.
  • step S703 the main network node is required to generate a global routing table.
  • the node routing table of the main network node is not a global routing table, because in step S702, the computing task of the entire network is still being executed, which may be There will be new computing tasks arriving at the network nodes in the distributed network, so between steps S702-S703 in FIG. 7, the node routing table of the master node may be updated.
  • Step S701, step S704-step S706 are the same as step S401, step S404-S406.
  • the main network node may not be set before step S701, and when any network node in the distributed network receives the first packet of the new computing task, the information of the packet may be uniformly broadcast to the entire network, so that The processing ensures that the node routing table of each network node is a global routing table, and the main network node can be selected when the network structure changes to prevent the removal of the main network node. But this will take up additional network transmission resources.
  • the load when the main network node summarizes the node routing table to generate the global routing table when the network structure changes, makes the data processing process more efficient.
  • An embodiment of the present invention further provides a data processing apparatus 800, as shown in FIG. 8 .
  • the data processing apparatus 800 is applied to a main network node, the main network node is used to connect a first network node and a second network node, the second network node is a newly added network node or a faulty network node, and the data processing apparatus 800 includes: Communication module 801 and processing module 802 .
  • the processing module 801 is configured to detect the second network node. Please refer to steps S301-S305 in Fig. 3, S401, S403, S406 in Fig. 4, S601-S606 in Fig. 6, S701, S703, S706 in Fig. 7 for the specific implementation of the functions performed by the processing module, It is not repeated here.
  • a communication module 802 configured to send a notification message to the first network node; and receive a computing task sent by the first network node that the first network node has never received before receiving the notification message. Please refer to step S305 in FIG. 3 , S402 , S404 , S406 in FIG. 4 , S604 in FIG. 6 , S402 , S404 , S406 in FIG. 7 for the specific implementation of the functions performed by the communication module, which are not repeated here. Repeat.
  • the processing module 801 is further configured to process the computing task sent by the first network node.
  • An embodiment of the present invention further provides a data processing apparatus 900, as shown in FIG. 9 .
  • the data processing apparatus 900 is applied to a first network node, the first network node is connected to the main network node, and the apparatus 900 includes: a communication module 901 and a judgment module 902 .
  • a communication module configured to receive a notification message sent by the primary network node, where the notification message is used to notify the primary network node that a second network node is detected, and the second network node is a newly added network node or a faulty network node; receive the first computing task.
  • the communication module please refer to steps S301-S305 in FIG. 3 , S402 and S405 in FIG. 4 , S601 and S604 in FIG. 6 , and S702 and S705 in FIG. 7 , which are not repeated here. Repeat.
  • the judgment module 902 is configured to determine that the received computing task is a computing task that has never been received before the notification message is received. Please refer to steps S301-S305 in FIG. 3 , S405 in FIG. 4 , S601 , S603 , S606 in FIG. 6 , and S705 in FIG. 7 for specific implementations of the functions performed by the judging module, which will not be repeated here.
  • the communication module 901 is further configured to send the computing task to the main network node.
  • the present invention also provides a network node system, the network node system includes a main network node and a first network node, and is characterized in that:
  • the primary network node is configured to execute the method performed by the primary network node in Embodiment 1 or Embodiment 2,
  • the first network node is configured to execute the method executed by the slave network node in Embodiment 1 or Embodiment 2.
  • the present invention also provides a network node, characterized in that the node includes: a processor and a memory, the memory stores program instructions, and the processor executes the program instructions to implement Embodiment 1 or Embodiment 1 The data processing method described in 2.
  • the present invention also provides a computer-readable storage medium, where computer programs or instructions are stored on the computer-readable storage medium, and when the computer programs or instructions are executed in a computer, the computer is made to execute the first embodiment or implement the The data processing method described in Example 2.
  • a seventh aspect of the present invention provides a computer program product, the computer program product including instructions, when executed on a computer, causes the computer to execute the data processing method described in Embodiment 1 or Embodiment 2 above.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

本申请实施例提供一种数据处理方法，主网络节点用于连接第一网络节点及第二网络节点，所述第二网络节点为新增网络节点或者故障网络节点；主网络节点当侦测到所述第二网络节点时，发送通知消息至所述第一网络节点；第一网络节点发送所述通知消息之前从未接收过的计算任务给主网络节点；主网络节点接收所述第一网络节点发送的所述第一网络节点在接收到所述通知消息之前从未接收过的计算任务并处理。本发明的数据处理方法解决现有技术中网内计算在结构发生变化时，需要中断计算任务处理的缺陷。

Description

数据处理方法及装置
本申请要求于2020年10月21日提交中国专利局、申请号为202011131010.0、申请名称为“数据处理方法及装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及网内计算领域，具体涉及一种数据处理方法。
背景技术
在相关技术中，路由器和交换机除了转发业务报文外，还可以具备业务报文处理能力。通过网内计算技术，可以将一些适合通过路由器和交换机处理的计算任务卸载至路由器和交换机等网络节点执行，从而减轻了计算节点的压力。一般情况下，属于同一计算任务的业务报文需要分配给同一网络节点执行，在网络节点接收到业务报文后，会对业务报文中携带的计算任务ID以网络内的网络节点的个数为长度进行哈希计算，以计算处理所述计算任务的网络节点，并在接收到所述计算任务的网络节点中的路由表中记录处理所述计算任务的网络节点。但是当网络中的网络节点的数量发生变化后，例如网络节点故障或者增加新的网络节点，则由于哈希长度的变化，在接收到之前已经接收到计算任务时，则通过哈希计算确定的处理所述计算任务的网络节点与网络节点发生变化之前计算的处理所述计算任务的网络节点不同。相关技术中，为了避免上述情况的发生，会停止发送计算任务，在所有网路节点中的路由表同步之后，再重新开始发送计算任务。
发明内容
本申请实施例旨在提供一种数据处理方法，以在分布式网络的结构发生变化时，不需要中断计算任务的执行。
为实现上述目的，本发明实施例第一方面提供了一种数据处理方法，由主网络节点执行，所述主网络节点用于连接第一网络节点及第二网络节点，所述第二网络节点为新增网络节点或者故障网络节点，所述方法包括：当侦测到所述第二网络节点时，发送通知消息至所述第一网络节点；接收所述第一网络节点发送的所述第一网络节点在接收到所述通知消息之前从未接收过的计算任务；处理所述第一网络节点发送的计算任务。
通过将第一网络节点发送的从未接收过的计算任务由主网络节点处理，避免在分布式网络中网络节点结构变化时，计算任务的处理中断，提高计算任务处理效率。
在一种具体的实施方式中，在所述发送通知消息至所述第一网络节点之后，所述方法还包括：接收所述第一网络节点发送的确认信息，所述确认信息用于指示所述第 一网络节点已接收到所述通知信息；将全局路由表发送至所述第一网络节点，所述全局路由表包括所述主网络节点接收到的第一计算任务及处理所述第一计算任务的网络节点，及所述第一网络节点在接收到所述通知消息之前所接收到的第二计算任务及处理所述第二计算任务的网络节点。
通过向第一网路节点发送全局路由表，保证分布式网络结构变化前后的所有节点路由表的一致性。
在一种具体的实施方式中，所述方法还包括：在发送通知消息至所述第一网络节点后，接收所述第一网络节点发送的第一网络节点中的节点路由表，所述第一网络节点中的节点路由表包括所述第一网络节点在接收到所述通知消息之前所接收到的第二计算任务及处理所述第二计算任务的网络节点；将所述第一网络节点中的节点路由表与所述主网络节点中的节点路由表合并为所述全局路由表，所述主网络节点的路由表包括所述主网络节点接收到的第一计算任务及处理所述第一计算任务的网络节点。
在一种具体的实施方式中，所述方法还包括，第一网络节点在正常工作期间，当接收到计算任务的第一个报文时，根据所述报文更新节点路由表，并将所述节点路由表更新的内容发送给主网络节点。
避免了主节点在路由表同步时期同步所有节点的节点路由表的压力。
在一种具体的实施方式中，在所述处理所述第一网络节点发送的计算任务之后，所述方法还包括：将所述计算任务及处理所述计算任务的网络节点存储至所述主网络节点的路由表。
在网络结构发生变化期间，主网络节点将计算任务的网络节点信息及时保存至主网络节点的路由表中，保证根据主网络节点的路由表生成的全网路由表的准确性。
在一种具体的实施方式中，接收所述第一网络节点发送的确认信息，所述确认信息用于指示所述第一网络节点已接收到所述通知信息之后，还包括，向第一网络节点发送哈希路由表，所述哈希路由表用于以所述主网络节点及所述第一网络节点所在的网络系统中的网络节点的数量为哈希长度计算处理所述计算任务。
在一种具体的实施方式中，在所述发送通知消息至所述第一网络节点之后，若超过第一时间阈值未收到第一网络节点的确认信息，重新向第一网络节点发送通知消息。
通过超时重发机制，保证每个第一网络节点都获取网络结构变化的通知消息。
本发明的第二方面提供了一种数据处理方法，由第一网络节点执行，所述第一网络节点连接至主网络节点，所述方法包括：接收所述主网络节点发送的通知消息，所述通知消息用于通知所述主网络节点侦测到第二网络节点，所述第二网络节点为新增网络节点或者故障网络节点；接收第一计算任务；在确定所接收的计算任务是在收到所述通知消息之前从未接收过的计算任务时，发送所述计算任务至所述主网络节点。
通过将第一网络节点发送的从未接受过的计算任务由主网络节点处理，避免在分布式网络结构变化时，计算任务的处理中断，提高计算任务处理效率。
在一种具体的实施方式中，所述第一网络节点包括节点路由表，所述节点路由表中记录了所述第一网络节点在收到所述通知消息之前接收到的计算任务及处理所述计算任务的网络节点。
在一种具体的实施方式中，所述方法还包括：在确定所接收的计算任务是在收到所述通知消息之前接收过的计算任务时，根据所述第一网络节点中的节点路由表将所述计算任务转发至所述计算任务对应的网络节点。
将原有的业务继续在第一网络节点本节点处理，减轻主网络节点的负载。
在一种具体的实施方式中，所述方法包括：在接收到所述通知消息之后，发送所述节点路由表至所述主网络节点。
在一种具体的实施方式中，所述方法还包括，第一网络节点在正常工作期间，当接收到计算任务的第一个报文时，根据所述报文更新节点路由表，并将所述节点路由表更新的内容发送给主网络节点。
在一种具体的实施方式中，所述方法包括：接收所述主网络节点发送的全局路由表，所述全局路由表包括所述主网络节点接收到的第一计算任务及处理所述第一计算任务的网络节点，及所述第一网络节点在接收到所述通知消息之前所接收到的第二计算任务及处理所述第二计算任务的网络节点；接收第二计算任务，当所述第二计算任务为所述全局路由表中的计算任务时，根据所述全局路由表确定处理所述第二计算任务的网络节点，并将所述第二计算任务发送至所确定的网络节点。
通过向第一网路节点发送全局路由表，保证分布式网络结构变化前后的所有节点路由表的一致性。
在一种具体的实施方式中，所述方法包括：接收所述主网络节点发送的全局路由表，所述全局路由表包括所述主网络节点接收到的第一计算任务及处理所述第一计算任务的网络节点，及所述第一网络节点在接收到所述通知消息之前所接收到的第二计算任务及处理所述第二计算任务的网络节点；接收第二计算任务，当所述第二计算任务不是所述全局路由表中的计算任务时，以所述主网络节点及所述第一网络节点所在的网络系统中的网络节点的数量为哈希长度计算处理所述第二计算任务的网络节点。
接收全局路由表标志着路由表更新的结束，按照新的网络结构的哈希路由表计算任务的目标网络节点减轻了主网络节点的负载。
本发明的第三方面提供了一种数据处理装置用于执行第一方面所述的方法。
本发明的第四方面提供了一种数据处理装置用于执行第二方面所述的方法。
本发明的第五方面提供了一种网络节点系统，所述网络节点系统包括主网络节点和第一网络节点，其特征在于：
所述主网络节点用于执行第一方面所述的方法，
所述第一网络节点用于执行第二方面所述的方法。
本发明的第五方面提供了一种网络节点，其特征在于所述节点包括：处理器和存储器，所述存储器中存储有程序指令，所述处理器执行所述程序指令以实现第一方面或第二方面所述的方法。
本发明的第六方面提供了一种计算机可读存储介质，所述计算机可读存储介质上存储有计算机程序或指令，当所述计算机程序或指令在计算机中执行时，令计算机执行第一或第二方面所述的方法。
本发明的第七方面提供了一种计算机程序产品，所述计算计算机程序产品包括指令，当其在计算机上运行时，使得计算机执行上述第一或第二方面所述的方法。
附图说明
通过结合附图描述本说申请实施例，可以使得本申请实施例更加清楚：
图1示出了本发明分布式网络的结构示意图；
图2示出了本发明分布式网络中网络节点的结构示意图；
图3示出了本发明中网络节点业务报文处理流程图；
图4示出了本发明的一种网络节点结构发生变化后路由表同步的流程图；
图5a示出了本发明实施例1中的移除网络节点的情形；
图5b示出了本发明实施例1中的增加网络节点的情形；
图6示出了本发明图4中报文处理的流程图；
图7示出了本发明的另一种分布式网络结构发生变化后路由表同步的流程图；
图8示出了本发明的数据处理装置的结构方框图；
图9示出了本发明的另一种数据处理装置的结构方框图。
具体实施方式
为了使本申请的上述目的、技术方案和优点更易于理解，下文提供了详细的描述。所述详细的描述通过使用方框图、流程图和/或示例提出了设备和/或过程的各种实施例。由于这些方框图、流程图和/或示例包含一个或多个功能和/或操作，所以本领域内人员将理解可以通过许多硬件、软件、固件或它们的任意组合单独和/或共同实施这些方框图、流程图或示例内的每个功能和/或操作。
图1为分布式网络100网络架构的简化方框图，分布式网络100中包括多个网络节点和计算节点，示例性的以四个网络节点和四个计算节点为例，分别为网络节点101、网络节点102、网络节点103、网络节点104和计算节点105，计算节点106，计算节点107，计算节点108。网络节点与对应的计算节点相连，如图所示，网络节点101与计算节点105相连，网络节点102与计算节点106相连，网络节点103与计算节点107相连，网络节点104与计算节点108相连；分布式网络100中的网络节点间是全互连的。
计算节点之间通过网络节点连接，计算节点用于数据处理，并通过网络节点进行数据的传输。其中计算节点是网络中具有数据处理能力的终端设备，如主机或存储设备等。网络节点用于转发来自计算设备的数据，其中网络节点路由器或交换机等。
由于目前的网络节点除了基本的业务报文转发的功能外，还可以做一些特定计算， 这样，对于计算节点之间传输的业务报文，可以在通过网络节点传输时，在网络节点中对所传输的业务报文进行特定的处理，从而减轻计算节点的负载，这就是网内计算技术。通过网内计算技术，可以将一些适合通过网络节点处理的计算任务卸载至网络节点执行。一般情况下，属于同一计算任务的业务报文需要分配给同一网络节点执行，在网络节点接收到业务报文后，会对业务报文中携带的计算任务标识以网络内的网络节点的个数为长度进行哈希计算，以计算处理所述计算任务的网络节点，并在接收到所述计算任务的网络节点中的路由表中记录处理所述计算任务的网络节点，然后将业务报文转发至所确定的网络节点执行，等下次再接收到同一计算任务的业务报文时，则直接根据计算任务ID查询路由表即可确定处理所述业务报文的网络节点。
图2为所述具有计算能力的网络节点200的结构图。网络节点200包括与一个或多个数据存储单元连接的处理器201。数据存储单元可以包括存储介质203和内存单元(cache)204。存储介质203可以是只读的，如只读存储器(ROM)，或是可读写的，如硬盘或闪存。内存单元204可以与处理器201集成在一起也可为独立的元件。处理器201是网络节点200的控制中心，用于执行程序代码以实现与所述程序指令对应的功能。可选地，处理器201包括一个或多个网络处理器单元(Network Processer Unit，NPU)，用于处理业务报文，例如，如图2所示的NPU0和NPU1。可选地，网络节点200包括一个以上的处理器，例如，如图2所示的处理器201和205。本文所使用的术语“处理器”指一个或多个用于处理计算机程序指令等数据的设备、电路和/或处理内核。可替换的，网络处理器可以是中央处理器(CPU)。
处理器201和/或205的CPU将执行的程序代码存储在内存单元204或存储介质203中。可选地，存储介质203中存储的程序代码可以复制到内存单元204中以便处理器执行。根据程序代码，在内存单元204中对接收到的数据报文进行处理。处理器可执行至少一个内核(例如，以LINUZ TM、UNIX TM、WINDOWS TM、ANDROID TM、IOS TM等品牌厂商出售的操作系统中的内核)，众所周知该内核通过控制其它程序的执行、控制与外围设备的通信以及控制网络节点资源的使用来控制网络节点200的运行。
网络节点200还包括通信接口202，网络节点200可通过该通信端口202直接或通过外部网络与另一设备或系统进行通信。网络节点200的上述元件可通过数据总线、地址总线、控制总线、扩展总线和本地总线等总线中的任一或任意组合互相连接。
作为实用示例，上述网络节点200可以是具有数据处理功能的交换机、具有数据处理功能的路由器或具有如图2所示类似结构的用于业务报文转发和处理的任何其它设备。
对于需要进行网内计算的业务报文，会在业务报文中携带需要通过网络节点执行的计算任务的标识(ID)。在支持网内计算的分布式网络100建立时，管理员会为网络中每一个网络节点配置两张路由表。在分布式网络网络100工作正常时，业务报文到达网络100中任意一个网络节点后，根据业务报文中携带的计算任务的ID两张路由表得到执行所述计算任务的目标网络节点地址，并将业务报文转发到目标网络节点进行处理。
具体的，两张路由表分别为节点路由表和哈希路由表。其中节点路由表包括计算任务标识与目标网络节点的对应关系，如表1所示；哈希路由表包括哈希值与目标网 络节点的对应关系，如表2所示。关于如何根据上述两张路由表确定执行业务报文中携带的计算任务的目标节点的方法将结合图3所述的流程图进行描述。
表1
计算任务标识 网络节点IP地址
00001 10.0.0.1(节点101)
00002 10.0.0.2(节点102)
表2
哈希值 网络节点IP地址
0 10.0.0.1(节点101)
1 10.0.0.2(节点102)
2 10.0.0.3(节点103)
3 10.0.0.4(节点104)
接下来，结合图3对网络节点确定执行业务报文中的计算任务的目标网络节点的流程进行介绍。
步骤S301、当业务报文到达网络节点时，确定业务报文中的计算任务标识是否存在所述节点路由表中。若不存在，则转至步骤S302；若存在，转至步骤S303。
如果业务报文中的计算任务标识不包含在所述节点路由表中，则说明所述网络节点之前没有接收过需要执行所述计算任务的业务报文，如果业务报文中的计算任务标识不包含在所述节点路由表中，则说明所述网络节点之前已经接收过需要执行所述计算任务的业务报文，
步骤S302、对计算任务标识进行哈希计算，得到哈希值，根据哈希值在哈希路由表中获取哈希值对应的目标网络节点的地址，并将哈希值及对应的目标网络节点的地址更新到节点路由表。
在对计算任务标识进行哈希计算时，所述哈希计算的哈希长度为分布式网络中的网络节点的数量。
示例性的，哈希计算公式为：id＝Hash(taskID)mod node_num；其中taskID为业务报文中的计算任务标识，node＿num为网络100中网络节点的数量。其中网络节点的数量为网络节点的哈希路由表中的行数。
步骤S303、获取所述计算任务标识对应的目标网络节点的地址，并根据目标网络节点地址判断本网络节点是否为目标网络节点。若是，则转至步骤S304；否则转至步骤S305。
根据目标网络节点的IP地址与本网络节点的IP地址进行匹配，若相同则判断本网络节点为目标网络节点。
步骤S304、在本网络节点处理业务报文。
步骤S305、本网络节点转发业务报文到目标网络节点，以使目标网络节点处理所述业务报文。
以上为分布式网络结构未发生变化时进行网内计算进行数据处理时的流程。
但是当网络中的网络节点的数量发生变化后，例如网络节点故障或者增加新的网络节点，对于第一次接收到一个计算任务的业务报文的网络节点，会按照新的哈希长 度计算哈希值，以确定处理所述业务报文的目标网络节点，然而这个计算任务的业务报文可能之前已被其他网络节点接收，并以之前的网络节点的数量作为哈希长度计算了哈希值，并确定了目标网络节点，这将导致网络节点的数量变化前后，对于属于相同计算任务的业务报文会确定出不同的目标网络节点，即属于相同计算任务的业务报文被不同的网络节点处理。相关技术中，为了避免上述情况的发生，计算节点会停止发送业务报文，在所有网路节点中的路由表同步之后，再重新开始发送计算任务。
本发明实施例提供的技术方案可以在分布式网络中的网络节点的数量发生变化后，从网络节点将所接收到之前从没有接收过的计算任务的业务报文转发至主网络节点处理，从而避免网络节点的数量发生变化后，计算节点停止发送业务报文，从而保证计算任务的连续执行。
本发明实施例提供的方案将结合图4、图5a，5b、图6详细描述。其中图4为本发明实施例中在网络节点的数量发生变化后，各网络节点中的路由表同步的流程示意图，图5a，5b中为分布式网络中100中移除故障网络节点或添加新的网络节点的示意图，图6为本发明实施例中各网络节点进行路由表同步过程中的对接收到的业务报文的处理流程。
在分布式网络100搭建时，管理员在分布式网络100的网络节点中选择一个网络节点作为主网络节点，并将主网络节点的地址发送给从网络节点。主网络节点用于管理整个路由表更新过程与路由表更新过程中的数据处理。通过主网络节点，保证网络节点路由表更新后网络节点间的路由表是一致的，并且保证整个网内计算的数据处理不中断。本发明未指定选择主网络节点的过程，可以通过现有技术中任意方式或方式组合进行选举。如通过所有的网络节点选举主网络节点，由管理员指定主网络节点，按照网络节点特性，如剩余带宽等进行排序选择带宽最大的网络节点为主网络节点等。
由于网络节点结构变化的不确定性，可能会出现主网络节点由于故障被移除的情况。为了防止主网络节点被移除的情况，管理员可以指定一个从网络节点作为主网络节点的备份网络节点，将主网络节点的节点路由表周期性的备份到该备份节点中。在主网络节点被移除导致的网络100结构变化的情况发生时，由备份网络节点监测到主网络节点故障，随后备份网络节点作为主网络节点，并广播该节点的地址给其他从网络节点。可选的，为了防止主网络节点故障的情况，主网络节点可以在网络发生异常后进行再设定。
步骤S401、主网络节点侦测到网络节点数量发生变化，向从网络节点发送通知消息。
所述通知消息用于通知从网络节点分布式网络100结构发生变化。
具体的，主网络节点可以与网络中所有的从网络节点进行心跳连接，当某个从网络节点的心跳连接断开时，判断该网络节点故障；或，新加入网络100的网络节点向网络中所有的网络节点与计算节点广播报文，主网络节点接收到该报文后，判断有新节点加入网络100。
步骤S402、从网络节点收到通知消息后，发送第一确认报文和本节点的节点路由表给主网络节点。
其中第一确认报文用于通知所述主网络节点所述从网络节点已经收到所述通知消 息。所述节点路由表为网络节点的数量发生变化前生成的节点路由表。
在步骤S402之后到步骤S406之前，对于到达任意一个网络节点的报文依旧进行处理，保证网内计算的计算任务的业务报文转发和处理不中断。具体的，报文处理方式如图6所示。
步骤S403、主网络节点在接收到所有的从网络节点的第一确认报文和节点路由表后，生成全局路由表。
具体的，主网络节点将所有网络节点(包括主节点本身)的节点路由表进行汇总，合并生成全局路由表，合并的过程是将多个节点路由表取并集，得到全网路由表，由于在网络结构发生变化前，所有网络节点的节点路由表中网络节点与计算任务标识的对应关系都是根据原哈希路由表生成的，在网络结构变化后，将所有的网络节点的节点路由表在主网络节点处汇总保证在路由表同步前后属于相同的已有的计算任务的的业务报文均记载在全局路由表中。
在步骤S403中，主网络节点还需要更新本网络节点的哈希路由表，对于移除或增加节点的情况，需要保证表2中哈希值的连续性。
示例性的，如图5a所示，若分布式网络100中的网络节点103因故障被移除，主网络节点的哈希路由表变更为表3所示的形式。步骤S303中的哈希计算公式中node_num数量减1。
表3
哈希值 网络节点IP地址
0 10.0.0.1(网络节点101)
1 10.0.0.2(网络节点102)
2 10.0.0.4(网络节点104)
如图5b所示，若分布式网络100中的新增节点105，主网络节点哈希路由表变更为表4所示的形式。步骤S303中的哈希计算公式中node＿num数量加1。
表4
哈希值 节点地址
0 10.0.0.1(网络节点101)
1 10.0.0.2(网络节点102)
2 10.0.0.3(网络节点103)
3 10.0.0.4(网络节点104)
4 10.0.0.5(网络节点109)
若超过第一时间阈值主网络节点仍未从网络节点发送的第一确认报文，表明超时。主网络节点根据哈希路由表中的记录统计。主网络节点向未发送确认报文的从网络节点重新进行步骤S401。
步骤S404、主网络节点将全局路由表发送给从网络节点。
在步骤S404后，主网络节点按照新的哈希路由表和全局路由表处理到达主网络节点的业务报文。
步骤S405、从网络节点收到全局路由表后替换节点路由表，并发送第二确认报文给主网络节点。
在步骤S405后，从网络节点按照新的哈希路由表和全局路由表处理到达从网络节点的业务报文。对于从节点的新的哈希路由表，可选的，在步骤S404中，主网络节点可以将更新后的哈希路由表发送给从网络节点。或，在步骤S401中，主网络节点在通知消息中携带新增或移除节点的标识信息发送给从网络节点，步骤S405后，从网络节点根据新增或移除节点的标识信息更新哈希路由表，具体的更新过程，与步骤S404中主网络节点的更新过程相同。
步骤S406、主网络节点在接收到所有的从网络节点的第二确认报文后，表明同步完成。
若超过第二时间阈值主网络节点仍未收到从网络节点发送第二确认报文，表明超时，则主网络节点向仍未发送第二确认报文的从网络节点重新发送全局路由表。
图6为图4中步骤S402－S405之间，到达分布式网络从网络节点的业务报文的处理流程图：
步骤S601、在从网络节点接收到业务报文后，根据该业务报文的计算任务标识在节点路由表中进行查找，若节点路由表中存在计算任务标识对应的目标网络节点的记录，转至步骤S602，若否，转至步骤S603。
步骤S602、从网络节点依据节点路由表处理业务报文。
具体的，步骤S602的处理流程与图3中S303－S305相同，在此不做赘述。
步骤S603、从网络节点根据业务报文中主网络节点字段判断业务报文是否是主网络节点发送的；若否，则转向步骤S604，若是，则转向步骤S606。
本实施例中在每个计算任务的业务报文中增加了主网络节点字段，主网络节点字段用于标识在路由表同步过程中业务报文被主网络节点接管。所述主网络节点字段可以在业务报文的预留字段中选择并设置。主网络节点字段用于标识该报文是否由在主网络节点处理过，示例性的，0表示该业务报文未被主网络节点处理过，1该业务报文表示被主网络节点处理过。对于任意到达分布式网络100的业务报文，默认的，该字段置为0。
具体的，业务报文中主网络节点字段为1，表明业务报文是否是主网络节点发送的，若主网络节点字段为0，表明业务报文未经过主网络节点处理。
步骤S604、从网络节点修改业务报文中主网络节点字段并将业务报文转发到主网络节点处理。
示例性的，从网络节点将业务报文中的主网络节点字段从0修改为1。
步骤S605、主网络节点处理该业务报文。
具体的，步骤S605主网络节点处理计算任务的业务报文的流程与图3中S301－S305相同，在此不做赘述。
步骤S606、将业务报文在本节点处理。
如果主网络节点字段为1，表明该报文为主网络节点转发的报文，说明本网络节点被主网络节点认定为目标网络节点，所以在本节点处理该报文。
对于到达主网络节点的业务报文来说，主网络节点一直按照图3的流程处理计算任务的业务报文。
在步骤S402－步骤403之间的业务报文均会在主网络节点的全局路由表中记录， 保证后续路由表同步的一致性，主网络节点在接收到业务报文时，按照网络结构发生变化前的节点路由表和未更新的哈希路由表按照图3的流程处理业务报文。对于步骤S404－S405之间的到达从网络节点的从网络节点从未处理过的计算任务的业务报文，从节点依旧会转发给主网络节点处理，主网络节点按照全局路由表和新的哈希路由表按照图3的流程处理该业务报文。在步骤S405后，到达从网络节点的从网络节点从未处理过的计算任务的业务报文，从网络节点不再转发该业务报文，从网络节点按照全局路由表和新的哈希路由表处理该业务报文。步骤S406后，表明路由表更新完成，所有网络节点依旧按照图3的步骤进行业务报文的处理。在节点处理报文时不再考虑主网络节点字段。
对于因故障移除节点的情况，如果到达分布式网络从网络节点的报文为计算任务的第一个报文且时间在步骤S402－S403之间，则该报文被转发到主网络节点。主网络节点可以将此业务报文先根据分布式网络结构变化前的原哈希路由表进行哈希运算，如果哈希值对应的网络节点IP地址为未移除的网络节点的IP地址，则将此报文按照计算得到的网络节点的IP地址进行转发并更新节点路由表，如果得到的结果为移除的节点的IP地址，将该报文指定任意一个未删除的网络节点作为目标节点进行转发并处理，并在将计算任务标识和目标节点的IP地址的对应关系更新主网络节点的节点路由表。需要解释的是，其中计算任务的第一个报文的结构与后续属于同一个计算任务的报文结构不同，通常，计算任务的第一个报文不包含数据，报文长度小于后续的属于同一个计算任务的报文，网络节点通过业务报文的长度判断该业务报文是否为计算任务的第一个业务报文。
对于分布式网络节点结构变化前，若属于同一个计算任务的业务报文已经在移除的网络节点上处理，则在分布式网络的网络结构变化后，各网络节点依旧按照节点路由表转发到移除节点。因为网络节点移除后已经无法获取分布式网络节点结构变化前属于该计算任务的报文处理结果，若继续通过其他网络节点处理属于该计算任务的报文是无用的。
实施例一在分布式网络结构发生变化后，通过主网络节点对系统中从网络节点路由表信息的整合，并承接从网络节点的从未处理过的计算任务的业务报文，保证系统业务转发策略的一致性的同时不中断计算任务的数据处理，提高了数据处理效率。
实施例二、
考虑到实施例一数据处理过程中所有的不包含在从网络节点的节点路由表中的业务报文均由主网络节点处理，除此之外，主网络节点还负责所有网络节点的节点路由表的汇总更新工作，这会导致主网络节点的负载压力较大。在此基础上，对实施例一中的路由表同步策略进行改进，具体参见图7的流程图：
与图4的流程图区别在于：在步骤S401之前：
在分布式网络建立时，在分布式网络中选择主网络节点，并将主网络节点的标识广播给所有的从网络节点。在未发生网络结构的变化的正常通信过程中，主网络节点周期性的获取所有从网络节点的节点路由表进行周期性的更新。由于周期性更新会占用大量的网络资源，考虑到新计算任务的首个业务报文的格式与后续的业务报文的格式不同，在任意一个网络节点接收到新计算任务的首个业务报文时，按照图3的步骤 处理并将本地的节点路由表中的关于该报文中的计算任务标识主动上报给主网络节点，主网络节点在收到该信息后更新主网络节点的节点路由表。这样保证了在网络结构发生变化前，主网络节点的节点路由表的信息是全网络的。由于后续网络结构变化的不确定性，可能会出现主网络节点被移除的情况。为了防止主网络节点被移除的情况，可以由主网络节点指定一个从网络节点作为备份网络节点，将主网络节点的路由表信息周期性的备份到该备份网络节点中。在主网络节点被移除导致的网络100结构变化的情况发生时，由备份网络节点作为主网络节点。
可选的，可以通过在报文中设置相应的字段用于标识业务的第一个报文或非第一个报文。
与实施例一图4的步骤不同的是：
步骤S702、从网络节点收到异常报文后，发送第一确认报文给主网络节点。
由于主网络节点在正常通信的过程中已经收集了全网络中所有业务的信息，步骤S702中从网络节点就不需要将节点路由表发送给主网络节点了。
步骤S703具体为主网络节点将主网络节点的节点路由表作为全局路由表。
需要解释的是，在步骤S703需要主网络节点生成全局路由表，在步骤S701之前，主网络节点的节点路由表并非全局路由表，因为在步骤S702时，全网的计算任务依旧在执行，可能会有新的计算任务到达分布式网络中的网络节点，所以，在图7的步骤S702－S703之间，主节点的节点路由表可能会进行更新。
步骤S701、步骤S704－步骤S706与步骤S401、步骤S404－S406相同。
可选的，可以在步骤S701之前不设置主网络节点，可以在分布式网络中任意一个网络节点收到新计算任务的第一个报文时，统一将该报文的信息广播全网，这样处理保证每一个网络节点的节点路由表均为全局路由表，可以在网络结构发生变化的时候再选择主网络节点，防止主网络节点被移除的情况发生。但这样会占用额外的更多的网络传输资源。
通过在正常业务处理过程中设置主网络节点收集全网计算任务信息，减轻了在网络结构发生变化时主网络节点汇总节点路由表生成全局路由表时的负载，使得数据处理过程更加高效的进行。
本发明的实施例还提供一种数据处理装置800，如图8所示。数据处理装置800应用于主网络节点，所述主网络节点用于连接第一网络节点及第二网络节点，所述第二网络节点为新增网络节点或者故障网络节点，数据处理装置800包括：通信模块801和处理模块802。
处理模块801，用于侦测所述第二网络节点。所述处理模块所执行的功能的具体实现方式请参考图3中步骤S301－S305、图4中的S401、S403、S406，图6中的S601－S606，图7中的S701、S703、S706，在此不再赘述。
通信模块802，用于发送通知消息至所述第一网络节点；接收所述第一网络节点发送的所述第一网络节点在接收到所述通知消息之前从未接收过的计算任务。所述通信模块所执行的功能的具体实现方式请参考图3中步骤S305、图4中的S402、S404、S406，图6中的S604，图7中的S402、S404、S406，在此不再赘述。
所述处理模块801还用于，处理所述第一网络节点发送的计算任务。
本发明的实施例还提供了一种数据处理装置900，如图9所述。数据处理装置900，应用于第一网络节点，所述第一网络节点连接至主网络节点，所述装置900包括：通信模块901和判断模块902。
通信模块，用于接收所述主网络节点发送的通知消息，所述通知消息用于通知所述主网络节点侦测到第二网络节点，所述第二网络节点为新增网络节点或者故障网络节点；接收第一计算任务。所述通信模块所执行的功能的具体实现方式请参考图3中步骤S301－S305、图4中的S402、S405，图6中的S601、S604，图7中的S702、S705，在此不再赘述。
判断模块902，用于确定所接收的计算任务是在收到所述通知消息之前从未接收过的计算任务。所述判断模块所执行的功能的具体实现方式请参考图3中步骤S301－S305、图4中的S405，图6中的S601、S603、S606，图7中的S705，在此不再赘述。
所述通信模块901还用于，发送所述计算任务至所述主网络节点。
本发明的还提供了一种网络节点系统，所述网络节点系统包括主网络节点和第一网络节点，其特征在于：
所述主网络节点用于执行实施例一或实施例二中主网络节点所执行的方法，
所述第一网络节点用于执行实施例一或实施例二中从网络节点所执行的方法。
本发明的还提供了一种网络节点，其特征在于所述节点包括：处理器和存储器，所述存储器中存储有程序指令，所述处理器执行所述程序指令以实现实施例一或实施例二所述的数据处理方法。
本发明的还提供了一种计算机可读存储介质，所述计算机可读存储介质上存储有计算机程序或指令，当所述计算机程序或指令在计算机中执行时，令计算机执行实施例一或实施例二所述的数据处理方法。
本发明的第七方面提供了一种计算机程序产品，所述计算计算机程序产品包括指令，当其在计算机上运行时，使得计算机执行上述实施例一或实施例二所述的数据处理方法。
需要理解，本文中的“第一”，“第二”等描述，仅仅为了描述的简单而对相似概念进行区分，并不具有其他限定作用。
本领域普通技术人员应该还可以进一步意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、计算机软件或者二者的结合来实现，为了清楚地说明硬件和软件的可互换性，在上述说明中已经按照功能一般性地描述了各示例的组成及步骤。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本发明的范围。
以上所述，仅为本发明较佳的具体实施方式，但本发明的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本发明揭露的技术范围内，可轻易想到的变化或替换，都应涵盖在本发明的保护范围之内。因此，本发明的保护范围应该以权利要求的保护范围为准。

Claims (20)

  1. 一种数据处理方法，由主网络节点执行，所述主网络节点用于连接第一网络节点及第二网络节点，所述第二网络节点为新增网络节点或者移除网络节点，其特征在于，所述方法包括：
    当侦测到所述第二网络节点时，发送通知消息至所述第一网络节点；
    接收所述第一网络节点发送的所述第一网络节点在接收到所述通知消息之前从未接收过的第一计算任务的业务报文；
    处理所述第一网络节点发送的所述计算任务的业务报文。
  2. 如权利要求1所述的方法，其特征在于，在所述发送通知消息至所述第一网络节点之后，所述方法还包括：接收所述第一网络节点发送的确认信息，所述确认信息用于指示所述第一网络节点已接收到所述通知信息；
    将全局路由表发送至所述第一网络节点，所述全局路由表包括所述主网络节点接收到的业务报文对应的第二计算任务及处理所述第二计算任务的业务报文的网络节点，及所述第一网络节点在接收到所述通知消息之前所接收到的业务报文对应的第三计算任务及处理所述第三计算任务的业务报文的网络节点。
  3. 如权利要求2所述的方法，其特征在于，所述方法还包括：
    在发送通知消息至所述第一网络节点后，接收所述第一网络节点发送的第一网络节点中的节点路由表，所述第一网络节点中的节点路由表包括所述第一网络节点在接收到所述通知消息之前所接收到的业务报文对应的第三计算任务及处理所述第三计算任务的业务报文的网络节点；
    将所述第一网络节点中的节点路由表与所述主网络节点中的节点路由表合并为所述全局路由表，所述主网络节点的路由表包括所述主网络节点接收到的业务报文对应的第二计算任务及处理所述第一计算任务的业务报文的网络节点。
  4. 如权利要求1－3任意一项所述的方法，其特征在于，在所述处理所述第一网络节点发送的所述第一计算任务的业务报文之后，所述方法还包括：
    将所述第一计算任务及处理所述第一计算任务的报文的网络节点存储至所述主网络节点的节点路由表。
  5. 一种数据处理方法，由第一网络节点执行，所述第一网络节点连接至主网络节点，其特征在于，所述方法包括：
    接收所述主网络节点发送的通知消息，所述通知消息用于通知侦测到第二网络节点，所述第二网络节点为新增网络节点或者移除网络节点；
    接收第一计算任务对应的业务报文；
    在确定所接收的第一计算任务对应的业务报文是在收到所述通知消息之前从未接收过的计算任务的业务报文时，发送所述第一计算任务对应的业务报文至所述主网络节点。
  6. 如权利要求5所述的方法，其特征在于，所述第一网络节点包括节点路由表，所述节点路由表中记录了所述第一网络节点在收到所述通知消息之前接收到业务报文对应的第二计算任务及处理所述第二计算任务的业务报文的网络节点。
  7. 如权利要求6所述的方法，其特征在于，所述方法还包括：
    在确定所接收的计算任务是在收到所述通知消息之前接收过的第二计算任务时，根据所述第一网络节点中的节点路由表将所述第二计算任务的业务报文转发至所述第二计算任务对应的网络节点。
  8. 如权利要求6所述的方法，其特征在于，所述方法包括：
    在接收到所述通知消息之后，发送所述节点路由表至所述主网络节点。
  9. 如权利要求5－8任意一项所述的方法，其特征在于，所述方法包括：
    接收所述主网络节点发送的全局路由表，所述全局路由表包括所述主网络节点接收到的第一计算任务及处理所述第一计算任务的网络节点，及所述第一网络节点在接收到所述通知消息之前所接收到的第二计算任务及处理所述第二计算任务的网络节点；
    接收第二计算任务，当所述第二计算任务为所述全局路由表中的计算任务时，根据所述全局路由表确定处理所述第二计算任务的网络节点，并将所述第二计算任务发送至所确定的网络节点。
  10. 如权利要求5－8任意一项所述的方法，其特征在于，所述方法包括：
    接收所述主网络节点发送的全局路由表，所述全局路由表包括所述主网络节点接收到的第一计算任务及处理所述第一计算任务的网络节点，及所述第一网络节点在接收到所述通知消息之前所接收到的第二计算任务及处理所述第二计算任务的网络节点；
    接收第二计算任务，当所述第二计算任务不是所述全局路由表中的计算任务时，以所述主网络节点及所述第一网络节点所在的网络系统中的网络节点的数量为哈希长度计算处理所述第二计算任务的网络节点。
  11. 一种数据处理装置，应用于主网络节点，所述主网络节点用于连接第一网络节点及第二网络节点，所述第二网络节点为新增网络节点或者故障网络节点，其特征在于，所述装置包括：
    处理模块，用于侦测所述第二网络节点；
    通信模块，用于发送通知消息至所述第一网络节点；接收所述第一网络节点发送的所述第一网络节点在接收到所述通知消息之前从未接收过的计算任务；
    所述处理模块还用于，处理所述第一网络节点发送的计算任务。
  12. 如权利要求11所述的装置，其特征在于，所述通信模块还用于，在所述发送通知消息至所述第一网络节点之后，接收所述第一网络节点发送的确认信息，所述确认信息用于指示所述第一网络节点已接收到所述通知信息；
    将全局路由表发送至所述第一网络节点，所述全局路由表包括所述主网络节点接收到的第一计算任务及处理所述第一计算任务的网络节点，及所述第一网络节点在接 收到所述通知消息之前所接收到的第二计算任务及处理所述第二计算任务的网络节点。
  13. 如权利要求12所述的装置，其特征在于，
    所述通信模块还用于，
    在发送通知消息至所述第一网络节点后，接收所述第一网络节点发送的第一网络节点中的节点路由表，所述第一网络节点中的节点路由表包括所述第一网络节点在接收到所述通知消息之前所接收到的第二计算任务及处理所述第二计算任务的网络节点；
    所述处理模块还用于，将所述第一网络节点中的节点路由表与所述主网络节点中的节点路由表合并为所述全局路由表，所述主网络节点的路由表包括所述主网络节点接收到的第一计算任务及处理所述第一计算任务的网络节点。
  14. 如权利要求11－13任意一项所述的装置，其特征在于，所述处理模块还用于，在所述处理所述第一网络节点发送的计算任务之后，将所述计算任务及处理所述计算任务的网络节点存储至所述主网络节点的节点路由表。
  15. 一种数据处理装置，应用于第一网络节点，所述第一网络节点连接至主网络节点，其特征在于，所述装置包括：
    通信模块，用于接收所述主网络节点发送的通知消息，所述通知消息用于通知所述主网络节点侦测到第二网络节点，所述第二网络节点为新增网络节点或者故障网络节点；接收第一计算任务；
    判断模块，用于确定所接收的计算任务是在收到所述通知消息之前从未接收过的计算任务；
    所述通信模块还用于，发送所述计算任务至所述主网络节点。
  16. 如权利要求15所述的装置，其特征在于，所述第一网络节点包括节点路由表，所述节点路由表中记录了所述第一网络节点在收到所述通知消息之前接收到的计算任务及处理所述计算任务的网络节点。
  17. 如权利要求16所述的装置，其特征在于，所述通信模块还用于，
    在确定所接收的计算任务是在收到所述通知消息之前接收过的计算任务时，根据所述第一网络节点中的节点路由表将所述计算任务转发至所述计算任务对应的网络节点。
  18. 如权利要求16所述的装置，其特征在于，所述通信模块还用于，
    在接收到所述通知消息之后，发送所述节点路由表至所述主网络节点。
  19. 如权利要求15－18任意一项所述的装置，其特征在于，
    所述通信模块还用于，接收所述主网络节点发送的全局路由表，所述全局路由表包括所述主网络节点接收到的第一计算任务及处理所述第一计算任务的网络节点，及所述第一网络节点在接收到所述通知消息之前所接收到的第二计算任务及处理所述第二计算任务的网络节点；
    接收第二计算任务；
    所述判断模块还用于，当所述第二计算任务为所述全局路由表中的计算任务时， 根据所述全局路由表确定处理所述第二计算任务的网络节点；
    所述通信模块还用于，将所述第二计算任务发送至所确定的网络节点。
  20. 如权利要求15－18任意一项所述的装置，其特征在于，
    所述通信模块还用于，接收所述主网络节点发送的全局路由表，所述全局路由表包括所述主网络节点接收到的第一计算任务及处理所述第一计算任务的网络节点，及所述第一网络节点在接收到所述通知消息之前所接收到的第二计算任务及处理所述第二计算任务的网络节点；接收第二计算任务；
    所述判断模块还用于，当所述第二计算任务不是所述全局路由表中的计算任务时，以所述主网络节点及所述第一网络节点所在的网络系统中的网络节点的数量为哈希长度计算处理所述第二计算任务的网络节点。
PCT/CN2021/123930 2020-10-21 2021-10-14 数据处理方法及装置 WO2022083503A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011131010.0A CN114448889A (zh) 2020-10-21 2020-10-21 数据处理方法及装置
CN202011131010.0 2020-10-21

Publications (1)

Publication Number Publication Date
WO2022083503A1 true WO2022083503A1 (zh) 2022-04-28

Family

ID=81289668

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/123930 WO2022083503A1 (zh) 2020-10-21 2021-10-14 数据处理方法及装置

Country Status (2)

Country Link
CN (1) CN114448889A (zh)
WO (1) WO2022083503A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050198370A1 (en) * 2004-02-01 2005-09-08 Phonex Broadband Corporation Method for creating, deleting, and maintaining logical networks
WO2009039796A1 (fr) * 2007-09-26 2009-04-02 Huawei Technologies Co., Ltd. Procédé, système et dispositif d'acheminement de messages et procédé et système de sélection de ressource de sauvegarde
CN106302841A (zh) * 2015-05-18 2017-01-04 中兴通讯股份有限公司 一种运营商级网络地址转换的方法及装置
CN109219102A (zh) * 2017-06-30 2019-01-15 华为技术有限公司 一种数据转发方法和装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050198370A1 (en) * 2004-02-01 2005-09-08 Phonex Broadband Corporation Method for creating, deleting, and maintaining logical networks
WO2009039796A1 (fr) * 2007-09-26 2009-04-02 Huawei Technologies Co., Ltd. Procédé, système et dispositif d'acheminement de messages et procédé et système de sélection de ressource de sauvegarde
CN106302841A (zh) * 2015-05-18 2017-01-04 中兴通讯股份有限公司 一种运营商级网络地址转换的方法及装置
CN109219102A (zh) * 2017-06-30 2019-01-15 华为技术有限公司 一种数据转发方法和装置

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
H. SONG: "Network Telemetry Framework; draft-ietf-opsawg-ntf-03.txt", INTERNET ENGINEERING TASK FORCE, IETF, 13 April 2020 (2020-04-13), pages 1 - 34, XP015139097 *

Also Published As

Publication number Publication date
CN114448889A (zh) 2022-05-06

Similar Documents

Publication Publication Date Title
US10938710B2 (en) Protection switching method and system, and network device
CN109344014B (zh) 一种主备切换方法、装置及通信设备
CN106330475B (zh) 一种通信系统中管理主备节点的方法和装置及高可用集群
US10560550B1 (en) Automatic configuration of a replacement network device in a high-availability cluster
WO2020030000A1 (zh) 容灾切换方法、相关设备及计算机存储介质
CN110324375B (zh) 一种信息备份方法及相关设备
US10530634B1 (en) Two-channel-based high-availability
CN113746733A (zh) 表项同步方法、网关设备、组网系统及存储介质
US11108623B2 (en) Rapid owner selection
US9246796B2 (en) Transmitting and forwarding data
CN113055196A (zh) 报文处理方法、网络设备及计算机可读存储介质
CN113810439B (zh) 一种以太网存储系统及其信息通告方法和相关装置
JP4344333B2 (ja) パケット転送装置、パケット転送ネットワークシステムおよびパケット転送方法
WO2022083503A1 (zh) 数据处理方法及装置
CN110661599B (zh) 一种主、备节点间的ha实现方法、装置及存储介质
US20060274646A1 (en) Method and apparatus for managing network connection
CN112260971B (zh) 网络设备系统的容错方法、装置、计算机设备和存储介质
CN115152192B (zh) Pce受控网络可靠性
US11432172B2 (en) Channel establishment method and base station
CN109818870B (zh) 一种组播选路方法、装置、业务板及机器可读存储介质
WO2021249173A1 (zh) 一种分布式存储系统及其异常处理方法和相关装置
WO2023124663A1 (zh) 一种集群仲裁方法、网络设备及系统
CN111107006B (zh) 路由的管理方法及装置
US20210211381A1 (en) Communication method and related device
CN114039695A (zh) 冗余网络的时间同步方法、冗余网络系统及设备

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21881921

Country of ref document: EP

Kind code of ref document: A1