Processing method of ipv6 packet, forwarding router and electronic device

Abstract

The application relates to the technical field of communication, and discloses an IPv6 message processing method, a forwarding router and electronic equipment, which comprises the following steps: acquiring an IPv6 message, wherein the IPv6 message comprises an IPv6 extension header, the IPv6 extension header comprises intelligent body semantic information, the intelligent body semantic information comprises semantic action type information, the semantic action type information is used for indicating that a processing action on the intelligent body is completed on a data plane of a routing device, the processing action type comprises any one of searching, calling and state synchronization, and the processing action on the IPv6 message is executed according to the semantic action type information. According to the method, the semantic action type information is used, the router can directly identify whether the current message belongs to searching, calling or state synchronization operation on the data plane, corresponding forwarding or processing logic can be executed according to different operation types, and semantic perception and classified processing of the intelligent body request are realized.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a method for processing IPv6 packets, a forwarding router, and an electronic device. Background Technology

[0002] With the rapid development of artificial intelligence technology, a large number of intelligent agents with different functions have emerged in the network, such as general-purpose large language models, legal consulting agents, and visual recognition agents. A significant characteristic of these agents is that multiple agents of different types may be deployed in the same network location, and agents of the same type may also be distributed in different network locations.

[0003] In this context, the traditional IPv6 routing method, which forwards packets based solely on the destination IP address, is insufficient for efficiently forwarding and scheduling IPv6 packets containing requests from intelligent agents. Summary of the Invention

[0004] The purpose of this invention is to provide at least one method for processing IPv6 packets, a forwarding router, and an electronic device, which can at least solve the technical problem that existing methods are difficult to efficiently forward and schedule IPv6 packets containing agent requests, and can at least achieve efficient forwarding and scheduling of IPv6 packets containing agent requests.

[0005] To address the aforementioned technical problems, at least one embodiment of this application provides a method for processing IPv6 packets, including: Obtain an IPv6 packet, the IPv6 packet including an IPv6 extension header, the IPv6 extension header including agent semantic information, the agent semantic information including semantic action type information, the semantic action type information being used to indicate the completion of processing actions on the agent in the data plane of the routing device, the processing action type including any one of search, call, and state synchronization; Based on the semantic action type information, perform the processing action for the IPv6 packet.

[0006] In some optional embodiments, the semantic information of the agent may further include semantic action indication information, which is matched with the semantic action type information and is used to indicate the action constraints that need to be referenced when performing processing actions for the IPv6 packet.

[0007] In some optional embodiments, when the semantic action type information is a search, the corresponding matching semantic action indication information includes: Intelligent agent semantic identifiers are used to identify the intelligent agent to be searched.

[0008] In some optional embodiments, when the semantic action type information is a search, the corresponding matching semantic action indication information further includes: Task context information is used to indicate that the search should be prioritized for agents that match the task context information from the downstream agents of the forwarding router.

[0009] In some optional embodiments, when the semantic action type information is a search, the corresponding matching semantic action indication information further includes: task computing power requirement information and / or path computing power requirement information; The task computing power requirement information is used to indicate the computing power requirement required for the searched intelligent agent to perform the target task; The path computing power requirement information is used to indicate the computing power requirements that the path for routing the IPv6 packet to the searched agent must meet.

[0010] In some optional embodiments, when the semantic action type information is a call, the corresponding matching semantic action indication information includes at least one of the following: region identifier, load level information, and task context information: The region identifier is used to indicate the call path selection based on the deployment location of the agent; The load level information is used to indicate the call path to be selected based on the agent's current load level; The task context information is used to instruct the agent-based task context information to select the invocation path.

[0011] In some optional embodiments, when the semantic action type information is state synchronization, the corresponding matching semantic action indication information includes load level information and / or task context information; The load level information is used to indicate the state synchronization of the agent's load level information in the local agent forwarding table; The task context information is used to instruct the agent to synchronize its state in the local agent forwarding table.

[0012] In some optional embodiments, the IPv6 extension header is characterized by being any one of a hop-by-hop option header, a destination option header, and a segmented routing header.

[0013] In some optional embodiments, the forwarding router is characterized in that it is a pre-configured forwarding router with the ability to recognize the semantic information of the intelligent agent; Specifically, for forwarding routers that do not have the ability to recognize the semantic information of the intelligent agent, after obtaining the IPv6 packet, they process it according to the standard forwarding rules of IPv6 packets.

[0014] At least one embodiment of this application also provides a forwarding router, comprising: The acquisition module is used to acquire IPv6 packets, which include IPv6 extension headers, agent semantic information, and semantic action type information. The semantic action type information is used to indicate the processing action of the agent to be completed in the data plane of the routing device. The processing action type includes any one of search, call path selection, and state synchronization. The execution module is used to perform processing actions on the IPv6 packet according to the semantic action type information.

[0015] At least one embodiment of this application also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the above-described IPv6 packet processing method.

[0016] The IPv6 packet processing method, forwarding router, and electronic device provided in the embodiments of this application effectively solve the technical problem of efficient forwarding and scheduling of IPv6 packets containing agent requests by introducing agent semantic information into the IPv6 extension header and defining the semantic action type information (search, call, state synchronization) contained therein. At the same time, it is clear that the information is used to instruct the processing action of the agent to be completed in the data plane of the routing device.

[0017] Specifically, traditional IPv6 routing forwards packets solely based on the destination IP address, failing to recognize whether a packet is related to an agent or differentiate between different types of agent requests. This results in all agent packets being treated the same, and routers cannot differentiate processing based on the semantic intent carried by the packets. This method, through semantic action type information, enables routers to directly identify whether a packet belongs to a search, call, or state synchronization operation at the data plane. This allows for the execution of corresponding forwarding or processing logic for different operation types, achieving semantic awareness and classification of agent requests. Secondly, in existing technologies, agent-related processing often requires sending packets to the control plane CPU for parsing and forwarding decisions. This not only increases processing latency but also easily causes control plane congestion or even crashes in large-scale, high-frequency agent interactions. This method explicitly limits the processing of semantic action type information to the routing device's data plane, eliminating the need to send it to the CPU. This significantly reduces processing latency, avoids control plane overload risks, and improves system scalability and stability. By using the "state synchronization" type in the semantic action type information, a data plane channel is provided for agents to actively report their own state. This enables the router to update the agent's state in real time without relying on the control plane protocol. This ensures that subsequent search and invocation actions can make decisions based on the latest state information, avoiding forwarding requests to unavailable or overloaded agents, and significantly improving the efficiency and reliability of communication between agents. Attached Figure Description

[0018] One or more embodiments are illustrated by way of example with reference to the accompanying drawings, and these illustrative descriptions do not constitute a limitation on the embodiments.

[0019] Figure 1 This is a flowchart of an IPv6 packet processing method provided in one embodiment of this application; Figure 2 This is a schematic diagram of the expansion head structure provided in one embodiment of this application; Figure 3 This is a schematic diagram of a forwarding router provided in another embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details are presented in the various embodiments of this application to facilitate a better understanding of the application. However, the technical solutions claimed in this application can be implemented even without these technical details and various variations and modifications based on the following embodiments. The division of the various embodiments below is for ease of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.

[0021] The following is a detailed description of the implementation details of the IPv6 packet processing method in this embodiment. The following content is only for the convenience of understanding and is not necessary for implementing this solution.

[0022] Example 1: The IPv6 packet processing method of this embodiment can be applied to forwarding routers with communication, computing, and data storage capabilities or electronic devices with similar functions. Its specific process can be as follows: Figure 1 As shown, it includes: Step 110: Obtain an IPv6 packet, which includes an IPv6 extension header, which includes agent semantic information, which includes semantic action type information. The semantic action type information is used to indicate the processing action of the agent to be completed on the data plane of the routing device. The processing action type includes any one of search, call, and state synchronization. Step 120: Perform processing actions for the IPv6 packet according to the semantic action type information.

[0023] Specifically, the agent's semantic information is encoded into the IPv6 extension header using the TLV (Type-Length-Value) format. The value field includes the agent's semantic information, which in turn includes various types of semantically related information, one of which is semantic action type information.

[0024] The forwarding router first receives an IPv6 packet and parses out the IPv6 extension header. Based on the parsed "semantic action type information," the forwarding router knows that the processing operation of the agent associated with the packet should be completed in the data plane (i.e., without the control plane CPU and without the intervention of the software protocol stack).

[0025] In this embodiment, the type of processing operation is set to one of three: search (e.g., finding an agent that meets the conditions), invocation (e.g., triggering an agent to perform a specific function), or state synchronization (e.g., updating the agent's load or context information).

[0026] By placing this type of information in the IPv6 extension header, the router can quickly identify which type of intelligent agent interaction behavior the current packet belongs to at the hardware forwarding level, and then perform corresponding forwarding, address rewriting or table update actions according to predefined rules, without having to send the packet to the control plane for complex parsing.

[0027] In summary, traditional IPv6 routing forwards packets solely based on the destination IP address, failing to recognize whether a packet is related to an agent or differentiate between different types of agent requests. This results in all agent packets being treated the same, and routers cannot differentiate processing based on the semantic intent carried by the packet. This method, through semantic action type information, enables the router to directly identify whether a packet belongs to a search, call, or state synchronization operation at the data plane. This allows for the execution of corresponding forwarding or processing logic for different operation types, achieving semantic awareness and classification of agent requests. Secondly, in existing technologies, agent-related processing often requires sending packets to the control plane CPU for parsing and forwarding decisions. This not only increases processing latency but also easily causes control plane congestion or even crashes in large-scale, high-frequency agent interactions. This method explicitly limits the processing of semantic action type information to the routing device's data plane, eliminating the need to send it to the CPU. This significantly reduces processing latency, avoids control plane overload risks, and improves system scalability and stability. By using the "state synchronization" type in the semantic action type information, a data plane channel is provided for agents to actively report their own state. This enables the router to update the agent's state in real time without relying on the control plane protocol. This ensures that subsequent search and invocation actions can make decisions based on the latest state information, avoiding forwarding requests to unavailable or overloaded agents, and significantly improving the efficiency and reliability of communication between agents.

[0028] In some optional embodiments, the semantic information of the agent may further include semantic action indication information, which is matched with the semantic action type information and is used to indicate the action constraints that need to be referenced when performing processing actions for the IPv6 packet.

[0029] Specifically, this embodiment defines that the semantic information of the intelligent agent also includes a "semantic action indication information". This indication information matches the aforementioned "semantic action type information", that is: when the semantic action type information is "search", the semantic action indication information provides the specific parameters (i.e., action constraints) required for the search operation; when the semantic action type information is "call", it provides the parameters required for the call operation; when the semantic action type information is "state synchronization", it provides the parameters required for the state synchronization operation.

[0030] Semantic action indication information (or action constraints) is carried to the router by the message sender. The router then performs corresponding forwarding, address rewriting, or table update operations on the data plane based on these conditions, thereby achieving refined and differentiated processing of the interactive behaviors of different intelligent agents. In this embodiment, the "message sender" refers to the network entity that constructs and sends the IPv6 message. Depending on the application scenario, the sender can include, but is not limited to, the following: End-user devices: such as applications running on smartphones, personal computers, or IoT terminals. When an application needs to invoke a specific intelligent agent in the network (such as a large language model or image recognition service), it actively constructs an IPv6 message, fills in the "search" or "invoke" semantic action type in the extension header, and attaches the corresponding action indication information (such as the desired intelligent agent type, computing power constraints, etc.), and then sends the message to the local gateway or directly connected router.

[0031] The agent itself: After performing a task, an agent in the network (such as a legal consulting agent or a visual recognition agent) needs to synchronize its latest operating status with the network. At this time, the agent, as the sender, constructs an IPv6 message carrying the "state synchronization" semantic action type, fills in its own operating status, load level, and other action indication information, and sends the message to the forwarding router it is connected to, so that the router can update the agent forwarding table it maintains locally on the data plane.

[0032] Network edge gateways or proxies: In some deployment architectures, user requests do not directly generate IPv6 packets with semantic extension headers. Instead, an edge gateway or agent relay node receives ordinary requests and converts them into IPv6 packet formats conforming to the definition of this scheme. This gateway acts as the packet sender, dynamically generating semantic action types and corresponding action constraints based on the content of the user request, and then injecting the packet into the routing network that supports this scheme.

[0033] In either case, the message sender is aware of the type of agent interaction to be performed (search, call, or state synchronization) and can fill in the necessary constraints for the interaction in a structured form in the "Semantic Action Instruction Information" field of the IPv6 extension header. The forwarding router only needs to read this pre-defined information on the data plane to complete the forwarding or state update according to the rules, without relying on complex protocol negotiation on the control plane.

[0034] In some optional embodiments, when the semantic action type information is a search, the corresponding matching semantic action indication information includes: an agent semantic identifier, used to identify the agent to be searched.

[0035] Specifically, when the semantic action type information is set to "search", the corresponding semantic action indication information includes at least one "agent semantic identifier" field. The core function of this field is to identify the type, capability, or identity characteristics of the target agent that the message sender is searching for, rather than a traditional network layer address (such as an IPv6 address).

[0036] An agent semantic identifier is a descriptor with semantic meaning. Its length can be predefined (e.g., 128 bits). It can take the form of an independent semantic fingerprint or have a deterministic mapping relationship with the agent's IPv6 address. When a forwarding router resolves a semantic action type as "search," it extracts the agent semantic identifier and uses it as a search keyword to match candidate agent entries that meet the criteria in the locally maintained Agent Forwarding Table (AFT). For example, if the value of the agent semantic identifier is "image recognition.low latency," the router will filter out all agent instances that have image recognition capabilities and can provide low-latency responses.

[0037] In some optional embodiments, when the semantic action type information is a search, the corresponding matching semantic action indication information further includes: task context information, which is used to indicate that an agent matching the task context information should be searched first from the downstream agents of the forwarding router.

[0038] Specifically, the semantic action indication information can further include a "task context information" field. The purpose of this field is to provide the forwarding router with an agent filtering capability based on task content characteristics, that is, to determine whether there are downstream agents that have cached data or intermediate states that match the current task context.

[0039] Task context information can be a semantic fingerprint or hash digest of the current task (e.g., a dialogue prompt, an image to be recognized, a set of query parameters, etc.). After parsing this field in the data plane, the forwarding router quickly compares it with the cached fingerprints (e.g., context hash values ​​collected through state synchronization actions) reported and recorded in the local index table by downstream agents. The comparison method can employ bitwise AND operations of Bloom filters, exact matching of hash tables, or similarity calculations.

[0040] If a downstream agent's cached fingerprint matches the currently carried task context information, it indicates that the agent has previously processed the same or highly similar tasks and has stored some or all of the intermediate results needed to complete the task. In this case, the forwarding router can prioritize directing the current search request to that agent, thereby avoiding resubmitting the task to an agent without a cache, which would require recalculation from scratch (such as the pre-filling stage in large model inference), significantly reducing processing latency and improving user experience.

[0041] If no matching agent is found, the router can continue to perform the regular agent search process based on other semantic action indication information (such as agent semantic identifier, computing power requirements, etc.).

[0042] Therefore, the task context information field introduces "content awareness" into the agent's routing decision, enabling the search action to not only be selected based on the agent's static capability attributes, but also to utilize the agent's local cache state to achieve semantic scheduling with content hit awareness. This is one of the key features that distinguishes this solution from traditional service discovery or load balancing mechanisms.

[0043] In some optional embodiments, when the semantic action type information is a search, the corresponding matching semantic action indication information further includes: task computing power requirement information and / or path computing power requirement information; the task computing power requirement information is used to indicate the computing power requirement required for the searched agent to perform the target task; the path computing power requirement information is used to indicate the computing power requirement satisfied by the path of the IPv6 packet route to the searched agent.

[0044] Specifically, when the semantic action type information is "search," the semantic action indication information can further include task computing power requirement information and / or path computing power requirement information. These two fields impose constraints on the search process from the two dimensions of "the target agent's own computing power" and "the computing power conditions of the network transmission path," respectively, thereby enabling the forwarding router to perform computing power-aware agent selection at the data plane.

[0045] Task computational power requirement information: This indicates the minimum or estimated amount of computational resources the message sender expects the target agent to perform the target task. After searching for candidate agents that meet the semantic conditions in the local agent forwarding table (AFT), the forwarding router further checks whether the current available computational power of each candidate agent can meet the task computational power requirement. Only agents with available computational power greater than or equal to the task computational power requirement are included in the final selection. This avoids forwarding requests to agents with insufficient computational power, which could lead to task timeouts or failures, and achieves agent scheduling based on computational resource awareness.

[0046] Path computational power requirement information: This indicates the computational power requirements that the entire transmission path must meet when the IPv6 packet is routed from the current router to the network location of the searched agent. Unlike task computational power requirements, path computational power requirements focus on the computational processing capabilities of the network path itself. This information can be expressed in the form of a path cost ceiling, minimum required link bandwidth, or maximum allowed hop count. When calculating the path from itself to each candidate agent, the forwarding router combines the path computational power requirement information to eliminate paths that do not meet the requirements (e.g., paths with insufficient bandwidth or those passing through routers with limited computational resources), and selects the path with the best overall cost from the remaining feasible paths.

[0047] This embodiment can simultaneously consider whether the absolute computing power of the target agent meets the task requirements and whether the computing power conditions of the network path meet the transmission requirements. For network environments with a large number of heterogeneous agents (such as GPU instances and edge NPU nodes), the computing power-aware search mechanism of this embodiment can significantly improve the success rate of task scheduling and resource utilization, avoid blindly forwarding requests to agents with insufficient execution capabilities or limited network paths, thereby improving the overall service quality.

[0048] In some optional embodiments, when the semantic action type information is a call, the corresponding matching semantic action indication information includes at least one of the following: region identifier, load level information, and task context information: the region identifier is used to indicate the selection of a call path based on the deployment location of the agent; the load level information is used to indicate the selection of a call path based on the current load level of the agent; and the task context information is used to indicate the selection of a call path based on the task context information of the agent.

[0049] Specifically, when the semantic action type information is "invoke," it indicates that the forwarding router already knows the target agent (e.g., it obtained its network address through a previous search action) and needs to perform path selection or rerouting operations on the IPv6 packet to ensure that the packet can reach the target agent in the optimal way. The corresponding matching semantic action indication information includes at least one of the following: area identifier, load level information, and task context information. Each type of information is used to provide different constraints or optimization criteria for path selection.

[0050] A region identifier indicates the current deployment location of the target intelligent agent, such as "device-side" (e.g., user terminal, IoT device), "edge-side" (e.g., edge computing node), or "cloud-side" (e.g., central cloud data center). After reading this identifier, the forwarding router can combine its network topology knowledge to prioritize paths located within the same region as the target intelligent agent. For example, it can forward requests from edge nodes to intelligent agents within the same edge network, avoiding long-distance cross-regional transmissions. Alternatively, it can forcibly avoid certain expensive cross-regional links. The existence of the region identifier allows the calling path to match the physical or logical location of the intelligent agent, reducing transmission latency and bandwidth costs.

[0051] Load level information indicates the current processing load level of the target agent (e.g., CPU / GPU utilization, task queue length, percentage of available computing power). When determining call paths, forwarding routers can use this information to determine if the target agent is overloaded. If the load level exceeds a preset threshold, the router can choose to reroute subsequent requests in the same packet flow to another agent instance with a lower load within the same area, or choose a path through a lightly loaded relay node to avoid exacerbating congestion at the target agent. Load level information helps routers achieve fine-grained load-aware routing.

[0052] Similar to the task context information mentioned above, task context information indicates the task type, quality of service requirements, or associated status of the task currently being performed by the agent. Based on the task context, the forwarding router can dynamically select the transmission path that best suits the current task requirements. For example, for latency-sensitive tasks (such as autonomous driving control commands), the forwarding router will choose the path with the least jitter and shortest latency; for large file transfer tasks, it will choose a path with sufficient bandwidth; for tasks requiring session persistence (multi-turn dialogue), the router will ensure that packets belonging to the same session always travel through the same path to avoid context loss due to path switching.

[0053] By including at least one of the above-mentioned indications, this scheme enables path selection in the invocation action to no longer rely on the minimum number of hops or static configuration of traditional routing protocols, but can be dynamically and finely adjusted according to the agent's real-time location, load status and task characteristics, thereby significantly improving the success rate of agent invocation, resource utilization and user experience.

[0054] In some optional embodiments, when the semantic action type information is state synchronization, the corresponding matching semantic action indication information includes load level information and / or task context information; the load level information is used to indicate that the load level information of the agent is synchronized in the local agent forwarding table; the task context information is used to indicate that the task context information of the agent is synchronized in the local agent forwarding table.

[0055] Specifically, the agent proactively and periodically sends IPv6 packets containing its operational status to the network, thus informing the forwarding routers along the path of its current operational status. This allows the forwarding routers to record or update the agent's dynamic attributes in their local agent forwarding table (AFT). The corresponding semantic action indication information includes load level information and / or task context information. Load level information carries the agent's current processing load level, such as quantitative indicators like CPU / GPU utilization, available memory, task queue depth, or overall computing power idleness. After parsing this information in the data plane, the forwarding router directly extracts its value and locates the entry in its local AFT corresponding to the agent's semantic identifier, updating the "load level" field of that entry to the latest value. In this way, when processing subsequent "search" or "call" requests to that agent, the router can filter agents or optimize paths based on the latest load level, avoiding assigning tasks to overloaded agents.

[0056] Similar to the embodiments described above, task context information carries the characteristic fingerprint of the task context currently being processed by the agent. The forwarding router also extracts this information from the data plane and updates the "context cache fingerprint" information of the corresponding agent entry in the AFT. Subsequently, when the forwarding router receives a "search" request carrying similar task context information, it can quickly determine whether a downstream agent has cached the corresponding context using bitwise AND operations, thereby directing the request to an agent with reusable computation results, significantly reducing redundant computation.

[0057] In this embodiment, the state synchronization operation is performed entirely in the data plane of the forwarding router. After recognizing that the semantic action type is state synchronization, the hardware forwarding engine directly reads the load level information and / or task context information from the IPv6 extension header and uses these values ​​to overwrite the corresponding fields of the corresponding entries in the AFT. The entire process does not generate control plane interruption or send data to the CPU.

[0058] In some optional embodiments, the IPv6 extension header is characterized by being any one of a hop-by-hop option header, a destination option header, and a segmented routing header.

[0059] Specifically, the IPv6 extension header carrying agent semantic information can be any one of the following: Hop-by-Hop Options Header, Destination Options Header, or Segment Routing Header (SRH). All three extension headers are optional headers defined in the IPv6 standard and can utilize agent semantic information in TLV format.

[0060] This embodiment does not limit the use of any of the three extension headers mentioned above, but allows selection based on deployment requirements and network capabilities. Regardless of the extension header chosen, as long as it supports TLV format options, the agent's semantic information (including semantic action types and corresponding action indication information) can be constructed in the same way. This flexibility ensures that this solution can be applied to traditional IPv6 networks (using hop-by-hop option headers or destination option headers) and seamlessly integrated into SRv6 networks (using SRH), exhibiting good compatibility and scalability.

[0061] In some optional embodiments, the forwarding router is characterized in that it is a pre-configured forwarding router with the ability to recognize the semantic information of the intelligent agent; wherein, for forwarding routers that do not have the ability to recognize the semantic information of the intelligent agent, after obtaining the IPv6 packet, they process it according to the standard forwarding rules of IPv6 packets.

[0062] Specifically, the routers involved in forwarding are divided into two categories: one is a forwarding router that is pre-configured with the ability to recognize the semantic information of the intelligent agent (hereinafter referred to as "intelligent agent-aware router"), and the other is an ordinary IPv6 router that does not have this recognition capability (hereinafter referred to as "traditional router").

[0063] For agent-aware routers, parsing logic for the semantic information of the agent is deployed in the data plane hardware or high-speed forwarding engine. When an IPv6 packet carrying agent semantic information is received, such a router can identify the IPv6 extension header and extract semantic action type information and corresponding semantic action indication information according to the rules defined in this scheme, and then perform processing actions such as searching, calling, or state synchronization in the data plane.

[0064] For traditional routers, lacking the capability to recognize the agent semantic information, they process the received IPv6 packets according to standard IPv6 forwarding rules. For example, since agent semantic information options are encoded in TLV format, and their Option Type high-order control bits can be set to "00" (i.e., skipping unrecognized options and continuing processing the rest of the packet), when a traditional router parses an unknown Option Type, it reads the Option Length (Opt Data Len) field, skips the entire agent semantic information option, and then continues forwarding the packet based on the Next Header or the next extension header in the IPv6 base header. The entire process does not discard packets or trigger any error messages.

[0065] This embodiment ensures the compatibility of the solution. Even if a large number of traditional routers are deployed in the network, as long as there is at least one intelligent agent sensing router, the semantic information of the intelligent agent can be identified and processed accordingly at the sensing router. Although the traditional router cannot understand the semantic information, it can still forward the packets normally and will not interrupt communication due to not recognizing the new options.

[0066] Example 2: Based on the above embodiments, this embodiment provides an application example of an IPv6 packet processing method. An Agent Awareness Option (AAO) is defined in the IPv6 Hop-by-Hop Options, Destination Options, or Segment Router Options. Specifically: I. Extended Head Structure like Figure 2 As shown, this is the extended header structure of this embodiment, which adopts the TLV (Type-Length-Value) format and specifically includes the following fields: Option Type (8-bit): TLV type; Opt Data Len (8 bits): TLV length; Semantic Action (16-bit): One of the semantic action indication information, corresponding to the semantic action type information in the above embodiment: three semantic actions, FIND / EXEC / SYNC (search / execute / synchronize). Agent Semantic ID (128bit): One of the semantic action indication information, corresponding to the agent semantic identifier in the above embodiment, which can be an independent semantic fingerprint or correspond to the IPv6 agent address; Compute Cost (16bit): One of the semantic action indication information, corresponding to task computing power requirement information or path computing power requirement information; Load Status (8 bits): One of the semantic action indication information, corresponding to the load level information in the above embodiment, reflecting the current load level of the agent (0~100%). Zone / Loc Info (8 bits): One of the semantic action indication information, corresponding to the area identifier in the above embodiment, such as the terminal / edge / cloud area identifier; Context Vector Hash (several bits): One of the semantic action indication information, corresponding to the task context information in the above embodiment.

[0067] II. Behavioral Description of Semantic Actions FIND (0x01): This message indicates that it is used for agent search. The destination IPv6 address can be a semantic anycast address. The router performs in-network discovery based on the semantic identifier and computing power requirements in the AAO. In the router's FIND process, the router maintains a local index table based on a Bloom filter. When a message carrying a Context Vector Hash is received, a bitwise AND operation is used to quickly determine whether the router's downstream agents can hit the cache. This can greatly reduce the pre-filling time for large model inference, thereby significantly improving the user experience. This is the key effect that distinguishes this semantic addressing from all traditional routing and computing power routing.

[0068] EXEC (0x02): This message indicates that it is used for agent invocation. The destination IPv6 address is the specific instance address, and AAO is used to suggest that the network may optionally optimize computing power.

[0069] SYNC (0x03): This indicates that the agent reports its own state (load, context) to the network, and routers along the route can update the AFT (Agent Forwarding Table) based on this information. It is important to note that the processing of SYNC packets involves directly updating registers or entries in the fast forwarding plane, without sending them to the CPU. This avoids the problem of router control plane crashes caused by a massive number of agents sending SYNC packets.

[0070] III. Router Processing Flow AAO-enabled Agent Forwarding Routers (AFRs) handle packet processing as follows: 1. Parse the IPv6 extension header and identify the AAO; 2. Process according to Semantic Action: FIND: Extract Agent Semantic ID, Compute Cost, and other information from AAO; Search for candidate agents that meet the semantic conditions in the local AFT; Calculate the overall cost (network latency + computing load), select the next hop, and rewrite the destination address if necessary.

[0071] EXEC: Fine-grained routing or rerouting is performed on the path based on Load Status / Zone / Context.

[0072] SYNC: Update the Load Status and Context Hash of the corresponding agent entry recorded in AFT.

[0073] 3. If it is a non-AFR node, unknown Option Types will be transparently transmitted or ignored according to IPv6 standard rules to ensure compatibility.

[0074] Example 3: Another embodiment of this application relates to a forwarding router. The implementation details of this forwarding router are described below. The following content is for ease of understanding and is not essential for implementing this solution. A schematic diagram of the forwarding router in this embodiment can be shown as follows: Figure 3 As shown, it includes an acquisition module 310 and an execution module 320.

[0075] The acquisition module 310 is used to acquire IPv6 packets, the IPv6 packets including IPv6 extension headers, the IPv6 extension headers including agent semantic information, the agent semantic information including semantic action type information, the semantic action type information being used to indicate the completion of processing actions on the agent in the data plane of the routing device, the processing action type including any one of search, call path selection and state synchronization; The execution module 320 is used to perform processing actions on the IPv6 packet according to the semantic action type information.

[0076] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed in this application; however, this does not mean that other units are absent in this embodiment.

[0077] In some alternative embodiments, the forwarding router can implement the IPv6 packet processing method described in any of the above embodiments.

[0078] Example 4: Another embodiment of this application relates to an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the IPv6 packet processing methods in the above embodiments.

[0079] In this embodiment, the memory and processor are connected via a bus, which can include any number of interconnected buses and bridges, connecting various circuits of one or more processors and the memory together. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be further described in this embodiment. A bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.

[0080] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0081] Example 5: Another embodiment of this application relates to a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the method embodiments described above.

[0082] That is, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0083] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing this application, and in practical applications, various changes can be made to them in form and detail without departing from the spirit and scope of this application.

Claims

1. A method for processing IPv6 packets, characterized in that, Applied to forwarding routers, including: Obtain an IPv6 packet, the IPv6 packet including an IPv6 extension header, the IPv6 extension header including agent semantic information, the agent semantic information including semantic action type information, the semantic action type information being used to indicate the completion of processing actions on the agent in the data plane of the routing device, the processing action type including any one of search, call, and state synchronization; Based on the semantic action type information, perform the processing action for the IPv6 packet.

2. The method for processing IPv6 packets according to claim 1, characterized in that, The semantic information of the agent also includes semantic action indication information, which is matched with the semantic action type information and is used to indicate the action constraints that need to be referenced when performing processing actions for the IPv6 packet.

3. The method for processing IPv6 packets according to claim 2, characterized in that, When the semantic action type information is a search, the corresponding matching semantic action indication information includes: Intelligent agent semantic identifiers are used to identify the intelligent agent to be searched.

4. The method for processing IPv6 packets according to claim 3, characterized in that, When the semantic action type information is a search, the corresponding matching semantic action indication information also includes: Task context information is used to indicate that the search should be prioritized for agents that match the task context information from the downstream agents of the forwarding router.

5. The method for processing IPv6 packets according to claim 3, characterized in that, When the semantic action type information is a search, the corresponding matching semantic action indication information also includes: task computing power requirement information and / or path computing power requirement information; The task computing power requirement information is used to indicate the computing power requirement required for the searched intelligent agent to perform the target task; The path computing power requirement information is used to indicate the computing power requirements that the path for routing the IPv6 packet to the searched agent must meet.

6. The method for processing IPv6 packets according to claim 2, characterized in that, When the semantic action type information is a call, the corresponding matching semantic action indication information includes at least one of the following: region identifier, load level information, and task context information: The region identifier is used to indicate the call path selection based on the deployment location of the agent; The load level information is used to indicate the call path to be selected based on the agent's current load level; The task context information is used to instruct the agent-based task context information to select the invocation path.

7. The method for processing IPv6 packets according to claim 2, characterized in that, When the semantic action type information is state synchronization, the corresponding matching semantic action indication information includes load level information and / or task context information; The load level information is used to indicate the state synchronization of the agent's load level information in the local agent forwarding table; The task context information is used to instruct the agent to synchronize its state in the local agent forwarding table.

8. The method for processing IPv6 packets according to any one of claims 1-7, characterized in that, The IPv6 extension header is any one of the following: hop-by-hop option header, destination option header, and segmented routing header.

9. The method for processing IPv6 packets according to any one of claims 1-7, characterized in that, The forwarding router is a pre-configured forwarding router with the ability to recognize the semantic information of the intelligent agent; Specifically, for forwarding routers that do not have the ability to recognize the semantic information of the intelligent agent, after obtaining the IPv6 packet, they process it according to the standard forwarding rules of IPv6 packets.

10. A forwarding router, characterized in that, include: The acquisition module is used to acquire IPv6 packets, which include IPv6 extension headers, agent semantic information, and semantic action type information. The semantic action type information is used to indicate the processing action of the agent to be completed in the data plane of the routing device. The processing action type includes any one of search, call path selection, and state synchronization. The execution module is used to perform processing actions on the IPv6 packet according to the semantic action type information.

11. An electronic device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the IPv6 packet processing method as described in any one of claims 1 to 9.

Images